CAPAN — Hu63 vs Iso — 2026-09-10

CAPAN
MTT
Hu63 vs Iso
Published

September 10, 2026

IC50 & Growth Kinetics — Capan Conjugates (Castor + MTT combined)

Two readouts are analysed side by side in one notebook:

Assay Plates Raw measurement Line style
Castor 1, 2 image-based confluence (fraction) dashed
MTT 1, 2 MTT absorbance (stored in the same Confluence column) solid

Plate numbers repeat across assays — Castor P1 and MTT P1 are different physical plates, so every readout file, every layout and every grouping below is keyed on (Assay, Plate), never on the plate number alone.

File naming this notebook expects in DATA_DIR:

  • readout data — <ASSAY>-<anything>_P<plate>_D<day>.csv (e.g. CASTOR-…_P1_D7.csv, MTT-…_P1_D7.csv)
  • plate layouts — Plate<plate>_layout_<Assay>.csv (an assay-less Plate<plate>_layout.csv is applied to every assay)

The same drug measured both ways stays two separate series: normalization, the reversal filter, LL.4 fits and every IC50 are computed within an assay.

Plot contract: colour = drug (identical across assays), linestyle/marker = assay.

Raw metric: Day 7 only — no Day-0 normalization. Responses are normalized to the untreated wells of the same assay/plate/drug (%Viable), which also puts the two assays’ different raw units on one comparable scale. IC50 computed two ways: fitted LL.4 plateau midpoint and absolute 50% of untreated by direct interpolation.

Show code
from google.colab import drive
drive.mount('/content/drive')
Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount("/content/drive", force_remount=True).
Show code
# ====================================================================================
#  USER INPUT — Navigate to experiment folder shortcut; copy relative path to DATA_DIR
# ====================================================================================
DATA_DIR  = '/content/drive/MyDrive/20260827 Cytotox - Hu64_VPE6_GE6/Capan Combine Analysis-20260910/Data'
Show code
from IPython.display import display, Markdown, Image as IPImage
import os
import re
import pandas as pd
import numpy as np
from scipy.optimize import curve_fit
import matplotlib
import matplotlib.pyplot as plt
from matplotlib.lines import Line2D
from matplotlib.patches import Patch
from matplotlib.legend_handler import HandlerTuple
import seaborn as sns
import warnings
warnings.filterwarnings('ignore')

sns.set_style('whitegrid')
plt.rcParams.update({'figure.figsize': (10, 6), 'font.size': 12})

# =====================================================================
#  USER INPUT — change only this section to run a different experiment
# =====================================================================
CELL_LINE    = 'Capan' # cell line name shown in plot titles
TEST_AB      = 'Hu63'    # test article antibody prefix
ISO_AB       = 'Iso'   # isotype control antibody prefix
R2_THRESHOLD = 0.5     # minimum R² for a "reliable" LL.4 fit

SCALE_MIN_CONC = 0         # dose-response x-axis lower bound (0 = untreated pseudo-conc)
SCALE_MAX_CONC = None       # None uses all concentrations retained by the reversal filter
SCALE_EXCL_ABS = []         # all retained drugs contribute to y-axis scaling
REVERSAL_FILTER_METRIC = 'Confluence'     # raw metric used to detect fragmentation artefacts
REVERSAL_INCREASE_THRESHOLD = 0.25        # exclude if >25% above the preceding concentration

# --- USER INPUT: assay merge (multi-readout experiments) -------------------
# Each readout CSV is tagged with an assay label so the same drug measured by two
# different readouts stays two separate series and is never pooled. Plate numbers
# may repeat across assays (Castor P1 and MTT P1 are different plates), so every
# grouping downstream is keyed on (Assay, Plate).
#
#   readout data  ->  <ASSAY>-<anything>_P<plate>_D<day>.csv
#   plate layouts ->  Plate<plate>_layout_<Assay>.csv
#                     (assay-less Plate<plate>_layout.csv applies to every assay)
#
# Assay resolution per readout file: FILE_ASSAY -> ASSAY_PATTERNS -> DEFAULT_ASSAY.
ASSAY_PATTERNS = [
    (r'^CASTOR[-_ ]', 'Castor'),   # Castor image cell counting
    (r'^MTT[-_ ]',    'MTT'),      # MTT chemistry
]
FILE_ASSAY    = {}          # per-file override, e.g. {'oddly_named_P1_D7.csv': 'MTT'}
DEFAULT_ASSAY = 'Castor'    # for files matching no pattern and no override

# Plot style contract: COLOUR = drug (identical across assays),
#                      LINESTYLE / MARKER / HATCH = assay.
# linestyle/linewidth/marker are the on-plot cue; hatch is the bar-chart cue.
# Castor's dash is (0, (6, 2)) rather than '--': long enough to stay obvious
# where curves overlap, and in a short legend sample.
ASSAY_STYLES = {
    'Castor': {'linestyle': (0, (6, 2)), 'linewidth': 2.2, 'marker': 'o', 'hatch': '//'},
    'MTT':    {'linestyle': '-',         'linewidth': 1.7, 'marker': 's', 'hatch': ''},
}

# Plots of the RAW metric (growth kinetics, dose-response) show quantities in
# different units — Castor confluence vs MTT absorbance. Sharing one y-axis
# pushes them into separate bands, so only their x-positions can be compared.
# The assay named here is drawn against a twin right-hand y-axis on its own
# scale: each readout then fills the full plot height and the two sets of
# points/curves roughly overlap, comparable by shape. Presentation only — fits,
# IC50s and %Viable are computed per assay and are unaffected.
# None puts every assay back on a single left axis.
RAW_RIGHT_AXIS_ASSAY = 'MTT'
# Per-assay label for that axis; unlisted assays fall back to '<assay> raw metric'.
ASSAY_RAW_YLABEL = {
    'Castor': 'Castor confluence (fraction)',
    'MTT':    'MTT absorbance (a.u.)',
}
# =====================================================================
PLOT_DIR  = os.path.join(DATA_DIR, 'plots')
# Stem for exported CSVs — the experiment folder name (DATA_DIR is often '.').
# If DATA_DIR is a generic 'Data' wrapper, name exports after the experiment
# folder above it (e.g. .../LS180/Data -> 'LS180') rather than after 'Data'.
_stem_dir = os.path.abspath(DATA_DIR)
if os.path.basename(_stem_dir).lower() == 'data':
    _stem_dir = os.path.dirname(_stem_dir)
EXPORT_STEM = os.path.basename(_stem_dir)
os.makedirs(PLOT_DIR, exist_ok=True)
plot_counter = [0]

def save_fig(title=''):
    plot_counter[0] += 1
    fname = os.path.join(PLOT_DIR, f'plot_{plot_counter[0]:02d}.png')
    plt.savefig(fname, dpi=120, bbox_inches='tight')
    plt.close()
    display(IPImage(filename=fname))

# Auto-discover data: readout CSVs in DATA_DIR ($ anchor keeps Windows
# ':Zone.Identifier' sidecar files out).
_fname_pat = re.compile(r'.*_P(?:late)?(\d+)_D(\d+)\.csv$', re.IGNORECASE)
csv_data_files = sorted([f for f in os.listdir(DATA_DIR) if _fname_pat.match(f)])

# Plate layout CSVs live in DATA_DIR; group 2 is the optional assay suffix.
_layout_pat = re.compile(r'Plate(\d+)_layout(?:_([A-Za-z0-9]+))?\.csv$', re.IGNORECASE)
PLATE_LAYOUTS = sorted([f for f in os.listdir(DATA_DIR) if _layout_pat.match(f)])
assert len(PLATE_LAYOUTS) >= 1, f'No Plate*_layout*.csv found in {DATA_DIR}'

print(f'Data dir : {DATA_DIR}')
print(f'Readouts : {len(csv_data_files)} data file(s)')
print(f'Layouts  : {len(PLATE_LAYOUTS)} file(s) — {PLATE_LAYOUTS}')
print(f'X-axis   : {SCALE_MIN_CONC} – {SCALE_MAX_CONC if SCALE_MAX_CONC is not None else "all retained concentrations"}')
print(f'Excl ABs : {SCALE_EXCL_ABS} (excluded from y-axis range only)')
print(f'Assays   : {[lbl for _, lbl in ASSAY_PATTERNS]} (default {DEFAULT_ASSAY})')
Data dir : /content/drive/MyDrive/20260827 Cytotox - Hu64_VPE6_GE6/Capan Combine Analysis-20260910/Data
Readouts : 4 data file(s)
Layouts  : 4 file(s) — ['Plate1_layout_Castor.csv', 'Plate1_layout_MTT.csv', 'Plate2_layout_Castor.csv', 'Plate2_layout_MTT.csv']
X-axis   : 0 – all retained concentrations
Excl ABs : [] (excluded from y-axis range only)
Assays   : ['Castor', 'MTT'] (default Castor)

Load Readout Data for all Plate(s) — tagged & merged by assay

Show code
# ===========================================================================
# 1. Load readout data, tag each file with its assay, merge by assay
#    Key is (Assay, Plate, Day). Plate numbers repeat across assays, so the
#    assay is what separates Castor P1 from MTT P1; a repeated key is skipped,
#    never silently averaged in.
# ===========================================================================
print('\n=== 1. Load readout data (merged by assay) ===')
csv_files = sorted([f for f in os.listdir(DATA_DIR) if f.lower().endswith('.csv')])
print(f'{len(csv_files)} CSV files found')


def assay_from_filename(fname):
    """Assay label implied by a readout filename, or None if no pattern matches."""
    for pat, label in ASSAY_PATTERNS:
        if re.search(pat, fname, re.IGNORECASE):
            return label
    return None


frames = []
seen = {}
for fname in csv_files:
    m = _fname_pat.match(fname)
    if not m or _layout_pat.match(fname):
        continue
    plate_num = int(m.group(1))
    day       = int(m.group(2))
    assay     = FILE_ASSAY.get(fname) or assay_from_filename(fname) or DEFAULT_ASSAY
    key = (assay, plate_num, day)
    if key in seen:
        print(f'  SKIP duplicate {assay} P{plate_num} D{day}: {fname} '
              f'(already loaded {seen[key]})')
        continue
    seen[key] = fname
    # utf-8-sig: the MTT exports carry a BOM on the first header cell.
    tmp = pd.read_csv(os.path.join(DATA_DIR, fname), index_col=0, encoding='utf-8-sig')
    tmp.index.name = 'Well'
    tmp = tmp.reset_index()
    tmp['Plate'] = plate_num
    tmp['Day']   = day
    tmp['Assay'] = assay
    frames.append(tmp)
    print(f'  {assay:<8} P{plate_num} D{day}  <-  {fname}')

assert frames, f'No readout CSVs matched {_fname_pat.pattern} in {DATA_DIR}'
df_raw = pd.concat(frames, ignore_index=True)
df_raw = df_raw.drop(columns=['Total cell area'], errors='ignore')

ASSAYS = sorted(df_raw.Assay.unique())
ASSAY_PLATES = {a: sorted(df_raw.loc[df_raw.Assay == a, 'Plate'].unique().tolist())
                for a in ASSAYS}

print(f'\nCombined raw data: {df_raw.shape[0]} rows  x  {df_raw.shape[1]} cols')
print(f'Assays: {ASSAYS}')
for _a in ASSAYS:
    _sub = df_raw[df_raw.Assay == _a]
    print(f'  {_a:<8} plates {ASSAY_PLATES[_a]}   '
          f'days {sorted(_sub.Day.unique().tolist())}   ({len(_sub)} wells)')
_shared = sorted(set(ASSAY_PLATES[ASSAYS[0]]).intersection(
    *[set(ASSAY_PLATES[a]) for a in ASSAYS[1:]])) if len(ASSAYS) > 1 else []
if _shared:
    print(f'  NOTE: plate number(s) {_shared} are reused across assays — '
          f'kept distinct by (Assay, Plate).')

=== 1. Load readout data (merged by assay) ===
12 CSV files found
  Castor   P1 D7  <-  CASTOR-20260827_PCM_SH_Capan_63GE8_IsoGE8_P1_D7.csv
  Castor   P2 D7  <-  CASTOR-PCM_SH_20260827_Capan_64GE6-64VPE6_P2_D7.csv
  MTT      P1 D7  <-  MTT-PCM_SH_Capan_20260827_63GE8-IsoGE8_P1_D7.csv
  MTT      P2 D7  <-  MTT-PCM_SH_Capan_20260827_64GE6-64VPE6_P2_D7.csv

Combined raw data: 240 rows  x  5 cols
Assays: ['Castor', 'MTT']
  Castor   plates [1, 2]   days [7]   (120 wells)
  MTT      plates [1, 2]   days [7]   (120 wells)
  NOTE: plate number(s) [1, 2] are reused across assays — kept distinct by (Assay, Plate).

Merge Plate Layouts

Show code
# ===========================================================================
# 2. Load plate layouts and merge on (Assay, Plate, Well)
#    Plate<n>_layout_<Assay>.csv binds a layout to one assay; an assay-less
#    Plate<n>_layout.csv is broadcast to every assay in the data.
# ===========================================================================
print('\n=== 2. Load plate layouts and merge ===')
layout_frames = []
for lf in PLATE_LAYOUTS:
    m = _layout_pat.match(lf)
    plate_num    = int(m.group(1))
    layout_assay = m.group(2)          # None -> applies to every assay

    if layout_assay is None:
        targets = list(ASSAYS)
    else:
        targets = [a for a in ASSAYS if a.lower() == layout_assay.lower()]
        if not targets:
            print(f'  WARNING: {lf} names assay {layout_assay!r}, which no readout '
                  f'file uses ({ASSAYS}) — layout ignored')
            continue

    tmp = pd.read_csv(os.path.join(DATA_DIR, lf), encoding='utf-8-sig')
    tmp.rename(columns={'Cell': 'Well'}, inplace=True)
    tmp['Plate'] = plate_num
    for a in targets:
        layout_frames.append(tmp.assign(Assay=a))
    print(f'  P{plate_num} [{", ".join(targets)}] drugs: {sorted(tmp.Drug.unique())} '
          f'<-  {lf}')

assert layout_frames, 'No plate layout matched an assay present in the readout data'
layouts = pd.concat(layout_frames, ignore_index=True)

_dups = layouts.duplicated(subset=['Assay', 'Plate', 'Well']).sum()
assert _dups == 0, (f'{_dups} duplicate (Assay, Plate, Well) layout entries — two '
                    f'layout files claim the same assay/plate')
print(f'Layout entries: {len(layouts)}  ({len(PLATE_LAYOUTS)} file(s), '
      f'{layouts.groupby(["Assay", "Plate"]).ngroups} assay×plate combination(s))')

df = df_raw.merge(layouts[['Assay', 'Plate', 'Well', 'Drug', 'Conc']],
                  on=['Assay', 'Plate', 'Well'], how='left')

unmatched = df.Drug.isna().sum()
if unmatched > 0:
    print(f'WARNING: {unmatched} wells did not match a layout entry')
    display(df[df.Drug.isna()].groupby(['Assay', 'Plate']).size()
              .rename('unmatched wells').to_frame())
else:
    print('All wells matched their layout.')

print(f'Drugs: {sorted(df.Drug.dropna().unique().tolist())}')
print(f'Concentrations: {sorted(df.Conc.dropna().unique().tolist())}')

# Drug x Assay coverage — every populated cell here becomes its own curve.
display(Markdown('### Wells per Drug × Assay — each pair is a separate series'))
display(pd.crosstab(df.Drug, df.Assay))

# Auto-derive concentration parameters from the data
nonzero_concs = df.loc[df.Conc > 0, 'Conc']
CONC_MIN   = nonzero_concs.min()
CONC_MAX   = nonzero_concs.max()
PSEUDO_CONC = CONC_MIN / 10.0

df['Conc_orig'] = df['Conc'].copy()
df.loc[df.Conc == 0, 'Conc'] = PSEUDO_CONC
print(f'Conc range: {CONC_MIN:.4g} – {CONC_MAX:.4g} nM')
print(f'Conc==0 replaced with pseudo-conc {PSEUDO_CONC:.4g} nM')

=== 2. Load plate layouts and merge ===
  P1 [Castor] drugs: ['63GE8', 'IsoGE8'] <-  Plate1_layout_Castor.csv
  P1 [MTT] drugs: ['63GE8', 'IsoGE8'] <-  Plate1_layout_MTT.csv
  P2 [Castor] drugs: ['64GE6', '64VPE6'] <-  Plate2_layout_Castor.csv
  P2 [MTT] drugs: ['64GE6', '64VPE6'] <-  Plate2_layout_MTT.csv
Layout entries: 240  (4 file(s), 4 assay×plate combination(s))
All wells matched their layout.
Drugs: ['63GE8', '64GE6', '64VPE6', 'IsoGE8']
Concentrations: [0.0, 0.006, 0.026, 0.12, 0.54, 2.44, 11.0, 49.4, 222.2, 1000.0]

Wells per Drug × Assay — each pair is a separate series

Assay Castor MTT
Drug
63GE8 30 30
64GE6 30 30
64VPE6 30 30
IsoGE8 30 30
Conc range: 0.006 – 1000 nM
Conc==0 replaced with pseudo-conc 0.0006 nM

Raw Metric QC & Fragmentation Reversal Filter

Show code
# ===========================================================================
# 3. Raw metric QC (no Day-0 normalization)
#    The raw metric is used as measured. Day-0 confluence is NOT used as a
#    per-well baseline; all dose-response normalization is done against the
#    untreated wells of the same assay/plate/drug/day (%Viable, sec. 7b).
#    NOTE the raw units differ between assays (Castor = confluence fraction,
#    MTT = absorbance) — they are only comparable once normalized to %Viable.
# ===========================================================================
print('\n=== 3. Raw metric QC (no Day-0 normalization) ===')

if 0 in set(df.Day.unique()):
    print('NOTE: a Day 0 scan is present but is NOT used as a baseline. '
          'It is treated as an ordinary timepoint and excluded from curve fitting.')
else:
    print('No Day 0 scan present — none required.')

display(Markdown('### Raw metric stats by Assay and Day:'))
display(df.groupby(['Assay', 'Day'])['Confluence'].describe().round(3))
display(Markdown('### Raw metric stats by Assay, Plate and Day:'))
display(df.groupby(['Assay', 'Plate', 'Day'])['Confluence'].describe().round(3))

# Drug-by-drug fragmentation filter, run independently within each assay.
# Compare the mean raw metric at each concentration with the immediately
# preceding concentration for that assay/drug/day.
_filter_means = (
    df.groupby(['Assay', 'Drug', 'Day', 'Conc'], dropna=False)[REVERSAL_FILTER_METRIC]
    .mean().rename('Reversal_Filter_Mean').reset_index()
    .sort_values(['Assay', 'Drug', 'Day', 'Conc'])
)
_filter_means['Reversal_Previous_Mean'] = (
    _filter_means.groupby(['Assay', 'Drug', 'Day'])['Reversal_Filter_Mean'].shift(1)
)
_prev = _filter_means['Reversal_Previous_Mean']
_curr = _filter_means['Reversal_Filter_Mean']
_filter_means['Reversal_Pct_Increase'] = np.where(
    _prev > 0, (_curr - _prev) / _prev,
    np.where((_prev == 0) & (_curr > 0), np.inf, np.nan)
)
_filter_means['Exclude_25pct_Reversal'] = (
    _filter_means['Reversal_Pct_Increase'] > REVERSAL_INCREASE_THRESHOLD
)
df = df.drop(columns=['Reversal_Filter_Mean', 'Reversal_Previous_Mean',
                      'Reversal_Pct_Increase', 'Exclude_25pct_Reversal'],
             errors='ignore')
df = df.merge(
    _filter_means[['Assay', 'Drug', 'Day', 'Conc', 'Reversal_Filter_Mean',
                   'Reversal_Previous_Mean', 'Reversal_Pct_Increase',
                   'Exclude_25pct_Reversal']],
    on=['Assay', 'Drug', 'Day', 'Conc'], how='left'
)
df['Exclude_25pct_Reversal'] = df['Exclude_25pct_Reversal'].fillna(False).astype(bool)

excluded_reversals = _filter_means[_filter_means.Exclude_25pct_Reversal].copy()
print(f'\n25% reversal filter metric: {REVERSAL_FILTER_METRIC} (applied within each assay)')
print(f'Excluded assay/drug/day/concentration groups: {len(excluded_reversals)}')
if not excluded_reversals.empty:
    display(excluded_reversals[['Assay', 'Drug', 'Day', 'Conc', 'Reversal_Previous_Mean',
                                'Reversal_Filter_Mean', 'Reversal_Pct_Increase']]
            .round(3).style.hide(axis='index'))

=== 3. Raw metric QC (no Day-0 normalization) ===
No Day 0 scan present — none required.

Raw metric stats by Assay and Day:

count mean std min 25% 50% 75% max
Assay Day
Castor 7 120.0 0.082 0.038 0.031 0.057 0.077 0.103 0.338
MTT 7 120.0 0.121 0.041 0.060 0.085 0.116 0.153 0.206

Raw metric stats by Assay, Plate and Day:

count mean std min 25% 50% 75% max
Assay Plate Day
Castor 1 7 60.0 0.090 0.042 0.038 0.064 0.087 0.101 0.338
2 7 60.0 0.074 0.033 0.031 0.046 0.064 0.106 0.147
MTT 1 7 60.0 0.127 0.040 0.061 0.099 0.129 0.157 0.206
2 7 60.0 0.115 0.041 0.060 0.077 0.107 0.149 0.204

25% reversal filter metric: Confluence (applied within each assay)
Excluded assay/drug/day/concentration groups: 4
Assay Drug Day Conc Reversal_Previous_Mean Reversal_Filter_Mean Reversal_Pct_Increase
Castor 63GE8 7 222.200000 0.053000 0.080000 0.494000
Castor 63GE8 7 1000.000000 0.080000 0.200000 1.497000
Castor IsoGE8 7 0.006000 0.081000 0.107000 0.323000
Castor IsoGE8 7 1000.000000 0.055000 0.076000 0.387000

Replicate Variance & Outlier Wells

Show code
# ===========================================================================
# 3b. Replicate variance & outlier well detection — per Drug × Concentration
#     Plate layout: replicates run DOWN a column. The three wells of one
#     triplicate share a column and differ only by plate ROW (e.g. E2/F2/G2),
#     while concentration varies ACROSS columns (col 2 = top dose ... col 11 =
#     untreated). So one replicate group is (Assay, Plate, Drug, Day, Conc) and
#     its members are distinguished by row.
#
#     The replicate axis is detected from the data rather than assumed, so this
#     still works if a plate is ever laid out the other way round.
#
#     Run on the RAW metric, before any normalization: Castor confluence and MTT
#     absorbance are assessed separately and never compared, since their scales
#     and their noise floors are completely different.
# ===========================================================================
print('\n=== 3b. Replicate variance & outlier wells (raw metric, per assay) ===')

# --- USER INPUT: what counts as a suspect replicate ------------------------
REPLICATE_CV_THRESHOLD = 25.0   # % CV above which a triplicate is flagged
REPLICATE_ROW_BIAS_PCT = 15.0   # mean % deviation above which a whole replicate
                                # position (plate row) is called systematically off
# ---------------------------------------------------------------------------

_REP_KEY = ['Assay', 'Plate', 'Drug', 'Day', 'Conc']
_rep_metric = 'Confluence'      # the raw column; units differ per assay

_well_pat = re.compile(r'^\s*([A-Za-z]+)\s*(\d+)\s*$')


def _well_row(w):
    m = _well_pat.match(str(w))
    return m.group(1).upper() if m else np.nan


def _well_col(w):
    m = _well_pat.match(str(w))
    return int(m.group(2)) if m else np.nan


def _cv_pct(values):
    """Coefficient of variation (%) — scale-free, so Castor and MTT are comparable."""
    v = np.asarray(values, dtype=float)
    v = v[np.isfinite(v)]
    if len(v) < 2:
        return np.nan
    m = v.mean()
    if m == 0:
        return np.nan
    return 100.0 * v.std(ddof=1) / m


_rep = df.dropna(subset=[_rep_metric]).copy()
_rep['Row'] = _rep.Well.map(_well_row)
_rep['Col'] = _rep.Well.map(_well_col)

# Deviation of each well from the median of its OWN triplicate. The median is
# used rather than the mean because with n = 3 a single bad well drags the mean
# toward itself and masks its own deviation.
_grp_median = _rep.groupby(_REP_KEY, dropna=False)[_rep_metric].transform('median')
_rep['Rep_Median'] = _grp_median
_rep['Dev_pct'] = np.where(_grp_median != 0,
                           100.0 * (_rep[_rep_metric] - _grp_median) / _grp_median,
                           np.nan)

# Which plate axis separates the members of a replicate group?
_axis_var = _rep.groupby(_REP_KEY, dropna=False)[['Row', 'Col']].nunique()
_rep_axis = ('Row' if _axis_var['Row'].mean() >= _axis_var['Col'].mean() else 'Col')
_other_axis = 'Col' if _rep_axis == 'Row' else 'Row'
print(f'  Replicate axis detected: members of a triplicate differ by {_rep_axis} '
      f'(constant {_other_axis}); {_other_axis} carries concentration.')

# --- Per-group statistics ---------------------------------------------------
_rep_rows = []
for _key, _g in _rep.groupby(_REP_KEY, dropna=False):
    _v = _g[_rep_metric].to_numpy(dtype=float)
    _w = _g.Well.to_numpy()
    _pos = _g[_rep_axis].to_numpy()
    _med = float(np.median(_v))
    _cv = _cv_pct(_v)

    # The well furthest from the triplicate median is the candidate culprit.
    _i = int(np.argmax(np.abs(_v - _med)))
    _keep = [j for j in range(len(_v)) if j != _i]
    _cv_drop = _cv_pct(_v[_keep]) if len(_keep) >= 2 else np.nan

    # Three-way verdict. "one suspect well" means the remaining replicates agree
    # once that well is removed — i.e. a single well is to blame. "scattered"
    # means the whole group disagrees and no single well explains it.
    if not np.isfinite(_cv) or _cv <= REPLICATE_CV_THRESHOLD:
        _verdict, _suspect = 'OK', ''
    elif np.isfinite(_cv_drop) and _cv_drop <= REPLICATE_CV_THRESHOLD:
        _verdict, _suspect = 'one suspect well', _w[_i]
    else:
        _verdict, _suspect = 'scattered', ''

    _others = ' / '.join(f'{x:.4g}' for x in _v[_keep])

    # Plate-layout coordinates, so the table can be read against the physical
    # plate: each drug occupies a contiguous block of replicate positions
    # (rows B-D, E-G, ...) and each group sits in one concentration position
    # (column). Sorting on these reproduces plate reading order.
    _pos_sorted = sorted(str(p) for p in set(_pos))
    _conc_pos = _g[_other_axis].dropna()
    _conc_pos = _conc_pos.iloc[0] if len(_conc_pos) else np.nan

    _rep_rows.append(dict(zip(_REP_KEY, _key)) | {
        'Wells': '/'.join(sorted(map(str, _w))),
        'Block': _pos_sorted[0] if _pos_sorted else '',
        'Conc_Pos': _conc_pos,
        'N': len(_v), 'Mean': _v.mean(), 'SD': _v.std(ddof=1) if len(_v) > 1 else np.nan,
        'CV_pct': _cv, 'Min': _v.min(), 'Max': _v.max(),
        'Worst_Well': _w[_i], 'Worst_Pos': _pos[_i],
        'Worst_Value': _v[_i], 'Worst_Dev_pct': 100.0 * (_v[_i] - _med) / _med if _med else np.nan,
        'Other_Reps': _others, 'CV_pct_drop1': _cv_drop,
        'Verdict': _verdict, 'Suspect_Well': _suspect,
        'Untreated': bool((_g.Conc_orig == 0).all()),
        'Reversal_Excluded': bool(_g.Exclude_25pct_Reversal.any()),
    })

# Canonical order = plate reading order: assay, plate, drug's replicate block
# (top block first), then across the concentration axis (column 2 -> 11, i.e.
# top dose -> untreated). Views that need worst-first re-sort explicitly.
df_replicate_qc = pd.DataFrame(_rep_rows).sort_values(
    ['Assay', 'Plate', 'Block', 'Conc_Pos']).reset_index(drop=True)

# --- Systematic bias along the replicate axis -------------------------------
# Only the replicate axis is interpretable here: the other axis IS concentration,
# so a "bias" along it would just be the dose-response curve.
_pos_bias = (_rep.groupby(['Assay', 'Plate', _rep_axis])['Dev_pct']
             .mean().rename('Mean_Dev_pct').reset_index())
_pos_bias['Systematic'] = _pos_bias.Mean_Dev_pct.abs() >= REPLICATE_ROW_BIAS_PCT
_bias_lookup = {(a, p, r): m for a, p, r, m, _s in _pos_bias.itertuples(index=False)}


def _likely_cause(row):
    """Distinguish a one-off bad well from a whole replicate position being off."""
    if not row.Suspect_Well:
        return ''
    _b = _bias_lookup.get((row.Assay, row.Plate, row.Worst_Pos), np.nan)
    if np.isfinite(_b) and abs(_b) >= REPLICATE_ROW_BIAS_PCT:
        return f'{_rep_axis} {row.Worst_Pos} biased ({_b:+.0f}% plate-wide)'
    return 'isolated well'


df_replicate_qc['Likely_Cause'] = df_replicate_qc.apply(_likely_cause, axis=1)

_sig4 = lambda x: f'{x:.4g}'

# Plate banding: alternating black / grey down the Plate column, so consecutive
# plate blocks are separable at a glance when scanning the table.
# Keyed on the plate number itself (odd = black, even = grey) rather than on row
# position, so a given plate keeps the same shade in every table.
# NB: Plate is kept as a DATA column rather than an index level — pandas
# sparsifies repeated MultiIndex labels and `Styler.map_index` then emits CSS
# for only the first row of each run, so index-level banding does not render.
_PLATE_BAND = {0: 'background-color: #757575; color: #ffffff;',   # even plates: grey
               1: 'background-color: #1f1f1f; color: #ffffff;'}   # odd plates:  black


def _plate_band(v):
    try:
        return _PLATE_BAND[int(v) % 2]
    except (TypeError, ValueError):
        return ''

# --- Table 1: per-assay overview -------------------------------------------
_ov = (df_replicate_qc.groupby('Assay')
       .agg(Groups=('CV_pct', 'size'),
            Median_CV=('CV_pct', 'median'),
            P90_CV=('CV_pct', lambda s: s.quantile(0.90)),
            Max_CV=('CV_pct', 'max'),
            Flagged=('Verdict', lambda s: int((s != 'OK').sum())),
            Suspect_Well=('Verdict', lambda s: int((s == 'one suspect well').sum())),
            Scattered=('Verdict', lambda s: int((s == 'scattered').sum()))))
_ov.insert(0, 'Raw metric', [ASSAY_RAW_YLABEL.get(a, 'raw metric') for a in _ov.index])

display(Markdown('### Replicate variance overview — per assay'))
display(Markdown(
    f'Replicate group = one triplicate: `{"`, `".join(_REP_KEY)}`. '
    f'A group is flagged when its CV exceeds **{REPLICATE_CV_THRESHOLD:.0f}%**. '
    'Castor and MTT are summarised separately — their raw units and noise floors '
    'are not comparable.'))
display(_ov.style.format({'Median_CV': '{:.1f}', 'P90_CV': '{:.1f}',
                          'Max_CV': '{:.1f}'}, na_rep='—'))

# --- Table 2: full per-replicate variance table, in plate order -------------
display(Markdown(f'### Replicate variance by Drug × Concentration (plate layout order)'))
display(Markdown(
    f'Rows follow the physical plate: assay, plate, each drug\'s '
    f'{_rep_axis.lower()} block top-first, then across the '
    f'{_other_axis.lower()}s (top dose → untreated) — so this table can be read '
    'side by side with the plate map. `Wells` lists the triplicate. '
    '`Worst_Well` is the well furthest from its triplicate median and '
    '`CV_pct_drop1` is what the CV would become without it; when dropping it '
    'brings the group under threshold, a single well is to blame.'))
_tbl2 = df_replicate_qc.set_index(['Assay', 'Drug', 'Day', 'Conc'])
display(
    _tbl2[['Plate', 'Wells', 'N', 'Mean', 'SD', 'CV_pct', 'Worst_Well', 'Worst_Value',
           'Worst_Dev_pct', 'Other_Reps', 'CV_pct_drop1', 'Verdict', 'Untreated']].style
    .format({'Mean': _sig4, 'SD': _sig4, 'CV_pct': '{:.1f}',
             'Worst_Value': _sig4, 'Worst_Dev_pct': '{:+.1f}',
             'CV_pct_drop1': '{:.1f}'}, na_rep='—')
    .format_index(lambda v: f'{v:g}', axis=0, level='Conc')
    .map(_plate_band, subset=['Plate'])
    .map(lambda v: 'font-weight: bold' if v in ('one suspect well', 'scattered') else None,
         subset=['Verdict'])
)

# --- Table 3: the actionable list ------------------------------------------
# Grouped by drug, inheriting the plate-layout order of the table above, so a
# drug's flagged concentrations read together and in dose order rather than
# being split apart by severity.
_flagged = df_replicate_qc[df_replicate_qc.Verdict != 'OK'].copy()
display(Markdown('### Suspect wells — groups above the CV threshold (grouped by drug)'))
if _flagged.empty:
    display(Markdown(f'_No replicate group exceeds {REPLICATE_CV_THRESHOLD:.0f}% CV._'))
else:
    _t3 = _flagged.set_index(['Assay', 'Drug', 'Conc'])
    display(
        _t3[['Plate', 'CV_pct', 'Worst_Well', 'Worst_Value', 'Other_Reps',
             'Worst_Dev_pct', 'CV_pct_drop1', 'Verdict', 'Likely_Cause', 'Untreated',
             'Reversal_Excluded']].style
        .format({'CV_pct': '{:.1f}', 'Worst_Value': _sig4,
                 'Worst_Dev_pct': '{:+.1f}', 'CV_pct_drop1': '{:.1f}'}, na_rep='—')
        .format_index(lambda v: f'{v:g}', axis=0, level='Conc')
        .map(_plate_band, subset=['Plate'])
    )

# --- Table 4: is a whole replicate position off? ----------------------------
display(Markdown(f'### Systematic bias by replicate position (plate {_rep_axis.lower()})'))
display(Markdown(
    f'Mean % deviation of each plate {_rep_axis.lower()} from its own triplicate '
    f'median. Near 0 = that replicate position tracks the others. '
    f'Beyond ±{REPLICATE_ROW_BIAS_PCT:.0f}% the position is systematically off, '
    'and individual wells flagged there are symptoms of that, not independent '
    'bad wells. The other plate axis is not shown: it carries concentration, so '
    'a "bias" along it would just be the dose response.'))
_bias_pivot = _pos_bias.pivot_table(index=['Assay', 'Plate'], columns=_rep_axis,
                                    values='Mean_Dev_pct')
_bias_pivot = _bias_pivot.rename_axis(columns=None)
display(
    _bias_pivot.style
    .format('{:+.1f}', na_rep='—')
    .map(lambda v: 'font-weight: bold' if (pd.notna(v) and abs(v) >= REPLICATE_ROW_BIAS_PCT)
         else None)
)

# --- Printed summary --------------------------------------------------------
print(f'\n  Replicate groups: {len(df_replicate_qc)} '
      f'({df_replicate_qc.N.min()}–{df_replicate_qc.N.max()} wells each)')
print(f'  Above {REPLICATE_CV_THRESHOLD:.0f}% CV: {int((df_replicate_qc.Verdict != "OK").sum())}'
      f'  [{int((df_replicate_qc.Verdict == "one suspect well").sum())} with a single '
      f'suspect well, {int((df_replicate_qc.Verdict == "scattered").sum())} scattered]')

_sys = _pos_bias[_pos_bias.Systematic]
if not _sys.empty:
    print(f'\n  Systematically biased replicate positions '
          f'(|mean deviation| >= {REPLICATE_ROW_BIAS_PCT:.0f}%):')
    for _r in _sys.itertuples(index=False):
        print(f'    {_r.Assay} plate {_r.Plate}, {_rep_axis.lower()} '
              f'{getattr(_r, _rep_axis)}: {_r.Mean_Dev_pct:+.1f}%')
    print('    -> re-check seeding/edge effects for these positions before '
          'blaming individual wells.')

_untx_bad = df_replicate_qc[(df_replicate_qc.Verdict != 'OK') & df_replicate_qc.Untreated]
if not _untx_bad.empty:
    print(f'\n  WARNING: {len(_untx_bad)} flagged group(s) are UNTREATED wells, which set '
          'the 100% baseline —')
    print('  an outlier there shifts %Viable for every concentration of that drug:')
    for _r in _untx_bad.itertuples(index=False):
        print(f'    {_r.Assay} plate {_r.Plate} {_r.Drug}: CV {_r.CV_pct:.1f}%, '
              f'worst well {_r.Worst_Well} ({_r.Worst_Dev_pct:+.0f}%)'
              + (f' — {_r.Likely_Cause}' if _r.Likely_Cause else ''))

=== 3b. Replicate variance & outlier wells (raw metric, per assay) ===
  Replicate axis detected: members of a triplicate differ by Row (constant Col); Col carries concentration.

Replicate variance overview — per assay

Replicate group = one triplicate: Assay, Plate, Drug, Day, Conc. A group is flagged when its CV exceeds 25%. Castor and MTT are summarised separately — their raw units and noise floors are not comparable.

  Raw metric Groups Median_CV P90_CV Max_CV Flagged Suspect_Well Scattered
Assay                
Castor Castor confluence (fraction) 40 11.8 22.7 64.9 3 1 2
MTT MTT absorbance (a.u.) 40 8.7 16.6 19.9 0 0 0

Replicate variance by Drug × Concentration (plate layout order)

Rows follow the physical plate: assay, plate, each drug’s row block top-first, then across the cols (top dose → untreated) — so this table can be read side by side with the plate map. Wells lists the triplicate. Worst_Well is the well furthest from its triplicate median and CV_pct_drop1 is what the CV would become without it; when dropping it brings the group under threshold, a single well is to blame.

        Plate Wells N Mean SD CV_pct Worst_Well Worst_Value Worst_Dev_pct Other_Reps CV_pct_drop1 Verdict Untreated
Assay Drug Day Conc                          
Castor 63GE8 7 1000 1 B2/C2/D2 3 0.1996 0.127 63.6 D2 0.338 +96.4 0.0886 / 0.1721 45.3 scattered False
222.2 1 B3/C3/D3 3 0.07992 0.0519 64.9 D3 0.1383 +121.4 0.039 / 0.06247 32.7 scattered False
49.4 1 B4/C4/D4 3 0.0535 0.01403 26.2 B4 0.03774 -35.0 0.06464 / 0.05811 7.5 one suspect well False
11 1 B5/C5/D5 3 0.0604 0.007224 12.0 B5 0.05221 -17.3 0.06586 / 0.06314 3.0 OK False
2.44 1 B6/C6/D6 3 0.06872 0.01508 21.9 B6 0.05142 -32.0 0.07565 / 0.0791 3.2 OK False
0.54 1 B7/C7/D7 3 0.06179 0.003831 6.2 B7 0.05738 -9.9 0.06367 / 0.06431 0.7 OK False
0.12 1 B8/C8/D8 3 0.07039 0.009447 13.4 B8 0.05992 -17.9 0.07296 / 0.07828 5.0 OK False
0.026 1 B9/C9/D9 3 0.08111 0.0153 18.9 B9 0.06541 -20.2 0.09597 / 0.08196 11.1 OK False
0.006 1 B10/C10/D10 3 0.1026 0.01251 12.2 D10 0.1164 +17.0 0.09197 / 0.09953 5.6 OK False
0.0006 1 B11/C11/D11 3 0.08701 0.007421 8.5 C11 0.07865 -12.2 0.08954 / 0.09283 2.5 OK True
IsoGE8 7 1000 1 E2/F2/G2 3 0.07569 0.01229 16.2 F2 0.08907 +21.9 0.07308 / 0.06491 8.4 OK False
222.2 1 E3/F3/G3 3 0.05456 0.01023 18.8 E3 0.04287 -27.3 0.05896 / 0.06186 3.4 OK False
49.4 1 E4/F4/G4 3 0.06804 0.009257 13.6 F4 0.05791 -17.5 0.07605 / 0.07016 5.7 OK False
11 1 E5/F5/G5 3 0.09628 0.005802 6.0 F5 0.08972 -8.8 0.09838 / 0.1007 1.7 OK False
2.44 1 E6/F6/G6 3 0.1193 0.02699 22.6 F6 0.09204 -23.2 0.146 / 0.1198 13.9 OK False
0.54 1 E7/F7/G7 3 0.1007 0.003161 3.1 G7 0.09705 -4.9 0.1029 / 0.102 0.6 OK False
0.12 1 E8/F8/G8 3 0.1167 0.01359 11.6 G8 0.1309 +13.3 0.1155 / 0.1038 7.5 OK False
0.026 1 E9/F9/G9 3 0.1103 0.008853 8.0 G9 0.1001 -12.8 0.1149 / 0.1159 0.7 OK False
0.006 1 E10/F10/G10 3 0.1069 0.012 11.2 F10 0.0932 -16.8 0.112 / 0.1155 2.1 OK False
0.0006 1 E11/F11/G11 3 0.08084 0.01884 23.3 E11 0.06041 -28.6 0.08458 / 0.09753 10.1 OK True
64GE6 7 1000 2 B2/C2/D2 3 0.04692 0.001962 4.2 D2 0.04907 +5.6 0.04522 / 0.04647 1.9 OK False
222.2 2 B3/C3/D3 3 0.0426 0.00829 19.5 D3 0.03381 -22.6 0.05028 / 0.0437 9.9 OK False
49.4 2 B4/C4/D4 3 0.04704 0.008324 17.7 D4 0.0555 +18.7 0.04677 / 0.03886 13.1 OK False
11 2 B5/C5/D5 3 0.05709 0.007112 12.5 C5 0.06463 +15.2 0.05051 / 0.05612 7.4 OK False
2.44 2 B6/C6/D6 3 0.05747 0.007416 12.9 B6 0.04939 -16.4 0.06396 / 0.05907 5.6 OK False
0.54 2 B7/C7/D7 3 0.06502 0.006575 10.1 B7 0.05767 -14.0 0.07035 / 0.06702 3.4 OK False
0.12 2 B8/C8/D8 3 0.07757 0.008874 11.4 B8 0.06773 -15.4 0.08496 / 0.08001 4.2 OK False
0.026 2 B9/C9/D9 3 0.1086 0.005476 5.0 D9 0.1149 +8.1 0.1047 / 0.1062 1.0 OK False
0.006 2 B10/C10/D10 3 0.107 0.005086 4.8 D10 0.1011 -7.8 0.1102 / 0.1097 0.3 OK False
0.0006 2 B11/C11/D11 3 0.1254 0.02004 16.0 C11 0.147 +20.6 0.1219 / 0.1074 8.9 OK True
64VPE6 7 1000 2 E2/F2/G2 3 0.04021 0.003385 8.4 G2 0.04402 +12.8 0.03757 / 0.03902 2.7 OK False
222.2 2 E3/F3/G3 3 0.04015 0.001939 4.8 G3 0.03792 -8.1 0.04129 / 0.04126 0.0 OK False
49.4 2 E4/F4/G4 3 0.04266 0.003499 8.2 F4 0.03884 -10.6 0.0457 / 0.04345 3.6 OK False
11 2 E5/F5/G5 3 0.03579 0.003929 11.0 F5 0.03126 -17.5 0.03822 / 0.03789 0.6 OK False
2.44 2 E6/F6/G6 3 0.05625 0.005301 9.4 G6 0.05018 -14.4 0.05995 / 0.05863 1.6 OK False
0.54 2 E7/F7/G7 3 0.08088 0.009098 11.2 G7 0.07039 -17.9 0.0857 / 0.08656 0.7 OK False
0.12 2 E8/F8/G8 3 0.1046 0.01442 13.8 E8 0.08912 -16.7 0.1176 / 0.107 6.7 OK False
0.026 2 E9/F9/G9 3 0.116 0.02198 19.0 F9 0.09303 -21.2 0.1369 / 0.118 10.4 OK False
0.006 2 E10/F10/G10 3 0.1207 0.006494 5.4 F10 0.1132 -8.9 0.1246 / 0.1243 0.2 OK False
0.0006 2 E11/F11/G11 3 0.1136 0.01014 8.9 F11 0.102 -14.2 0.1201 / 0.1189 0.7 OK True
MTT 63GE8 7 1000 1 B2/C2/D2 3 0.083 0.01004 12.1 B2 0.0944 +19.3 0.0755 / 0.0791 3.3 OK False
222.2 1 B3/C3/D3 3 0.08533 0.0141 16.5 B3 0.1013 +26.5 0.0801 / 0.0746 5.0 OK False
49.4 1 B4/C4/D4 3 0.09197 0.005095 5.5 D4 0.0864 -7.2 0.0964 / 0.0931 2.5 OK False
11 1 B5/C5/D5 3 0.1013 0.003041 3.0 B5 0.1048 +5.0 0.0998 / 0.0993 0.4 OK False
2.44 1 B6/C6/D6 3 0.1036 0.009886 9.5 D6 0.0933 -10.8 0.1046 / 0.113 5.5 OK False
0.54 1 B7/C7/D7 3 0.1224 0.01632 13.3 D7 0.1046 -16.9 0.1258 / 0.1367 5.9 OK False
0.12 1 B8/C8/D8 3 0.1342 0.02569 19.1 D8 0.1067 -22.8 0.1382 / 0.1576 9.3 OK False
0.026 1 B9/C9/D9 3 0.1666 0.006151 3.7 D9 0.1595 -6.2 0.17 / 0.1703 0.1 OK False
0.006 1 B10/C10/D10 3 0.1845 0.02458 13.3 D10 0.1579 -16.5 0.1891 / 0.2064 6.2 OK False
0.0006 1 B11/C11/D11 3 0.1936 0.009802 5.1 B11 0.1837 -5.2 0.1938 / 0.2033 3.4 OK True
IsoGE8 7 1000 1 E2/F2/G2 3 0.0669 0.005462 8.2 G2 0.0606 -13.2 0.0698 / 0.0703 0.5 OK False
222.2 1 E3/F3/G3 3 0.06837 0.00206 3.0 F3 0.0707 +4.6 0.0668 / 0.0676 0.8 OK False
49.4 1 E4/F4/G4 3 0.1028 0.004543 4.4 E4 0.108 +7.1 0.1008 / 0.0996 0.8 OK False
11 1 E5/F5/G5 3 0.1313 0.02265 17.3 F5 0.1573 +30.6 0.1161 / 0.1204 2.6 OK False
2.44 1 E6/F6/G6 3 0.1343 0.01157 8.6 E6 0.1214 -12.0 0.1379 / 0.1437 2.9 OK False
0.54 1 E7/F7/G7 3 0.1357 0.002816 2.1 F7 0.1388 +2.8 0.135 / 0.1333 0.9 OK False
0.12 1 E8/F8/G8 3 0.1452 0.006892 4.7 G8 0.1376 -6.5 0.151 / 0.1471 1.9 OK False
0.026 1 E9/F9/G9 3 0.1557 0.01996 12.8 F9 0.1785 +21.3 0.1414 / 0.1472 2.8 OK False
0.006 1 E10/F10/G10 3 0.1593 0.03162 19.9 F10 0.194 +27.8 0.1518 / 0.1321 9.8 OK False
0.0006 1 E11/F11/G11 3 0.1763 0.02261 12.8 F11 0.2022 +21.9 0.1607 / 0.1659 2.3 OK True
64GE6 7 1000 2 B2/C2/D2 3 0.0707 0.00652 9.2 C2 0.0637 -11.3 0.0766 / 0.0718 4.6 OK False
222.2 2 B3/C3/D3 3 0.08053 0.008179 10.2 D3 0.0899 +16.9 0.0769 / 0.0748 2.0 OK False
49.4 2 B4/C4/D4 3 0.0828 0.006393 7.7 C4 0.0755 -11.7 0.0855 / 0.0874 1.6 OK False
11 2 B5/C5/D5 3 0.09037 0.01157 12.8 C5 0.0771 -19.4 0.0956 / 0.0984 2.0 OK False
2.44 2 B6/C6/D6 3 0.09617 0.007504 7.8 C6 0.0878 -10.8 0.1023 / 0.0984 2.7 OK False
0.54 2 B7/C7/D7 3 0.1123 0.002205 2.0 C7 0.1099 -2.7 0.1142 / 0.1129 0.8 OK False
0.12 2 B8/C8/D8 3 0.1347 0.006412 4.8 C8 0.1421 +8.3 0.1308 / 0.1312 0.2 OK False
0.026 2 B9/C9/D9 3 0.1598 0.004291 2.7 D9 0.1549 -4.3 0.1619 / 0.1627 0.3 OK False
0.006 2 B10/C10/D10 3 0.1773 0.01467 8.3 D10 0.1615 -10.2 0.1798 / 0.1905 4.1 OK False
0.0006 2 B11/C11/D11 3 0.1728 0.02343 13.6 C11 0.1481 -15.7 0.1756 / 0.1947 7.3 OK True
64VPE6 7 1000 2 E2/F2/G2 3 0.0666 0.005011 7.5 F2 0.072 +9.6 0.0657 / 0.0621 4.0 OK False
222.2 2 E3/F3/G3 3 0.0653 0.007134 10.9 F3 0.0733 +16.3 0.063 / 0.0596 3.9 OK False
49.4 2 E4/F4/G4 3 0.07283 0.006341 8.7 F4 0.0798 +11.9 0.0713 / 0.0674 4.0 OK False
11 2 E5/F5/G5 3 0.07137 0.007778 10.9 F5 0.0795 +12.6 0.0706 / 0.064 6.9 OK False
2.44 2 E6/F6/G6 3 0.09227 0.01055 11.4 F6 0.1043 +18.7 0.0879 / 0.0846 2.7 OK False
0.54 2 E7/F7/G7 3 0.1211 0.007941 6.6 F7 0.1302 +11.1 0.1172 / 0.1158 0.8 OK False
0.12 2 E8/F8/G8 3 0.1322 0.009465 7.2 E8 0.122 -8.9 0.1407 / 0.1339 3.5 OK False
0.026 2 E9/F9/G9 3 0.1634 0.01667 10.2 G9 0.1455 -12.4 0.1661 / 0.1785 5.1 OK False
0.006 2 E10/F10/G10 3 0.1618 0.01004 6.2 G10 0.1506 -8.6 0.17 / 0.1648 2.2 OK False
0.0006 2 E11/F11/G11 3 0.169 0.03102 18.4 F11 0.204 +29.2 0.145 / 0.1579 6.0 OK True

Suspect wells — groups above the CV threshold (grouped by drug)

      Plate CV_pct Worst_Well Worst_Value Other_Reps Worst_Dev_pct CV_pct_drop1 Verdict Likely_Cause Untreated Reversal_Excluded
Assay Drug Conc                      
Castor 63GE8 1000 1 63.6 D2 0.338 0.0886 / 0.1721 +96.4 45.3 scattered False True
222.2 1 64.9 D3 0.1383 0.039 / 0.06247 +121.4 32.7 scattered False True
49.4 1 26.2 B4 0.03774 0.06464 / 0.05811 -35.0 7.5 one suspect well Row B biased (-23% plate-wide) False False

Systematic bias by replicate position (plate row)

Mean % deviation of each plate row from its own triplicate median. Near 0 = that replicate position tracks the others. Beyond ±15% the position is systematically off, and individual wells flagged there are symptoms of that, not independent bad wells. The other plate axis is not shown: it carries concentration, so a “bias” along it would just be the dose response.

    B C D E F G
Assay Plate            
Castor 1 -22.6 +2.0 +25.1 -2.5 -5.4 +1.0
2 -4.4 +3.8 -1.0 +0.5 -6.2 -2.8
MTT 1 +4.9 +3.6 -8.3 -1.4 +11.0 -3.1
2 +1.1 -5.9 +1.8 -1.4 +12.2 -5.2

  Replicate groups: 80 (3–3 wells each)
  Above 25% CV: 3  [1 with a single suspect well, 2 scattered]

  Systematically biased replicate positions (|mean deviation| >= 15%):
    Castor plate 1, row B: -22.6%
    Castor plate 1, row D: +25.1%
    -> re-check seeding/edge effects for these positions before blaming individual wells.

Drug Colour & Assay Style Setup

Show code
# ===========================================================================
# 4. Style mapping — COLOUR = drug, LINESTYLE/MARKER/HATCH = assay
#    A drug keeps one colour in every plot and in every legend; the assay it was
#    measured with is what changes the line (Castor dashed, MTT solid).
# ===========================================================================
all_drugs  = sorted(df.Drug.dropna().unique())
all_assays = sorted(df.Assay.dropna().unique())
all_days   = sorted(df.Day.unique())
# Day 0, if scanned, is a seeding-check timepoint and is not fitted.
fit_days = [d for d in all_days if d > 0] or all_days

# One series = one (Drug, Assay) pair actually present in the data.
drug_assay_pairs = sorted(
    df[['Drug', 'Assay']].dropna().drop_duplicates()
      .itertuples(index=False, name=None)
)


def series_label(drug, assay):
    """Legend/table label for one curve; the assay suffix is dropped if there is only one."""
    return f'{drug} · {assay}' if len(all_assays) > 1 else str(drug)


def _split_ab_lp(name):
    """Split 'AB-LP' on the first dash. LP may contain further dashes."""
    idx = name.find('-')
    return (name[:idx], name[idx+1:]) if idx >= 0 else (name, None)


# Unique AB prefixes and LP suffixes (still used by the linker-payload trellises)
_all_abs_setup = sorted({_split_ab_lp(d)[0] for d in all_drugs})
_all_lps_setup = sorted({lp for d in all_drugs
                          for lp in [_split_ab_lp(d)[1]] if lp is not None})

# --- Colour: one per drug, stable across assays ---
_palette = (sns.color_palette('tab10', len(all_drugs)) if len(all_drugs) <= 10
            else sns.color_palette('husl', len(all_drugs)))
drug_colors = {d: _palette[i] for i, d in enumerate(all_drugs)}

# --- Line style / marker / hatch: one per assay ---
# Fallback cycles for assays not named in ASSAY_STYLES. The dash patterns are
# deliberately coarse — matplotlib's default '--'/':' are hard to tell apart on
# a crowded log axis and vanish in a short legend sample.
_ls_cycle = ['-', (0, (6, 2)), (0, (1, 1.8)), (0, (8, 2, 1.5, 2)), (0, (3, 1.5)),
             (0, (6, 2, 1.5, 2, 1.5, 2)), (0, (10, 3))]
_lw_cycle = [1.7, 2.2, 2.4, 2.2, 2.0, 2.2, 2.2]
_mk_cycle = ['o', 's', '^', 'D', 'v', 'P', 'X']
_ht_cycle = ['', '//', '\\\\', 'xx', '..', '++']
assay_linestyles, assay_linewidths, assay_markers, assay_hatches = {}, {}, {}, {}
for i, a in enumerate(all_assays):
    _st = ASSAY_STYLES.get(a, {})
    assay_linestyles[a] = _st.get('linestyle', _ls_cycle[i % len(_ls_cycle)])
    assay_linewidths[a] = _st.get('linewidth', _lw_cycle[i % len(_lw_cycle)])
    assay_markers[a]    = _st.get('marker',    _mk_cycle[i % len(_mk_cycle)])
    assay_hatches[a]    = _st.get('hatch',     _ht_cycle[i % len(_ht_cycle)])


# --- Raw-metric y-axis split (see RAW_RIGHT_AXIS_ASSAY in the user input) ---
# Only meaningful with more than one assay in the data; a single-assay run keeps
# one left axis and these helpers become no-ops.
_raw_right_assay = (RAW_RIGHT_AXIS_ASSAY
                    if len(all_assays) > 1 and RAW_RIGHT_AXIS_ASSAY in all_assays
                    else None)
_raw_left_assays = [a for a in all_assays if a != _raw_right_assay]


def _raw_ylabel(assays):
    """Axis label for a raw-metric axis carrying `assays`."""
    return ' / '.join(ASSAY_RAW_YLABEL.get(a, f'{a} raw metric') for a in assays)


def raw_metric_axes(ax, assays=None):
    """Axes to draw each assay's raw metric on, splitting off a twin if asked.

    Returns `(axis_for, sides)`: a drug-agnostic assay -> axis map, and a list of
    `(axis, spine_side, assays)` for the caller to scale and label. Each axis
    ends up holding one assay's artists only, so autoscaling (or `set_ylim` from
    that assay's rows) gives each readout the full plot height and the curves
    overlap instead of stacking. `sides` has one entry when unsplit, two when
    split, so `len(sides) > 1` is the "is this plot split" test.
    """
    assays = list(all_assays) if assays is None else list(assays)
    right = _raw_right_assay if (len(assays) > 1 and _raw_right_assay in assays) else None
    if right is None:
        return {a: ax for a in assays}, [(ax, 'left', assays)]
    ax2 = ax.twinx()
    ax2.grid(False)                  # one set of gridlines, owned by the left axis
    return ({a: (ax2 if a == right else ax) for a in assays},
            [(ax, 'left', [a for a in assays if a != right]),
             (ax2, 'right', [right])])


def style_raw_axis(ax, side, assays, split):
    """Y-label for one side of a raw-metric plot, plus a spine cue when split.

    With one assay per side the spine is drawn in that assay's line style, so the
    axis itself repeats the linestyle=assay encoding the curves already use.
    """
    ax.set_ylabel(_raw_ylabel(assays))
    if split and len(assays) == 1:
        sp = ax.spines[side]
        sp.set_visible(True)
        sp.set_linestyle(assay_linestyles[assays[0]])
        sp.set_linewidth(1.4)
        sp.set_color('0.35')


def assay_axis_labels(sides):
    """assay -> 'Name (left/right axis)' for a split plot; empty dict if unsplit."""
    if len(sides) < 2:
        return {}
    return {a: f'{a} ({side} axis)' for _, side, assays in sides for a in assays}


LEGEND_SAMPLE_LW = 1.5   # see the note in style_legend_handles below


def style_legend_handles(drugs=None, assays=None, drug_title='Drug', assay_title='Assay',
                         assay_labels=None):
    """Factored legend: one colour swatch per drug + one line sample per assay.

    Factored rather than one entry per (drug, assay) curve, so the legend states
    the encoding itself instead of repeating every combination.

    `assay_labels` overrides the text of individual assay entries (e.g. to note
    which y-axis an assay is drawn against); assays left out keep their name.
    """
    drugs  = list(all_drugs)  if drugs  is None else list(drugs)
    assays = list(all_assays) if assays is None else list(assays)
    assay_labels = assay_labels or {}
    handles = [Line2D([], [], linestyle='none', marker='', label=drug_title)]
    handles += [Line2D([], [], color=drug_colors[d], lw=2.5, label=str(d)) for d in drugs]
    if len(assays) > 1:
        handles += [Line2D([], [], linestyle='none', marker='', label=assay_title)]
        # NB: matplotlib scales dash patterns by line width, so the sample is
        # drawn thin (not at the assay's plot width) or one dash fills the key.
        handles += [Line2D([], [], color='0.35', lw=LEGEND_SAMPLE_LW,
                           linestyle=assay_linestyles[a], marker=assay_markers[a],
                           markersize=5, label=str(assay_labels.get(a, a)))
                    for a in assays]
    return handles


def marker_key_handle(label, assays=None):
    """Legend key reading ○/◻ — one hollow marker per assay, slash-separated.

    IC50 points are drawn with the assay's marker in the drug's colour, so the
    key shows every shape in use rather than picking one drug's colour, and one
    assay's marker, to stand for all of them.
    """
    assays = list(all_assays) if assays is None else list(assays)
    parts = []
    for a in assays:
        if parts:                                    # separator between shapes
            parts.append(Line2D([], [], marker='$/$', markersize=7,
                                color='black', linestyle='none',
                                label='_nolegend_'))
        parts.append(Line2D([], [], marker=assay_markers[a], markersize=7,
                            markerfacecolor='none', markeredgecolor='black',
                            markeredgewidth=1.2, linestyle='none',
                            label='_nolegend_'))
    parts[0].set_label(label)                        # label rides on the first part
    return tuple(parts)


def draw_legend(ax, handles, **kw):
    """ax.legend for handle lists that may contain tuples (multi-marker keys)."""
    labels = []
    for h in handles:
        if isinstance(h, tuple):
            labels.append(next((x.get_label() for x in h
                                if not x.get_label().startswith('_')), ''))
        else:
            labels.append(h.get_label())
    kw['handler_map'] = {tuple: HandlerTuple(ndivide=None, pad=0.1),
                         **kw.get('handler_map', {})}
    kw.setdefault('handlelength', 3.4)
    return ax.legend(handles, labels, **kw)


print(f'\n{len(all_drugs)} drugs: {all_drugs}')
print(f'  AB prefixes  ({len(_all_abs_setup)}): {_all_abs_setup}')
print(f'  Linker-payloads ({len(_all_lps_setup)}): {_all_lps_setup}')
print(f'  Assays ({len(all_assays)}): {all_assays}')
for a in all_assays:
    _side = 'right' if a == _raw_right_assay else 'left'
    print(f'    {a:<8} linestyle={assay_linestyles[a]!r}  marker={assay_markers[a]!r}'
          + (f'  raw-metric axis={_side}' if _raw_right_assay else ''))
print(f'  Series (drug × assay): {len(drug_assay_pairs)}')
for d, a in drug_assay_pairs:
    print(f'    {series_label(d, a)}')

4 drugs: ['63GE8', '64GE6', '64VPE6', 'IsoGE8']
  AB prefixes  (4): ['63GE8', '64GE6', '64VPE6', 'IsoGE8']
  Linker-payloads (0): []
  Assays (2): ['Castor', 'MTT']
    Castor   linestyle=(0, (6, 2))  marker='o'  raw-metric axis=left
    MTT      linestyle='-'  marker='s'  raw-metric axis=right
  Series (drug × assay): 8
    63GE8 · Castor
    63GE8 · MTT
    64GE6 · Castor
    64GE6 · MTT
    64VPE6 · Castor
    64VPE6 · MTT
    IsoGE8 · Castor
    IsoGE8 · MTT

Growth Kinetics Plots

Show code
# ===========================================================================
# 5. Growth kinetics plots (per drug, both assays overlaid)
#    Colour = concentration here (the drug is fixed per figure);
#    linestyle/marker still = assay.
#    The metric is raw and its units differ by assay, so one assay is drawn
#    against a twin right-hand axis (see raw_metric_axes).
# ===========================================================================
print('\n=== 4. Growth kinetics plots ===')
for drug in all_drugs:
    sub_drug = df[df.Drug == drug]
    if sub_drug.empty:
        continue
    assays_here = [a for a in all_assays if not sub_drug[sub_drug.Assay == a].empty]
    concs = sorted(sub_drug.Conc.unique())
    conc_colors = dict(zip(concs, sns.color_palette('viridis', len(concs))))

    fig, ax = plt.subplots(figsize=(8, 5))
    axis_for, sides = raw_metric_axes(ax, assays_here)
    for assay in assays_here:
        _ax = axis_for[assay]
        for conc in concs:
            csub = sub_drug[(sub_drug.Assay == assay) & (sub_drug.Conc == conc)]
            if csub.empty:
                continue
            means = csub.groupby('Day')['Confluence'].mean()
            sds   = csub.groupby('Day')['Confluence'].std()
            _ax.errorbar(means.index, means.values, yerr=sds.values,
                         marker=assay_markers[assay], markersize=5, capsize=2,
                         color=conc_colors[conc],
                         linestyle=assay_linestyles[assay],
                         linewidth=assay_linewidths[assay], elinewidth=0.9)
    ax.set_xlabel('Day')
    # Each side holds one assay's curves only, so matplotlib's own autoscaling
    # already gives each readout the full plot height — no explicit limits.
    _split = len(sides) > 1
    for _ax, _side, _assays in sides:
        style_raw_axis(_ax, _side, _assays, _split)
    ax.set_xticks(all_days)

    _axis_labels = assay_axis_labels(sides)
    handles = [Line2D([], [], color=conc_colors[c], lw=2.5,
                      label='untreated' if c == PSEUDO_CONC else f'{c} nM')
               for c in concs]
    if len(assays_here) > 1:
        handles += [Line2D([], [], linestyle='none', marker='', label='Assay')]
        handles += [Line2D([], [], color='0.35', lw=LEGEND_SAMPLE_LW,
                           linestyle=assay_linestyles[a], marker=assay_markers[a],
                           markersize=5, label=str(_axis_labels.get(a, a)))
                    for a in assays_here]
    # Legend on the topmost axis, so the twin's curves cannot be drawn over it.
    draw_legend(sides[-1][0], handles, title='Conc', fontsize=8, title_fontsize=9)

    fig.suptitle(f'{drug} — Growth Kinetics ({" + ".join(assays_here)})', fontsize=14)
    plt.tight_layout()
    save_fig()

=== 4. Growth kinetics plots ===

LL.4 Model & Interpolation Functions

Show code
# ===========================================================================
# 6. LL.4 fitting functions
# ===========================================================================
print('\n=== 5. LL.4 model ===')

def ll4(x, slope, lower, upper, ec50):
    return lower + (upper - lower) / (1.0 + (x / ec50) ** slope)


def compute_r2(y_obs, y_pred):
    ss_res = np.sum((y_obs - y_pred) ** 2)
    ss_tot = np.sum((y_obs - np.mean(y_obs)) ** 2)
    if ss_tot == 0:
        return 0.0
    return 1.0 - ss_res / ss_tot


def fit_ll4(conc, response, p0=None):
    """Fit LL.4 using the points supplied by the shared analysis filter."""
    conc = np.asarray(conc, dtype=float)
    response = np.asarray(response, dtype=float)
    mask = (conc > 0) & np.isfinite(response)
    conc, response = conc[mask], response[mask]
    if len(conc) < 4 or len(np.unique(conc)) < 3:
        return None, None

    cmin, cmax = conc.min(), conc.max()
    if p0 is None:
        p0 = [1.0, np.min(response), np.max(response), np.median(conc)]

    bounds = (
        [0.01,   -np.inf, -np.inf, cmin * 0.01],
        [np.inf,  np.inf,  np.inf, cmax * 100 ]
    )

    try:
        popt, _ = curve_fit(ll4, conc, response, p0=p0,
                            maxfev=10000, bounds=bounds)
    except Exception:
        return None, None

    y_pred = ll4(conc, *popt)
    r2 = compute_r2(response, y_pred)
    return popt, r2


def fit_ll4_pct_viable(conc, response, upper=100.0):
    """Fit %Viable LL.4 with the untreated-normalized upper plateau fixed at 100%."""
    conc = np.asarray(conc, dtype=float)
    response = np.asarray(response, dtype=float)
    mask = (conc > 0) & np.isfinite(response)
    conc, response = conc[mask], response[mask]
    if len(conc) < 4 or len(np.unique(conc)) < 3:
        return None, None

    cmin, cmax = conc.min(), conc.max()

    def _ll4_fixed_upper(x, slope, lower, ec50):
        return ll4(x, slope, lower, upper, ec50)

    try:
        popt_fixed, _ = curve_fit(
            _ll4_fixed_upper, conc, response,
            p0=[1.0, np.min(response), np.median(conc)],
            bounds=([0.01, -np.inf, cmin * 0.01],
                    [np.inf, np.inf, cmax * 100]),
            maxfev=10000,
        )
    except Exception:
        return None, None

    slope, lower, ec50 = popt_fixed
    popt = np.array([slope, lower, upper, ec50])
    r2 = compute_r2(response, ll4(conc, *popt))
    return popt, r2


def classify_ic50(row):
    if row['R2'] < R2_THRESHOLD:
        return 'poor fit'
    if row['IC50_nM'] < CONC_MIN * 0.5:
        return f'<{CONC_MIN} nM'
    if row['IC50_nM'] > CONC_MAX * 2:
        return f'>{CONC_MAX} nM'
    return f"{row['IC50_nM']:.3g}"


def interpolate_ic50(concs, pct_viable, target=50.0):
    """
    Find concentration where mean %Viable crosses `target` (default 50%).
    Linear interpolation on log10(conc).
    Exclusions are applied before this function is called.
    Returns NaN if the curve never crosses the target.
    """
    means = pct_viable.groupby(concs).mean()
    c = np.array(means.index, dtype=float)
    v = np.array(means.values, dtype=float)
    order = np.argsort(c)
    c, v = c[order], v[order]

    for i in range(len(v) - 1):
        if (v[i] >= target and v[i+1] < target) or \
           (v[i] <= target and v[i+1] > target):
            log_c1, log_c2 = np.log10(c[i]), np.log10(c[i+1])
            frac = (target - v[i]) / (v[i+1] - v[i])
            return 10 ** (log_c1 + frac * (log_c2 - log_c1))

    if v[-1] < target and v[0] >= target:
        return c[-1]
    return np.nan


print(f'LL.4 defined.  R² threshold = {R2_THRESHOLD}')
print('Interpolation IC50 function defined.')
print('Drug/day-specific >25% reversal exclusions active for all IC50 calculations.')

=== 5. LL.4 model ===
LL.4 defined.  R² threshold = 0.5
Interpolation IC50 function defined.
Drug/day-specific >25% reversal exclusions active for all IC50 calculations.

Dose-Response Fitting & Plotting

Show code
# ===========================================================================
# 7. Dose-response fitting & plotting — per (Drug, Assay)
#    Every fit, normalization and IC50 below is keyed on (Drug, Assay, Day):
#    Castor and MTT measurements of the same drug are never pooled.
# ===========================================================================
metric_name = 'Confluence'
minfo = {
    'col': 'Confluence',
    'ylabel_pv': '% Viable (untreated-normalized, within assay)',
}

all_ic50_tables = {}

# x-axis window derived from user-defined scale bounds
_xlo = PSEUDO_CONC if SCALE_MIN_CONC == 0 else SCALE_MIN_CONC
_xhi = CONC_MAX if SCALE_MAX_CONC is None else SCALE_MAX_CONC


def _analysis_rows(sub_df):
    """Rows allowed to affect scaling, fitted curves, and IC50 calculations."""
    return sub_df[~sub_df['Exclude_25pct_Reversal']].copy()


def _series_rows(drug, assay, day, value_col):
    """Wells for one (drug, assay, day) series with `value_col` present."""
    sub = df[(df.Drug == drug) & (df.Assay == assay) & (df.Day == day)].copy()
    return sub.dropna(subset=[value_col])


def _fit_window(sub_df):
    """Analysis-eligible rows inside the visible x-window."""
    return _analysis_rows(sub_df[(sub_df.Conc >= _xlo) & (sub_df.Conc <= _xhi)])


def _axis_limits(sub_df, col, xlo, xhi, pad=0.10):
    """Y-limits from data within [xlo, xhi], honouring SCALE_EXCL_ABS."""
    vis = _analysis_rows(sub_df[(sub_df['Conc'] >= xlo) & (sub_df['Conc'] <= xhi)])
    if SCALE_EXCL_ABS and 'Drug' in vis.columns:
        vis = vis[~vis['Drug'].str.split('-').str[0].isin(SCALE_EXCL_ABS)]
    if vis.empty:
        return None
    m = vis.groupby('Conc')[col].mean()
    s = vis.groupby('Conc')[col].std().fillna(0)
    lo, hi = (m - s).min(), (m + s).max()
    mg = (hi - lo) * pad if hi > lo else max(abs(hi), abs(lo)) * pad + 0.1
    return lo - mg, hi + mg


def _plot_series(ax, drug, assay, sub, value_col, fit_params, day):
    """Points + LL.4 curve for one series. Colour = drug, linestyle/marker = assay.

    With 8 series on one axis the SD whiskers otherwise dominate, so they are
    drawn faint and behind the fitted curves; the markers stay full strength and
    carry a thin white edge so overlapping points still read as separate.
    """
    means = sub.groupby('Conc')[value_col].mean()
    sds   = sub.groupby('Conc')[value_col].std()
    ax.errorbar(means.index, means.values, yerr=sds.values,
                marker=assay_markers[assay], markersize=5,
                markeredgecolor='white', markeredgewidth=0.6,
                color=drug_colors[drug], linestyle='none',
                ecolor=matplotlib.colors.to_rgba(drug_colors[drug], 0.40),
                elinewidth=0.9, capsize=2, capthick=0.9, zorder=2)
    key = (drug, assay, day)
    if key not in fit_params:
        return
    curve_sub = _fit_window(sub)
    if curve_sub.empty:
        return
    x_pred = np.geomspace(curve_sub.Conc.min(), curve_sub.Conc.max(), 200)
    ax.plot(x_pred, ll4(x_pred, *fit_params[key]),
            color=drug_colors[drug], linewidth=assay_linewidths[assay],
            linestyle=assay_linestyles[assay],
            solid_capstyle='round', dash_capstyle='round', zorder=3)


col = minfo['col']
print(f'\n{"="*60}')
print(f'=== METRIC: {metric_name} — assays: {", ".join(all_assays)} ===')
print(f'{"="*60}')

# --- 7a. Dose-response plots (raw metric) ---
print(f'\n--- Dose-response plots (raw {metric_name}) ---')
fit_params_dr = {}
ic50_dr_records = []

for drug, assay in drug_assay_pairs:
    for day in fit_days:
        sub = _fit_window(_series_rows(drug, assay, day, col).dropna(subset=['Conc']))
        if sub.empty:
            continue
        popt, r2 = fit_ll4(sub.Conc.values, sub[col].values)
        rec = {'Drug': drug, 'Assay': assay, 'Day': day}
        if popt is not None:
            fit_params_dr[(drug, assay, day)] = popt
            rec.update({'IC50_nM': popt[3], 'Slope': popt[0],
                        'Lower': popt[1], 'Upper': popt[2], 'R2': r2})
        else:
            rec.update({'IC50_nM': np.nan, 'Slope': np.nan,
                        'Lower': np.nan, 'Upper': np.nan, 'R2': np.nan})
        ic50_dr_records.append(rec)

for day in fit_days:
    fig, ax = plt.subplots(figsize=(12, 7))
    axis_for, sides = raw_metric_axes(ax)
    _split = len(sides) > 1

    for drug, assay in drug_assay_pairs:
        sub = _series_rows(drug, assay, day, col)
        if sub.empty:
            continue
        _plot_series(axis_for[assay], drug, assay, sub, col, fit_params_dr, day)

    # Each side is scaled from its own assays' rows, so both readouts fill the
    # plot height and overlap instead of stacking into separate bands.
    for _ax, _side, _assays in sides:
        _vis = df[(df.Day == day) & df.Assay.isin(_assays)].dropna(subset=[col])
        _lim = _axis_limits(_vis, col, _xlo, _xhi)
        if _lim:
            _ax.set_ylim(*_lim)
        style_raw_axis(_ax, _side, _assays, _split)

    ax.set_xscale('log')
    ax.set_xlabel('Concentration (nM, log scale)')
    ax.set_title(
        f'Day {day} — Dose Response (raw metric, one axis per assay: '
        f'{"/".join(_raw_left_assays)} left, {_raw_right_assay} right)' if _split
        else f'Day {day} — Dose Response (raw metric, units differ by assay)')
    # Legend on the topmost axis, so the twin's curves cannot be drawn over it.
    draw_legend(sides[-1][0],
                style_legend_handles(assay_labels=assay_axis_labels(sides)),
                fontsize=8, loc='lower left', ncol=2)
    plt.tight_layout()
    save_fig()

df_ic50_dr = pd.DataFrame(ic50_dr_records)
df_ic50_dr['Reliable'] = (
    (df_ic50_dr['R2'] >= R2_THRESHOLD) &
    (df_ic50_dr['IC50_nM'] >= CONC_MIN * 0.5) &
    (df_ic50_dr['IC50_nM'] <= CONC_MAX * 2)
)
df_ic50_dr['Status'] = df_ic50_dr.apply(classify_ic50, axis=1)

print(f'\n  IC50 ({metric_name} direct): {df_ic50_dr.Reliable.sum()} reliable '
      f'of {len(df_ic50_dr)}')

# --- 7b. %Viable calculation — untreated baseline per Assay/Plate/Drug/Day ---
# Normalizing within the plate keeps each assay on its own baseline, which is
# what makes the two readouts comparable on one %Viable axis.
pv_col   = f'PctViable_{metric_name}'
untx_col = f'UntxMean_{metric_name}'
untx_mean = (
    df[df.Conc_orig == 0]
    .groupby(['Assay', 'Plate', 'Drug', 'Day'])[col]
    .mean()
    .reset_index()
    .rename(columns={col: untx_col})
)
df = df.drop(columns=[untx_col, pv_col], errors='ignore')   # idempotent re-run
df = df.merge(untx_mean, on=['Assay', 'Plate', 'Drug', 'Day'], how='left')
df[pv_col] = 100.0 * df[col] / df[untx_col]

# --- 7c. Absolute IC50: interpolated crossing at 50% of untreated ---
interp_records = []
for drug, assay in drug_assay_pairs:
    for day in fit_days:
        sub = _fit_window(_series_rows(drug, assay, day, pv_col).dropna(subset=['Conc']))
        if sub.empty:
            continue
        ic50_interp = interpolate_ic50(sub.Conc, sub[pv_col], target=50.0)
        interp_records.append({
            'Drug': drug, 'Assay': assay, 'Day': day,
            'Absolute_IC50_nM': ic50_interp, 'IC50_interp': ic50_interp,
        })
df_interp = pd.DataFrame(interp_records)
all_ic50_tables[f'{metric_name}_interpolated'] = df_interp

print(f'\n  Absolute IC50 at 50% untreated ({metric_name}): '
      f'{df_interp.IC50_interp.notna().sum()} of {len(df_interp)} crossed 50%')

# --- 7d. Fitted IC50: LL.4 midpoint between fitted upper/lower plateaus ---
fit_params_pv = {}
ic50_pv_records = []

for drug, assay in drug_assay_pairs:
    for day in fit_days:
        sub = _fit_window(_series_rows(drug, assay, day, pv_col).dropna(subset=['Conc']))
        if sub.empty:
            continue
        popt, r2 = fit_ll4_pct_viable(sub.Conc.values, sub[pv_col].values)
        rec = {'Drug': drug, 'Assay': assay, 'Day': day}
        if popt is not None:
            fit_params_pv[(drug, assay, day)] = popt
            rec.update({'IC50_nM': popt[3], 'Fitted_IC50_nM': popt[3],
                        'Midpoint_Response': (popt[1] + popt[2]) / 2.0,
                        'Slope': popt[0], 'Lower': popt[1], 'Upper': popt[2], 'R2': r2})
        else:
            rec.update({'IC50_nM': np.nan, 'Fitted_IC50_nM': np.nan,
                        'Midpoint_Response': np.nan, 'Slope': np.nan,
                        'Lower': np.nan, 'Upper': np.nan, 'R2': np.nan})
        ic50_pv_records.append(rec)

# --- 7e. %Viable dose-response plots with absolute IC50 markers ---
for day in fit_days:
    _all_vis_pv = df[df.Day == day].dropna(subset=[pv_col])
    _ylim_pv = _axis_limits(_all_vis_pv, pv_col, _xlo, _xhi)
    if _ylim_pv:
        _ylim_pv = (0, _ylim_pv[1])

    fig, ax = plt.subplots(figsize=(12, 7))
    _marked = False
    for drug, assay in drug_assay_pairs:
        sub = _series_rows(drug, assay, day, pv_col)
        if sub.empty:
            continue
        _plot_series(ax, drug, assay, sub, pv_col, fit_params_pv, day)

        row = df_interp[(df_interp.Drug == drug) & (df_interp.Assay == assay) &
                        (df_interp.Day == day)]
        if not row.empty and pd.notna(row.IC50_interp.values[0]):
            ax.plot(row.IC50_interp.values[0], 50, assay_markers[assay],
                    color=drug_colors[drug], markersize=11,
                    markeredgecolor='black', markeredgewidth=1.2, zorder=6)
            _marked = True

    ax.set_xscale('log')
    if _ylim_pv:
        ax.set_ylim(*_ylim_pv)
    ax.set_xlabel('Concentration (nM, log scale)')
    ax.set_ylabel(minfo['ylabel_pv'])
    ax.set_title(f'Day {day} — %Viable — Absolute IC50 at 50% untreated')
    ax.axhline(50, color='red', ls='--', alpha=0.7)
    ax.axhline(100, color='grey', ls='--', alpha=0.3)

    _extra = [Line2D([], [], color='red', ls='--', alpha=0.7, label='50% viable')]
    if _marked:
        _extra.append(marker_key_handle('Absolute IC50 (50% untx)'))
    draw_legend(ax, style_legend_handles() + _extra,
                fontsize=8, loc='lower left', ncol=2)
    plt.tight_layout()
    save_fig()

# --- 7f. %Viable dose-response plots with fitted-midpoint IC50 markers ---
for day in fit_days:
    _all_vis_pv = df[df.Day == day].dropna(subset=[pv_col])
    _ylim_pv = _axis_limits(_all_vis_pv, pv_col, _xlo, _xhi, pad=0.05)

    fig, ax = plt.subplots(figsize=(12, 7))
    _marked = False
    for drug, assay in drug_assay_pairs:
        sub = _series_rows(drug, assay, day, pv_col)
        if sub.empty:
            continue
        _plot_series(ax, drug, assay, sub, pv_col, fit_params_pv, day)

        key = (drug, assay, day)
        if key in fit_params_pv:
            popt = fit_params_pv[key]
            fitted_ic50 = popt[3]
            fitted_midpoint = (popt[1] + popt[2]) / 2.0
            if _xlo <= fitted_ic50 <= _xhi:
                ax.plot(fitted_ic50, fitted_midpoint, assay_markers[assay],
                        color=drug_colors[drug], markersize=11,
                        markeredgecolor='black', markeredgewidth=1.2, zorder=6)
                _marked = True

    ax.set_xscale('log')
    if _ylim_pv:
        ax.set_ylim(*_ylim_pv)
    ax.set_xlabel('Concentration (nM, log scale)')
    ax.set_ylabel(minfo['ylabel_pv'])
    ax.set_title(f'Day {day} — %Viable — Fitted IC50 at LL.4 plateau midpoint')

    _extra = []
    if _marked:
        _extra.append(marker_key_handle('Fitted IC50 (LL.4 midpoint)'))
    draw_legend(ax, style_legend_handles() + _extra,
                fontsize=8, loc='lower left', ncol=2)
    plt.tight_layout()
    save_fig()

df_ic50_pv = pd.DataFrame(ic50_pv_records)
df_ic50_pv['Reliable'] = (
    (df_ic50_pv['R2'] >= R2_THRESHOLD) &
    (df_ic50_pv['IC50_nM'] >= CONC_MIN * 0.5) &
    (df_ic50_pv['IC50_nM'] <= CONC_MAX * 2)
)
df_ic50_pv['Status'] = df_ic50_pv.apply(classify_ic50, axis=1)

print(f'  Fitted-midpoint IC50 (%Viable {metric_name}): {df_ic50_pv.Reliable.sum()} reliable '
      f'of {len(df_ic50_pv)}')

all_ic50_tables[f'{metric_name}_direct'] = df_ic50_dr
all_ic50_tables[f'{metric_name}_pctViable'] = df_ic50_pv

============================================================
=== METRIC: Confluence — assays: Castor, MTT ===
============================================================

--- Dose-response plots (raw Confluence) ---


  IC50 (Confluence direct): 7 reliable of 8

  Absolute IC50 at 50% untreated (Confluence): 6 of 8 crossed 50%

  Fitted-midpoint IC50 (%Viable Confluence): 6 reliable of 8

Dose-Response Fitting & Plotting — Suspect Replicates Dropped

Show code
# ===========================================================================
# 7g. Dose-response fitting & plotting — suspect replicates DROPPED
#     Same LL.4 model, same normalization, same plots as section 7, re-run on a
#     copy of the data with the wells section 3b flagged removed first.
#
#     Section 7 is deliberately left intact: nothing here rebinds `df`,
#     `fit_params_pv`, `df_ic50_pv`, `df_interp` or `all_ic50_tables`, which the
#     LL.5 fit, the trellises, the summary tables and the export all read. This
#     is a parallel view to compare against, not a replacement — so the two can
#     be read side by side and the effect of dropping is visible rather than
#     assumed.
# ===========================================================================
print('\n=== 7g. Dose-response with suspect replicates dropped ===')

# --- USER INPUT: what counts as droppable ----------------------------------
DROP_SUSPECT_WELLS = True           # 'one suspect well' groups: drop that well,
                                    # leaving the two replicates that agree.
SCATTERED_POLICY   = 'drop_worst'   # 'scattered' groups — no single well explains
                                    # the spread, so choose how to treat them:
                                    #   'drop_group' drop all of its wells; that
                                    #                concentration is lost from
                                    #                the curve entirely
                                    #   'drop_worst' drop only the well furthest
                                    #                from the median, leaving n=2
                                    #   'keep'       leave the group untouched
REGISTER_QC_TABLES_IN_EXPORTS = False   # True also feeds these IC50s into the
                                        # section-8 summary tables and the
                                        # section-10 CSV export. Off by default
                                        # so this section changes nothing
                                        # downstream.
# ---------------------------------------------------------------------------

assert 'df_replicate_qc' in globals(), 'Run section 3b (replicate variance) first.'
assert 'pv_col' in globals(), 'Run section 7 (dose-response fitting) first.'
assert SCATTERED_POLICY in ('drop_group', 'drop_worst', 'keep'), \
    f'SCATTERED_POLICY must be drop_group/drop_worst/keep, got {SCATTERED_POLICY!r}'

# --- Which wells does the section-3b verdict remove? -----------------------
# A well is addressed by (Assay, Plate, Day, Well): plate numbers repeat across
# assays and well IDs repeat across plates, so `Well` alone identifies nothing.
_drop_keys, _drop_log = set(), []

for _r in df_replicate_qc.itertuples(index=False):
    if _r.Verdict == 'one suspect well' and DROP_SUSPECT_WELLS:
        _wells, _why = [_r.Suspect_Well], 'suspect well'
    elif _r.Verdict == 'scattered' and SCATTERED_POLICY == 'drop_group':
        _wells, _why = _r.Wells.split('/'), 'scattered — whole group'
    elif _r.Verdict == 'scattered' and SCATTERED_POLICY == 'drop_worst':
        _wells, _why = [_r.Worst_Well], 'scattered — worst well'
    else:
        continue
    for _w in _wells:
        _drop_keys.add((_r.Assay, _r.Plate, _r.Day, _w))
        _drop_log.append({
            'Assay': _r.Assay, 'Plate': _r.Plate, 'Drug': _r.Drug, 'Day': _r.Day,
            'Conc': _r.Conc, 'Well': _w, 'CV_pct': _r.CV_pct,
            'CV_pct_drop1': _r.CV_pct_drop1, 'Verdict': _r.Verdict,
            'Reason': _why, 'Untreated': _r.Untreated,
            'Likely_Cause': _r.Likely_Cause,
        })

df_dropped_wells = pd.DataFrame(_drop_log)

# Section 3b judges every well, including wells at concentrations the 25%
# reversal filter or the x-window had already removed from the fit. Dropping one
# of those changes nothing, so mark which dropped wells were actually reaching
# the curve — otherwise a series shows wells dropped and a zero IC50 shift and
# looks like a bug.
_fit_wells = set(zip(*(_fit_window(df)[c] for c in
                       ['Assay', 'Plate', 'Day', 'Well'])))
if not df_dropped_wells.empty:
    df_dropped_wells['In_Fit_Window'] = [
        k in _fit_wells for k in zip(df_dropped_wells.Assay, df_dropped_wells.Plate,
                                     df_dropped_wells.Day, df_dropped_wells.Well)
    ]

# --- The cleaned copy -------------------------------------------------------
# `df` itself is never modified — every later section still sees all wells.
df_clean = df.copy()
df_clean['Dropped_Replicate_QC'] = [
    k in _drop_keys
    for k in zip(df_clean.Assay, df_clean.Plate, df_clean.Day, df_clean.Well)
]
_n_before = len(df_clean)
df_clean = df_clean[~df_clean.Dropped_Replicate_QC].copy()
_n_dropped = _n_before - len(df_clean)
_n_effective = (0 if df_dropped_wells.empty
                else int(df_dropped_wells.In_Fit_Window.sum()))

print(f'  Flagged groups in 3b : {(df_replicate_qc.Verdict != "OK").sum()} '
      f'of {len(df_replicate_qc)}')
print(f'  Wells dropped        : {_n_dropped} of {_n_before} '
      f'({100.0 * _n_dropped / _n_before:.1f}%)')
print(f'  ...reaching the fit  : {_n_effective} '
      f'({_n_dropped - _n_effective} were already outside it — reversal filter '
      f'or x-window — so they cannot move any curve)')
print(f'  Policy               : suspect wells={"drop" if DROP_SUSPECT_WELLS else "keep"}, '
      f'scattered={SCATTERED_POLICY}')

if df_dropped_wells.empty:
    display(Markdown('_No wells met the drop criteria — the curves below are '
                     'identical to section 7._'))
else:
    display(Markdown('### Wells dropped before fitting'))
    display(Markdown(
        '`In_Fit_Window` is whether the well was still reaching the section-7 '
        'fit; a `False` here means the reversal filter had already removed its '
        'concentration, so dropping it cannot change any IC50.'))
    display(
        df_dropped_wells.sort_values(['Assay', 'Plate', 'Drug', 'Conc'])
        .style.hide(axis='index')
        .format({'CV_pct': '{:.1f}', 'CV_pct_drop1': '{:.1f}', 'Conc': '{:g}'},
                na_rep='—')
        .map(lambda v: 'color: #757575' if v == False else None,
             subset=['In_Fit_Window'])
    )

# --- %Viable, recomputed on the cleaned data --------------------------------
# This cannot be inherited from section 7. If a dropped well was an untreated
# well, the baseline that every point on that curve is divided by has changed,
# so %Viable must be rebuilt from the surviving wells. The columns keep their
# section-7 names because this is a different DataFrame — `df` is untouched.
_untx_clean = (
    df_clean[df_clean.Conc_orig == 0]
    .groupby(['Assay', 'Plate', 'Drug', 'Day'])[col]
    .mean().reset_index().rename(columns={col: untx_col})
)
df_clean = df_clean.drop(columns=[untx_col, pv_col], errors='ignore')
df_clean = df_clean.merge(_untx_clean, on=['Assay', 'Plate', 'Drug', 'Day'], how='left')
df_clean[pv_col] = 100.0 * df_clean[col] / df_clean[untx_col]


def _series_rows_clean(drug, assay, day, value_col):
    """`_series_rows`, reading the QC-cleaned frame instead of `df`."""
    sub = df_clean[(df_clean.Drug == drug) & (df_clean.Assay == assay) &
                   (df_clean.Day == day)].copy()
    return sub.dropna(subset=[value_col])


# Every other section-7 helper (`_analysis_rows`, `_fit_window`, `_axis_limits`,
# `_plot_series`) already takes its rows as an argument, so they are reused as
# they are — the reversal filter, the x-window and the plot styling stay
# identical and only the underlying wells differ.

_QC_TAG = 'suspect replicates dropped'

# --- 7g-a. Dose-response fits & plots (raw metric) --------------------------
print(f'\n--- Dose-response plots (raw {metric_name}, {_QC_TAG}) ---')
fit_params_dr_qc = {}
ic50_dr_qc_records = []

for drug, assay in drug_assay_pairs:
    for day in fit_days:
        sub = _fit_window(_series_rows_clean(drug, assay, day, col).dropna(subset=['Conc']))
        if sub.empty:
            continue
        popt, r2 = fit_ll4(sub.Conc.values, sub[col].values)
        rec = {'Drug': drug, 'Assay': assay, 'Day': day}
        if popt is not None:
            fit_params_dr_qc[(drug, assay, day)] = popt
            rec.update({'IC50_nM': popt[3], 'Slope': popt[0],
                        'Lower': popt[1], 'Upper': popt[2], 'R2': r2})
        else:
            rec.update({'IC50_nM': np.nan, 'Slope': np.nan,
                        'Lower': np.nan, 'Upper': np.nan, 'R2': np.nan})
        ic50_dr_qc_records.append(rec)

for day in fit_days:
    fig, ax = plt.subplots(figsize=(12, 7))
    axis_for, sides = raw_metric_axes(ax)
    _split = len(sides) > 1

    for drug, assay in drug_assay_pairs:
        sub = _series_rows_clean(drug, assay, day, col)
        if sub.empty:
            continue
        _plot_series(axis_for[assay], drug, assay, sub, col, fit_params_dr_qc, day)

    for _ax, _side, _assays in sides:
        _vis = df_clean[(df_clean.Day == day) & df_clean.Assay.isin(_assays)].dropna(subset=[col])
        _lim = _axis_limits(_vis, col, _xlo, _xhi)
        if _lim:
            _ax.set_ylim(*_lim)
        style_raw_axis(_ax, _side, _assays, _split)

    ax.set_xscale('log')
    ax.set_xlabel('Concentration (nM, log scale)')
    ax.set_title(f'Day {day} — Dose Response (raw metric, {_QC_TAG})')
    draw_legend(sides[-1][0],
                style_legend_handles(assay_labels=assay_axis_labels(sides)),
                fontsize=8, loc='lower left', ncol=2)
    plt.tight_layout()
    save_fig()

df_ic50_dr_qc = pd.DataFrame(ic50_dr_qc_records)
df_ic50_dr_qc['Reliable'] = (
    (df_ic50_dr_qc['R2'] >= R2_THRESHOLD) &
    (df_ic50_dr_qc['IC50_nM'] >= CONC_MIN * 0.5) &
    (df_ic50_dr_qc['IC50_nM'] <= CONC_MAX * 2)
)
df_ic50_dr_qc['Status'] = df_ic50_dr_qc.apply(classify_ic50, axis=1)
print(f'\n  IC50 ({metric_name} direct, {_QC_TAG}): '
      f'{df_ic50_dr_qc.Reliable.sum()} reliable of {len(df_ic50_dr_qc)}')

# --- 7g-b. Absolute IC50: interpolated crossing at 50% of untreated ---------
interp_qc_records = []
for drug, assay in drug_assay_pairs:
    for day in fit_days:
        sub = _fit_window(_series_rows_clean(drug, assay, day, pv_col).dropna(subset=['Conc']))
        if sub.empty:
            continue
        ic50_interp = interpolate_ic50(sub.Conc, sub[pv_col], target=50.0)
        interp_qc_records.append({
            'Drug': drug, 'Assay': assay, 'Day': day,
            'Absolute_IC50_nM': ic50_interp, 'IC50_interp': ic50_interp,
        })
df_interp_qc = pd.DataFrame(interp_qc_records)
print(f'  Absolute IC50 at 50% untreated ({_QC_TAG}): '
      f'{df_interp_qc.IC50_interp.notna().sum()} of {len(df_interp_qc)} crossed 50%')

# --- 7g-c. Fitted IC50: LL.4 midpoint between fitted plateaus ---------------
fit_params_pv_qc = {}
ic50_pv_qc_records = []

for drug, assay in drug_assay_pairs:
    for day in fit_days:
        sub = _fit_window(_series_rows_clean(drug, assay, day, pv_col).dropna(subset=['Conc']))
        if sub.empty:
            continue
        popt, r2 = fit_ll4_pct_viable(sub.Conc.values, sub[pv_col].values)
        rec = {'Drug': drug, 'Assay': assay, 'Day': day}
        if popt is not None:
            fit_params_pv_qc[(drug, assay, day)] = popt
            rec.update({'IC50_nM': popt[3], 'Fitted_IC50_nM': popt[3],
                        'Midpoint_Response': (popt[1] + popt[2]) / 2.0,
                        'Slope': popt[0], 'Lower': popt[1], 'Upper': popt[2], 'R2': r2})
        else:
            rec.update({'IC50_nM': np.nan, 'Fitted_IC50_nM': np.nan,
                        'Midpoint_Response': np.nan, 'Slope': np.nan,
                        'Lower': np.nan, 'Upper': np.nan, 'R2': np.nan})
        ic50_pv_qc_records.append(rec)

# --- 7g-d. %Viable plots with absolute IC50 markers ------------------------
for day in fit_days:
    _all_vis_pv = df_clean[df_clean.Day == day].dropna(subset=[pv_col])
    _ylim_pv = _axis_limits(_all_vis_pv, pv_col, _xlo, _xhi)
    if _ylim_pv:
        _ylim_pv = (0, _ylim_pv[1])

    fig, ax = plt.subplots(figsize=(12, 7))
    _marked = False
    for drug, assay in drug_assay_pairs:
        sub = _series_rows_clean(drug, assay, day, pv_col)
        if sub.empty:
            continue
        _plot_series(ax, drug, assay, sub, pv_col, fit_params_pv_qc, day)

        row = df_interp_qc[(df_interp_qc.Drug == drug) & (df_interp_qc.Assay == assay) &
                           (df_interp_qc.Day == day)]
        if not row.empty and pd.notna(row.IC50_interp.values[0]):
            ax.plot(row.IC50_interp.values[0], 50, assay_markers[assay],
                    color=drug_colors[drug], markersize=11,
                    markeredgecolor='black', markeredgewidth=1.2, zorder=6)
            _marked = True

    ax.set_xscale('log')
    if _ylim_pv:
        ax.set_ylim(*_ylim_pv)
    ax.set_xlabel('Concentration (nM, log scale)')
    ax.set_ylabel(minfo['ylabel_pv'])
    ax.set_title(f'Day {day} — %Viable — Absolute IC50 at 50% untreated ({_QC_TAG})')
    ax.axhline(50, color='red', ls='--', alpha=0.7)
    ax.axhline(100, color='grey', ls='--', alpha=0.3)

    _extra = [Line2D([], [], color='red', ls='--', alpha=0.7, label='50% viable')]
    if _marked:
        _extra.append(marker_key_handle('Absolute IC50 (50% untx)'))
    draw_legend(ax, style_legend_handles() + _extra,
                fontsize=8, loc='lower left', ncol=2)
    plt.tight_layout()
    save_fig()

# --- 7g-e. %Viable plots with fitted-midpoint IC50 markers -----------------
for day in fit_days:
    _all_vis_pv = df_clean[df_clean.Day == day].dropna(subset=[pv_col])
    _ylim_pv = _axis_limits(_all_vis_pv, pv_col, _xlo, _xhi, pad=0.05)

    fig, ax = plt.subplots(figsize=(12, 7))
    _marked = False
    for drug, assay in drug_assay_pairs:
        sub = _series_rows_clean(drug, assay, day, pv_col)
        if sub.empty:
            continue
        _plot_series(ax, drug, assay, sub, pv_col, fit_params_pv_qc, day)

        key = (drug, assay, day)
        if key in fit_params_pv_qc:
            popt = fit_params_pv_qc[key]
            fitted_ic50 = popt[3]
            fitted_midpoint = (popt[1] + popt[2]) / 2.0
            if _xlo <= fitted_ic50 <= _xhi:
                ax.plot(fitted_ic50, fitted_midpoint, assay_markers[assay],
                        color=drug_colors[drug], markersize=11,
                        markeredgecolor='black', markeredgewidth=1.2, zorder=6)
                _marked = True

    ax.set_xscale('log')
    if _ylim_pv:
        ax.set_ylim(*_ylim_pv)
    ax.set_xlabel('Concentration (nM, log scale)')
    ax.set_ylabel(minfo['ylabel_pv'])
    ax.set_title(f'Day {day} — %Viable — Fitted IC50 at LL.4 plateau midpoint ({_QC_TAG})')

    _extra = []
    if _marked:
        _extra.append(marker_key_handle('Fitted IC50 (LL.4 midpoint)'))
    draw_legend(ax, style_legend_handles() + _extra,
                fontsize=8, loc='lower left', ncol=2)
    plt.tight_layout()
    save_fig()

df_ic50_pv_qc = pd.DataFrame(ic50_pv_qc_records)
df_ic50_pv_qc['Reliable'] = (
    (df_ic50_pv_qc['R2'] >= R2_THRESHOLD) &
    (df_ic50_pv_qc['IC50_nM'] >= CONC_MIN * 0.5) &
    (df_ic50_pv_qc['IC50_nM'] <= CONC_MAX * 2)
)
df_ic50_pv_qc['Status'] = df_ic50_pv_qc.apply(classify_ic50, axis=1)
print(f'  Fitted-midpoint IC50 (%Viable, {_QC_TAG}): '
      f'{df_ic50_pv_qc.Reliable.sum()} reliable of {len(df_ic50_pv_qc)}')

all_ic50_tables_qc = {
    f'{metric_name}_direct_QCdropped':      df_ic50_dr_qc,
    f'{metric_name}_pctViable_QCdropped':   df_ic50_pv_qc,
    f'{metric_name}_interpolated_QCdropped': df_interp_qc,
}
if REGISTER_QC_TABLES_IN_EXPORTS:
    all_ic50_tables.update(all_ic50_tables_qc)
    print(f'  Registered {len(all_ic50_tables_qc)} QC tables into all_ic50_tables '
          f'(they will appear in the section-8 summaries and section-10 export).')

# --- 7g-f. What did dropping actually change? ------------------------------
# The point of the section: the two IC50 sets side by side. A large shift means
# the fit was leaning on a well that section 3b called suspect.
_KEYS = ['Drug', 'Assay', 'Day']


def _delta_pct(new, old):
    return np.where((old > 0) & np.isfinite(new), 100.0 * (new - old) / old, np.nan)


_cmp_fit = (
    df_ic50_pv[_KEYS + ['IC50_nM', 'R2', 'Reliable']]
    .rename(columns={'IC50_nM': 'IC50_all', 'R2': 'R2_all', 'Reliable': 'Rel_all'})
    .merge(df_ic50_pv_qc[_KEYS + ['IC50_nM', 'R2', 'Reliable']]
           .rename(columns={'IC50_nM': 'IC50_QC', 'R2': 'R2_QC', 'Reliable': 'Rel_QC'}),
           on=_KEYS, how='outer')
)
_cmp_fit['IC50_delta_%'] = _delta_pct(_cmp_fit.IC50_QC, _cmp_fit.IC50_all)
_cmp_fit['R2_delta'] = _cmp_fit.R2_QC - _cmp_fit.R2_all

_cmp_int = (
    df_interp[_KEYS + ['IC50_interp']].rename(columns={'IC50_interp': 'Abs_all'})
    .merge(df_interp_qc[_KEYS + ['IC50_interp']].rename(columns={'IC50_interp': 'Abs_QC'}),
           on=_KEYS, how='outer')
)
_cmp_int['Abs_delta_%'] = _delta_pct(_cmp_int.Abs_QC, _cmp_int.Abs_all)

_cmp = _cmp_fit.merge(_cmp_int, on=_KEYS, how='outer')

# How many wells each series lost, and how many of those were reaching the fit.
# `Dropped_in_fit == 0` must come with a zero delta — that is the sanity check on
# the whole section. `Wells_dropped` alone does not: a series can lose a whole
# group at a concentration the reversal filter had already excluded and be
# completely unaffected.
_lost_all = ({} if df_dropped_wells.empty
             else df_dropped_wells.groupby(_KEYS).size().to_dict())
_lost_fit = ({} if df_dropped_wells.empty
             else df_dropped_wells[df_dropped_wells.In_Fit_Window]
             .groupby(_KEYS).size().to_dict())
_cmp['Wells_dropped'] = [_lost_all.get(k, 0) for k in
                         zip(_cmp.Drug, _cmp.Assay, _cmp.Day)]
_cmp['Dropped_in_fit'] = [_lost_fit.get(k, 0) for k in
                          zip(_cmp.Drug, _cmp.Assay, _cmp.Day)]

display(Markdown('### IC50 before vs after dropping suspect replicates'))
display(Markdown(
    '`IC50_all` / `Abs_all` are section 7 (every well); `_QC` are the same '
    'quantities refitted without the dropped wells. `Dropped_in_fit` is the '
    'number of dropped wells that were still reaching the fit — a series with '
    '`0` there must show a delta of `+0.0`, and a large delta where only one '
    'well went means the LL.4 fit was being steered by that single well. '
    '`Abs_delta_%` moves furthest when a whole concentration is dropped, since '
    'the 50% crossing then has to be interpolated across a wider gap.'))
display(
    _cmp.sort_values('IC50_delta_%', key=lambda s: s.abs(), ascending=False)
    .style.hide(axis='index')
    .format({'IC50_all': '{:.4g}', 'IC50_QC': '{:.4g}', 'IC50_delta_%': '{:+.1f}',
             'R2_all': '{:.3f}', 'R2_QC': '{:.3f}', 'R2_delta': '{:+.3f}',
             'Abs_all': '{:.4g}', 'Abs_QC': '{:.4g}', 'Abs_delta_%': '{:+.1f}'},
            na_rep='—')
    .map(lambda v: 'font-weight: bold' if isinstance(v, (int, float, np.floating))
         and np.isfinite(v) and abs(v) >= 20 else None,
         subset=['IC50_delta_%', 'Abs_delta_%'])
    .map(lambda v: 'color: #b00020' if v == False else None, subset=['Rel_all', 'Rel_QC'])
)

print(f'\n  Series whose fitted IC50 moved >20%: '
      f'{int((_cmp["IC50_delta_%"].abs() > 20).sum())} of {len(_cmp)}')
print(f'  Series that became reliable: '
      f'{int((~_cmp.Rel_all.fillna(False) & _cmp.Rel_QC.fillna(False)).sum())}; '
      f'lost reliability: '
      f'{int((_cmp.Rel_all.fillna(False) & ~_cmp.Rel_QC.fillna(False)).sum())}')
print('  `df` and every section-7 table are unchanged — later sections still use all wells.')

=== 7g. Dose-response with suspect replicates dropped ===
  Flagged groups in 3b : 3 of 80
  Wells dropped        : 3 of 240 (1.2%)
  ...reaching the fit  : 1 (2 were already outside it — reversal filter or x-window — so they cannot move any curve)
  Policy               : suspect wells=drop, scattered=drop_worst

Wells dropped before fitting

In_Fit_Window is whether the well was still reaching the section-7 fit; a False here means the reversal filter had already removed its concentration, so dropping it cannot change any IC50.

Assay Plate Drug Day Conc Well CV_pct CV_pct_drop1 Verdict Reason Untreated Likely_Cause In_Fit_Window
Castor 1 63GE8 7 49.4 B4 26.2 7.5 one suspect well suspect well False Row B biased (-23% plate-wide) True
Castor 1 63GE8 7 222.2 D3 64.9 32.7 scattered scattered — worst well False False
Castor 1 63GE8 7 1000 D2 63.6 45.3 scattered scattered — worst well False False

--- Dose-response plots (raw Confluence, suspect replicates dropped) ---


  IC50 (Confluence direct, suspect replicates dropped): 7 reliable of 8
  Absolute IC50 at 50% untreated (suspect replicates dropped): 6 of 8 crossed 50%

  Fitted-midpoint IC50 (%Viable, suspect replicates dropped): 6 reliable of 8

IC50 before vs after dropping suspect replicates

IC50_all / Abs_all are section 7 (every well); _QC are the same quantities refitted without the dropped wells. Dropped_in_fit is the number of dropped wells that were still reaching the fit — a series with 0 there must show a delta of +0.0, and a large delta where only one well went means the LL.4 fit was being steered by that single well. Abs_delta_% moves furthest when a whole concentration is dropped, since the 50% crossing then has to be interpolated across a wider gap.

Drug Assay Day IC50_all R2_all Rel_all IC50_QC R2_QC Rel_QC IC50_delta_% R2_delta Abs_all Abs_QC Abs_delta_% Wells_dropped Dropped_in_fit
63GE8 Castor 7 0.08313 0.578 True 0.0688 0.584 True -17.2 +0.006 — — — 3 1
63GE8 MTT 7 0.1366 0.905 True 0.1366 0.905 True +0.0 +0.000 22.69 22.69 +0.0 0 0
64GE6 Castor 7 0.09133 0.911 True 0.09133 0.911 True +0.0 +0.000 0.8566 0.8566 +0.0 0 0
64GE6 MTT 7 0.2644 0.934 True 0.2644 0.934 True +0.0 +0.000 24.17 24.17 +0.0 0 0
64VPE6 Castor 7 0.693 0.935 True 0.693 0.935 True +0.0 +0.000 2.357 2.357 +0.0 0 0
64VPE6 MTT 7 0.5068 0.923 True 0.5068 0.923 True +0.0 +0.000 4.275 4.275 +0.0 0 0
IsoGE8 Castor 7 49.61 -0.080 False 49.61 -0.080 False +0.0 +0.000 — — — 0 0
IsoGE8 MTT 7 5715 0.833 False 5715 0.833 False +0.0 +0.000 93.73 93.73 +0.0 0 0

  Series whose fitted IC50 moved >20%: 0 of 8
  Series that became reliable: 0; lost reliability: 0
  `df` and every section-7 table are unchanged — later sections still use all wells.

Trellis: Test vs Isotype by Linker-Payload

Show code
# ===========================================================================
# 7g. Trellis: All Antibodies per Linker-Payload, one subplot per Day
#      Colour = drug, linestyle/marker = assay (both assays overlaid).
# ===========================================================================
print('\n=== Trellis: dose-response by LP (all antibodies, both assays) ===')

# Build lp -> {ab: drug_name} from all drugs
_lp_ab_map = {}
for d in all_drugs:
    ab, lp = _split_ab_lp(d)
    if lp is not None:
        _lp_ab_map.setdefault(lp, {})[ab] = d

_all_lps  = sorted(_lp_ab_map.keys())
_markers  = ['o', 's', '^', 'D', 'v', 'P', '*', 'X', 'h', '<']
print(f'  Linker-payloads: {_all_lps}')
for lp, abmap in sorted(_lp_ab_map.items()):
    print(f'    {lp}: {sorted(abmap.keys())}')

# Derive TEST_AB / ISO_AB matched pairs for section 7g (TI trellis)
_test_drugs  = {lp: ab_drug[TEST_AB] for lp, ab_drug in _lp_ab_map.items() if TEST_AB in ab_drug}
_iso_drugs   = {lp: ab_drug[ISO_AB]  for lp, ab_drug in _lp_ab_map.items() if ISO_AB  in ab_drug}
_matched_lps = sorted(set(_test_drugs) & set(_iso_drugs))

metric_name = 'Confluence'
pv_col = f'PctViable_{metric_name}'


def _assays_for(drug):
    """Assays this drug was actually measured with, in canonical order."""
    return [a for (d, a) in drug_assay_pairs if d == drug]


for lp in _all_lps:
    ab_drug    = _lp_ab_map[lp]
    abs_sorted = sorted(ab_drug.keys())
    lp_drugs   = [ab_drug[ab] for ab in abs_sorted]

    # Pre-scan: y-range within SCALE_MIN_CONC–SCALE_MAX_CONC, excluding SCALE_EXCL_ABS
    all_vals = []
    for day in fit_days:
        for ab, drug in ab_drug.items():
            if ab in SCALE_EXCL_ABS:
                continue
            for assay in _assays_for(drug):
                vis = _fit_window(_series_rows(drug, assay, day, pv_col))
                if vis.empty:
                    continue
                means = vis.groupby('Conc')[pv_col].mean()
                sds   = vis.groupby('Conc')[pv_col].std().fillna(0)
                all_vals.extend((means + sds).tolist())
                all_vals.extend((means - sds).tolist())
    if not all_vals:
        continue
    y_lo   = min(all_vals)
    y_hi   = max(all_vals)
    margin = (y_hi - y_lo) * 0.08
    ylim   = (y_lo - margin, y_hi + margin)

    ncols = len(fit_days)
    fig, axes = plt.subplots(1, ncols, figsize=(4.6 * ncols, 5), sharey=True)
    if ncols == 1:
        axes = [axes]

    lp_assays = sorted({a for drug in lp_drugs for a in _assays_for(drug)},
                       key=all_assays.index)
    for ax, day in zip(axes, fit_days):
        for ab in abs_sorted:
            drug = ab_drug[ab]
            for assay in _assays_for(drug):
                sub = _series_rows(drug, assay, day, pv_col)
                if sub.empty:
                    continue
                _plot_series(ax, drug, assay, sub, pv_col, fit_params_pv, day)

        ax.set_xscale('log')
        ax.set_ylim(ylim)
        ax.set_title(f'Day {day}', fontsize=11)
        ax.set_xlabel('Conc (nM)')
        ax.axhline(50, color='red', ls=':', alpha=0.5)
        ax.axhline(100, color='grey', ls=':', alpha=0.3)
        if ax is axes[0]:
            ax.set_ylabel(minfo['ylabel_pv'])
        draw_legend(ax, style_legend_handles(drugs=lp_drugs, assays=lp_assays),
                    fontsize=7, loc='lower left')

    fig.suptitle(f'{lp} — %Viable ({" + ".join(lp_assays)})', fontsize=14, y=1.02)
    plt.tight_layout()
    save_fig()

=== Trellis: dose-response by LP (all antibodies, both assays) ===
  Linker-payloads: []

Trellis: Therapeutic Index by Concentration

Show code
# ===========================================================================
# 7g. Trellis: Therapeutic Index per concentration, per LP x Day
#     TI = %Viable(ISO_AB) / %Viable(each test AB), computed WITHIN an assay.
# ===========================================================================
print('\n=== Trellis: TI by concentration (all test ABs vs ISO_AB, per assay) ===')

metric_name = 'Confluence'
pv_col = f'PctViable_{metric_name}'

_ti_data    = {}   # (lp, ab, assay, day) -> (conc_array, ti_array)
_lp_yranges = {}   # lp -> (ymin, ymax)

for lp in _all_lps:
    ab_drug  = _lp_ab_map[lp]
    if ISO_AB not in ab_drug:
        continue
    iso_drug = ab_drug[ISO_AB]
    test_abs = sorted(ab_drug.keys())  # includes ISO_AB so TI=1 is shown
    if not test_abs:
        continue

    all_ti = []
    for ab in test_abs:
        drug = ab_drug[ab]
        # Only assays where BOTH the test AB and its isotype were measured.
        for assay in [a for a in _assays_for(drug) if a in _assays_for(iso_drug)]:
            for day in fit_days:
                sub_t = _analysis_rows(_series_rows(drug, assay, day, pv_col))
                sub_i = _analysis_rows(_series_rows(iso_drug, assay, day, pv_col))
                if sub_t.empty or sub_i.empty:
                    continue
                mean_t = sub_t.groupby('Conc')[pv_col].mean()
                mean_i = sub_i.groupby('Conc')[pv_col].mean()
                shared = sorted(set(mean_t.index) & set(mean_i.index))
                if not shared:
                    continue
                concs = np.array(shared)
                ti    = np.array([mean_i[c] / mean_t[c] if mean_t[c] != 0 else np.nan
                                  for c in shared])
                _ti_data[(lp, ab, assay, day)] = (concs, ti)
                all_ti.extend(ti[np.isfinite(ti)].tolist())

    if all_ti:
        _lp_yranges[lp] = (min(all_ti), max(all_ti))

if not _lp_yranges:
    print(f'  No TI data for {metric_name}')
else:
    lps_with_ti = [lp for lp in _all_lps if lp in _lp_yranges]
    nrows = len(lps_with_ti)
    ncols = len(fit_days)
    fig, axes = plt.subplots(nrows, ncols,
                             figsize=(3.9 * ncols, 3.5 * nrows),
                             squeeze=False)

    for row, lp in enumerate(lps_with_ti):
        ab_drug  = _lp_ab_map[lp]
        test_abs = sorted(ab_drug.keys())  # includes ISO_AB so TI=1 is shown
        lp_drugs = [ab_drug[ab] for ab in test_abs]
        y_lo, y_hi = _lp_yranges[lp]
        margin = (y_hi - y_lo) * 0.1 if y_hi > y_lo else 0.5
        ylim = (y_lo - margin, y_hi + margin)
        lp_assays = sorted({a for (l, ab, a, d) in _ti_data if l == lp},
                           key=all_assays.index)

        for col_i, day in enumerate(fit_days):
            ax = axes[row][col_i]
            for ab in test_abs:
                drug = ab_drug[ab]
                for assay in lp_assays:
                    key = (lp, ab, assay, day)
                    if key not in _ti_data:
                        continue
                    concs, ti = _ti_data[key]
                    ax.plot(concs, ti,
                            marker=assay_markers[assay],
                            linestyle=assay_linestyles[assay],
                            color=drug_colors[drug], markersize=5,
                            linewidth=assay_linewidths[assay],
                            markeredgecolor='white', markeredgewidth=0.5)

            ax.axhline(1, color='red', ls='--', alpha=0.6)
            ax.set_xscale('log')
            ax.set_ylim(ylim)
            draw_legend(ax, style_legend_handles(drugs=lp_drugs, assays=lp_assays),
                        fontsize=6, loc='best')

            if row == 0:
                ax.set_title(f'Day {day}', fontsize=11)
            if col_i == 0:
                ax.set_ylabel(f'{lp}\nTI ({ISO_AB} / AB)', fontsize=9)
            else:
                ax.set_yticklabels([])
            if row == nrows - 1:
                ax.set_xlabel('Conc (nM)')
            else:
                ax.set_xticklabels([])

    fig.suptitle(f'Therapeutic Index — %Viable ({metric_name})\n'
                 f'TI = {ISO_AB} / each test AB, within assay   (red dashed = TI 1)',
                 fontsize=14, y=1.02)
    plt.tight_layout()
    save_fig()

=== Trellis: TI by concentration (all test ABs vs ISO_AB, per assay) ===
  No TI data for Confluence

IC50 Summary Tables (Curve Fit)

Show code
# ===========================================================================
# 8. Summary tables — side-by-side comparison (plain display version)
#    Tables are indexed by Drug x Assay so the two readouts stay distinguishable.
# ===========================================================================
pd.set_option('display.float_format', lambda x: f'{x:.3g}')  # applies to all displayed tables
sig3 = lambda x: f'{x:.3g}'  # same 3-sig-fig rule, for use inside .style (doesn't inherit the option above)

_ID_COLS = ['Drug', 'Assay']


def _display_pivot(pivot):
    """Return a display-only copy with 'Day' columns prefixed, leaving the original pivot untouched."""
    disp = pivot.copy()
    disp = disp.rename(columns={c: f'Day {c}' for c in disp.columns if c not in _ID_COLS})
    return disp


display(Markdown('## IC50 SUMMARY — COMPARISON'))

for label, df_tbl in all_ic50_tables.items():
    if 'interpolated' in label:
        continue  # handled separately below
    # Reliable IC50 pivot
    pivot = df_tbl[df_tbl.Reliable].pivot_table(
        index=_ID_COLS, columns='Day', values='IC50_nM')
    pivot = pivot.rename_axis(columns=None).reset_index()

    display(Markdown(f'### {label} — IC50 (nM), reliable only (R²≥{R2_THRESHOLD}, in range)'))
    if pivot.empty:
        display(Markdown('_(no reliable fits)_'))
    else:
        display(_display_pivot(pivot).style.hide(axis='index'))

# --- Absolute IC50 tables (direct crossing at 50% of untreated) ---
display(Markdown('## ABSOLUTE IC50 AT 50% OF UNTREATED (no curve fitting)'))

for metric_name in ['Confluence']:
    label = f'{metric_name}_interpolated'
    df_interp_tbl = all_ic50_tables[label]
    pivot = df_interp_tbl.pivot_table(
        index=_ID_COLS, columns='Day', values='IC50_interp')
    pivot = pivot.rename_axis(columns=None).reset_index()

    display(Markdown(f'### {metric_name} — Absolute IC50 (nM) at 50% of untreated'))
    display(_display_pivot(pivot).style.hide(axis='index'))

# Side-by-side: absolute 50%-untreated crossing vs fitted midpoint
display(Markdown('### Absolute 50%-untreated vs fitted-midpoint IC50 (per Drug × Assay)'))
interp = all_ic50_tables['Confluence_interpolated'][_ID_COLS + ['Day', 'IC50_interp']]
fitted = all_ic50_tables['Confluence_pctViable']
fitted_rel = fitted[fitted.Reliable][_ID_COLS + ['Day', 'IC50_nM']].rename(
    columns={'IC50_nM': 'IC50_LL4'})
compare = interp.merge(fitted_rel, on=_ID_COLS + ['Day'], how='outer')
compare = compare.dropna(subset=['IC50_interp', 'IC50_LL4'], how='all')
compare['Diff_%'] = ((compare.IC50_interp - compare.IC50_LL4) /
                      compare.IC50_LL4 * 100).round(1)

compare = compare.set_index(_ID_COLS + ['Day']).sort_index()

display(
    compare.style
    .format({'IC50_interp': sig3, 'IC50_LL4': sig3, 'Diff_%': '{:.1f}'}, na_rep='—')
)

# --- Cross-assay agreement: same drug, one readout against the other ---
if len(all_assays) > 1:
    display(Markdown('## CROSS-ASSAY AGREEMENT — fitted IC50 (nM) per drug'))
    for tbl_label, value_col, src in [
        ('Fitted LL.4 midpoint (reliable only)', 'IC50_nM',
         all_ic50_tables['Confluence_pctViable'].pipe(lambda t: t[t.Reliable])),
        ('Absolute 50% of untreated', 'IC50_interp',
         all_ic50_tables['Confluence_interpolated']),
    ]:
        x = src.pivot_table(index=['Drug', 'Day'], columns='Assay', values=value_col)
        x = x.rename_axis(columns=None)
        present = [a for a in all_assays if a in x.columns]
        if len(present) == 2:
            a0, a1 = present
            x[f'{a0}/{a1}'] = x[a0] / x[a1]
        display(Markdown(f'### {tbl_label}'))
        if x.empty:
            display(Markdown('_(no values)_'))
        else:
            display(x.style.format(sig3, na_rep='—'))

IC50 SUMMARY — COMPARISON

Confluence_direct — IC50 (nM), reliable only (R²≥0.5, in range)

Drug Assay Day 7
63GE8 Castor 0.050689
63GE8 MTT 0.091955
64GE6 Castor 0.074488
64GE6 MTT 0.207777
64VPE6 Castor 0.613688
64VPE6 MTT 0.492295
IsoGE8 Castor 28.833824

Confluence_pctViable — IC50 (nM), reliable only (R²≥0.5, in range)

Drug Assay Day 7
63GE8 Castor 0.083133
63GE8 MTT 0.136591
64GE6 Castor 0.091326
64GE6 MTT 0.264357
64VPE6 Castor 0.692950
64VPE6 MTT 0.506750

ABSOLUTE IC50 AT 50% OF UNTREATED (no curve fitting)

Confluence — Absolute IC50 (nM) at 50% of untreated

Drug Assay Day 7
63GE8 MTT 22.693997
64GE6 Castor 0.856597
64GE6 MTT 24.174933
64VPE6 Castor 2.357239
64VPE6 MTT 4.275096
IsoGE8 MTT 93.729343

Absolute 50%-untreated vs fitted-midpoint IC50 (per Drug × Assay)

      IC50_interp IC50_LL4 Diff_%
Drug Assay Day      
63GE8 Castor 7 — 0.0831 —
MTT 7 22.7 0.137 16514.6
64GE6 Castor 7 0.857 0.0913 838.0
MTT 7 24.2 0.264 9044.8
64VPE6 Castor 7 2.36 0.693 240.2
MTT 7 4.28 0.507 743.6
IsoGE8 MTT 7 93.7 — —

CROSS-ASSAY AGREEMENT — fitted IC50 (nM) per drug

Fitted LL.4 midpoint (reliable only)

    Castor MTT Castor/MTT
Drug Day      
63GE8 7 0.0831 0.137 0.609
64GE6 7 0.0913 0.264 0.345
64VPE6 7 0.693 0.507 1.37

Absolute 50% of untreated

    Castor MTT Castor/MTT
Drug Day      
63GE8 7 — 22.7 —
64GE6 7 0.857 24.2 0.0354
64VPE6 7 2.36 4.28 0.551
IsoGE8 7 — 93.7 —

Absolute IC50 Tables

Show code
# ===========================================================================
# 9. IC50 bar plot (reliable, %Viable) — one bar per Drug × Assay
#    Bar colour = drug, hatch = assay, matching the curve plots' encoding.
# ===========================================================================
for label, df_tbl in [
    ('Confluence %Viable', all_ic50_tables['Confluence_pctViable']),
]:
    plot_df = df_tbl[df_tbl.Reliable].dropna(subset=['IC50_nM']).copy()
    if plot_df.empty:
        print(f'{label}: no reliable IC50 values to plot.')
        continue

    pairs = [(d, a) for (d, a) in drug_assay_pairs
             if not plot_df[(plot_df.Drug == d) & (plot_df.Assay == a)].empty]
    day_vals = sorted(plot_df.Day.unique())
    x = np.arange(len(pairs))
    width = 0.8 / max(len(day_vals), 1)

    fig, ax = plt.subplots(figsize=(max(8, 1.6 * len(pairs)), 6))
    for i, day in enumerate(day_vals):
        vals = []
        for (d, a) in pairs:
            row = plot_df[(plot_df.Drug == d) & (plot_df.Assay == a) &
                          (plot_df.Day == day)]
            vals.append(row.IC50_nM.values[0] if len(row) else np.nan)
        bars = ax.bar(x + i * width, vals, width,
                      color=[drug_colors[d] for (d, _) in pairs],
                      edgecolor='black', linewidth=0.8,
                      alpha=1.0 - 0.15 * i)
        for bar, (d, a) in zip(bars, pairs):
            bar.set_hatch(assay_hatches[a])

    ax.set_xticks(x + width * (len(day_vals) - 1) / 2)
    ax.set_xticklabels([f'{d}\n{a}' for (d, a) in pairs], fontsize=9)
    ax.set_ylabel('IC50 (nM)')
    ax.set_yscale('log')
    ax.set_title(f'IC50 by Drug × Assay — {label}'
                 + (f'  (day groups: {", ".join(f"Day {d}" for d in day_vals)})'
                    if len(day_vals) > 1 else f'  (Day {day_vals[0]})'))

    handles = [Patch(facecolor=drug_colors[d], edgecolor='black', label=str(d))
               for d in sorted({d for (d, _) in pairs})]
    handles += [Patch(facecolor='white', edgecolor='black', hatch=assay_hatches[a],
                      label=str(a)) for a in sorted({a for (_, a) in pairs},
                                                    key=all_assays.index)]
    ax.legend(handles=handles, fontsize=9, ncol=2)
    plt.tight_layout()
    save_fig()

IC50 Bar Plots

Show code
# ===========================================================================
# 10. ExportDATA_SUBDIR
# ===========================================================================
print('\n=== 10. Export ===')
out_long = os.path.join(DATA_DIR, f'{EXPORT_STEM}_long.csv')
df.to_csv(out_long, index=False)
print(f'Saved long data: {out_long}')

for label, df_tbl in all_ic50_tables.items():
    out = os.path.join(DATA_DIR, f'{EXPORT_STEM}_IC50_{label}.csv')
    df_tbl.to_csv(out, index=False)
    print(f'Saved: {out}')

print(f'\nTotal plots saved: {plot_counter[0]}')
print('Done.')

=== 10. Export ===
Saved long data: /content/drive/MyDrive/20260827 Cytotox - Hu64_VPE6_GE6/Capan Combine Analysis-20260910/Data/Capan Combine Analysis-20260910_long.csv
Saved: /content/drive/MyDrive/20260827 Cytotox - Hu64_VPE6_GE6/Capan Combine Analysis-20260910/Data/Capan Combine Analysis-20260910_IC50_Confluence_interpolated.csv
Saved: /content/drive/MyDrive/20260827 Cytotox - Hu64_VPE6_GE6/Capan Combine Analysis-20260910/Data/Capan Combine Analysis-20260910_IC50_Confluence_direct.csv
Saved: /content/drive/MyDrive/20260827 Cytotox - Hu64_VPE6_GE6/Capan Combine Analysis-20260910/Data/Capan Combine Analysis-20260910_IC50_Confluence_pctViable.csv

Total plots saved: 11
Done.
Show code
# ===========================================================================
# 11. Therapeutic Windows — Relative IC50 (midpoint between plateaus)
#     Test vs isotype are compared WITHIN an assay; each assay gets its own
#     tables and bar charts so Castor and MTT windows are never mixed.
# ===========================================================================
print(f'\n=== 11. Therapeutic Windows ({TEST_AB} vs {ISO_AB}) ===')
print(f'    IC50 = relative IC50: conc at midpoint between untreated & max-dose plateaus')
print(f'    Assays analysed independently: {all_assays}')

def parse_drug(name, test_ab, iso_ab):
    """Return (antibody, linker-payload) or (None, None) if not matching."""
    for ab in [test_ab, iso_ab]:
        prefix = ab + '-'
        if name.startswith(prefix):
            return ab, name[len(prefix):]
    return None, None


def relative_ic50(concs, responses):
    """
    Relative IC50: concentration where response = midpoint between
    the untreated plateau (lowest conc) and the max-dose plateau (highest conc).
    Interpolates on log10(conc). Returns (ic50, target_pct, confident).
    confident=False if the curve never monotonically crosses the midpoint.
    """
    means = responses.groupby(concs).mean()
    c = np.array(means.index, dtype=float)
    v = np.array(means.values, dtype=float)
    order = np.argsort(c)
    c, v = c[order], v[order]

    untreated = v[0]
    max_dose  = v[-1]
    target    = (untreated + max_dose) / 2.0
    spread    = abs(untreated - max_dose)

    if spread < 5.0:
        return np.nan, target, False

    for i in range(len(v) - 1):
        if (v[i] >= target >= v[i+1]) or (v[i] <= target <= v[i+1]):
            log_c1, log_c2 = np.log10(max(c[i], 1e-6)), np.log10(max(c[i+1], 1e-6))
            frac = (target - v[i]) / (v[i+1] - v[i])
            ic50 = 10 ** (log_c1 + frac * (log_c2 - log_c1))
            confident = (spread >= 20.0) and (CONC_MIN <= ic50 <= CONC_MAX)
            return ic50, target, confident

    return np.nan, target, False


METRICS_FOR_TI = {
    'Confluence': 'PctViable_Confluence',
}

all_summary_rows = []

for metric_label, pv_col in METRICS_FOR_TI.items():
    if pv_col not in df.columns:
        print(f'\n  {metric_label}: column {pv_col} not found, skipping.')
        continue

    for assay in all_assays:
        assay_df = df[df.Assay == assay]
        drugs_in_assay = sorted(assay_df.Drug.dropna().unique())
        drug_ab = {d: parse_drug(d, TEST_AB, ISO_AB)[0] for d in drugs_in_assay}
        drug_lp = {d: parse_drug(d, TEST_AB, ISO_AB)[1] for d in drugs_in_assay}
        test_lps = {drug_lp[d] for d in drugs_in_assay if drug_ab[d] == TEST_AB}
        iso_lps  = {drug_lp[d] for d in drugs_in_assay if drug_ab[d] == ISO_AB}
        matched_lps = sorted(test_lps & iso_lps)

        if not matched_lps:
            print(f'\n  {metric_label} [{assay}]: no matched LPs between '
                  f'{TEST_AB} and {ISO_AB}.')
            continue

        days = sorted(assay_df[assay_df.Day.isin(fit_days)].Day.unique())

        print(f'\n{"="*60}')
        print(f'  {metric_label} (%Viable) — assay: {assay}')
        print(f'  Matched LPs: {matched_lps}')
        print(f'{"="*60}')

        for day in days:
            day_df = assay_df[assay_df.Day == day].copy()
            rows = []
            for lp in matched_lps:
                test_drug = f'{TEST_AB}-{lp}'
                iso_drug  = f'{ISO_AB}-{lp}'
                test_sub = day_df[(day_df.Drug == test_drug)].dropna(subset=['Conc', pv_col])
                iso_sub  = day_df[(day_df.Drug == iso_drug)].dropna(subset=['Conc', pv_col])
                test_sub = _analysis_rows(test_sub)
                iso_sub  = _analysis_rows(iso_sub)
                if test_sub.empty or iso_sub.empty:
                    continue

                t_ic50, t_target, t_conf = relative_ic50(test_sub.Conc, test_sub[pv_col])
                i_ic50, i_target, i_conf = relative_ic50(iso_sub.Conc, iso_sub[pv_col])

                t_untx = test_sub.groupby('Conc')[pv_col].mean().sort_index().iloc[0]
                t_maxd = test_sub.groupby('Conc')[pv_col].mean().sort_index().iloc[-1]
                i_untx = iso_sub.groupby('Conc')[pv_col].mean().sort_index().iloc[0]
                i_maxd = iso_sub.groupby('Conc')[pv_col].mean().sort_index().iloc[-1]

                ti = i_ic50 / t_ic50 if (not np.isnan(t_ic50) and t_ic50 > 0
                                          and not np.isnan(i_ic50)) else np.nan
                ti_conf = t_conf and i_conf and not np.isnan(ti)

                def fmt(val, conf):
                    if np.isnan(val):
                        return 'NA'
                    return f'{val:.3g}' if conf else f'{val:.3g}*'

                rows.append({
                    'Assay': assay,
                    'LP': lp,
                    f'{TEST_AB} untx→max': f'{t_untx:.0f}→{t_maxd:.0f}%',
                    f'{TEST_AB} IC50 (nM)': fmt(t_ic50, t_conf),
                    f'{ISO_AB} untx→max': f'{i_untx:.0f}→{i_maxd:.0f}%',
                    f'{ISO_AB} IC50 (nM)': fmt(i_ic50, i_conf),
                    f'TI ({ISO_AB}/{TEST_AB})': fmt(ti, ti_conf),
                })
                all_summary_rows.append({
                    'Method': metric_label, 'Assay': assay, 'Day': day, 'LP': lp,
                    f'{TEST_AB}_IC50_nM': t_ic50, f'{TEST_AB}_confident': t_conf,
                    f'{TEST_AB}_untreated_%': t_untx, f'{TEST_AB}_maxdose_%': t_maxd,
                    f'{ISO_AB}_IC50_nM': i_ic50, f'{ISO_AB}_confident': i_conf,
                    f'{ISO_AB}_untreated_%': i_untx, f'{ISO_AB}_maxdose_%': i_maxd,
                    'TI': ti, 'TI_confident': ti_conf,
                })

            if not rows:
                print(f'\n  [{assay}] Day {day}: no matched LP pairs with data.')
                continue

            tbl = pd.DataFrame(rows)
            print(f'\n--- {metric_label} [{assay}] — Day {day} ---')
            print(f'    IC50 = conc at midpoint of (untreated, max-dose)')
            print(f'    TI = IC50({ISO_AB}) / IC50({TEST_AB})  ;  * = low confidence')
            print(tbl.to_string(index=False))

            # Grouped bar chart: IC50 by LP
            plot_rows = [r for r in rows
                         if r[f'{TEST_AB} IC50 (nM)'] != 'NA' or r[f'{ISO_AB} IC50 (nM)'] != 'NA']
            if not plot_rows:
                continue
            lps_plot = [r['LP'] for r in plot_rows]

            fig, ax = plt.subplots(figsize=(max(6, len(lps_plot) * 2.5), 5))
            x = np.arange(len(lps_plot))
            width = 0.35
            colors = {TEST_AB: '#2196F3', ISO_AB: '#FF9800'}

            for j, ab in enumerate([TEST_AB, ISO_AB]):
                col_name = f'{ab} IC50 (nM)'
                vals, labels, hatches = [], [], []
                for r in plot_rows:
                    raw = r[col_name]
                    if raw == 'NA':
                        vals.append(0); labels.append('NA'); hatches.append('//')
                    else:
                        v = float(raw.rstrip('*'))
                        vals.append(v)
                        labels.append(raw)
                        hatches.append('//' if raw.endswith('*') else '')
                bars = ax.bar(x + (j - 0.5) * width, vals, width, label=ab,
                              color=colors[ab], alpha=0.85)
                for bar, lbl, hatch in zip(bars, labels, hatches):
                    bar.set_hatch(hatch)
                    if lbl != 'NA' and bar.get_height() > 0:
                        ax.text(bar.get_x() + bar.get_width() / 2, bar.get_height(),
                                lbl, ha='center', va='bottom', fontsize=9)

            ax.set_xticks(x)
            ax.set_xticklabels(lps_plot, fontsize=11)
            ax.set_ylabel('Relative IC50 (nM)')
            ax.set_title(f'{metric_label} [{assay}] — Day {day}\n'
                         f'Relative IC50 by Linker-Payload  (* = low confidence)',
                         fontsize=13)
            ax.legend()
            ax.set_yscale('log')
            plt.tight_layout()
            save_fig()

            # TI bar chart
            ti_rows = [r for r in plot_rows if r[f'TI ({ISO_AB}/{TEST_AB})'] != 'NA']
            if ti_rows:
                ti_vals = [float(r[f'TI ({ISO_AB}/{TEST_AB})'].rstrip('*')) for r in ti_rows]
                ti_lps  = [r['LP'] for r in ti_rows]
                ti_conf = [not r[f'TI ({ISO_AB}/{TEST_AB})'].endswith('*') for r in ti_rows]

                fig, ax = plt.subplots(figsize=(max(5, len(ti_vals) * 2), 4.5))
                bar_colors = ['#4CAF50' if c else '#FFC107' for c in ti_conf]
                bar_hatches = ['' if c else '//' for c in ti_conf]
                bars = ax.bar(range(len(ti_vals)), ti_vals, color=bar_colors, alpha=0.85)
                for bar, hatch, v, c in zip(bars, bar_hatches, ti_vals, ti_conf):
                    bar.set_hatch(hatch)
                    lbl = f'{v:.1f}' + ('' if c else '*')
                    ax.text(bar.get_x() + bar.get_width() / 2, bar.get_height(),
                            lbl, ha='center', va='bottom', fontsize=10, fontweight='bold')
                ax.axhline(1.0, color='red', linestyle='--', linewidth=1, label='TI = 1')
                ax.set_xticks(range(len(ti_vals)))
                ax.set_xticklabels(ti_lps, fontsize=11)
                ax.set_ylabel(f'Therapeutic Index\nIC50({ISO_AB}) / IC50({TEST_AB})')
                ax.set_title(f'{metric_label} [{assay}] — Day {day}\n'
                             f'Therapeutic Index  (* = low confidence)',
                             fontsize=13)
                ax.legend()
                plt.tight_layout()
                save_fig()

# Export combined summary
if all_summary_rows:
    df_ti_summary = pd.DataFrame(all_summary_rows)
    out = os.path.join(DATA_DIR, f'{EXPORT_STEM}_therapeutic_index_{TEST_AB}_vs_{ISO_AB}.csv')
    df_ti_summary.to_csv(out, index=False)
    print(f'\nSaved combined summary: {out}')
    print('\n' + '='*60)
    print('  Combined Therapeutic Index Summary (all assays)')
    print('='*60)
    print(df_ti_summary.to_string(index=False, float_format=lambda x: f'{x:.3g}'))

=== 11. Therapeutic Windows (Hu63 vs Iso) ===
    IC50 = relative IC50: conc at midpoint between untreated & max-dose plateaus
    Assays analysed independently: ['Castor', 'MTT']

  Confluence [Castor]: no matched LPs between Hu63 and Iso.

  Confluence [MTT]: no matched LPs between Hu63 and Iso.