Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261015T141849Z-rare-cell-7k-8k-panels/diffusion/dataset/server_1/model8/plans/rare_gene_context_expansion_20261015/coordination/ilc_training_support/audit_training_ilc.py

@@ -0,0 +1,349 @@
+#!/usr/bin/env python3
+"""Bounded, CPU-only source-labelled TRAIN ILC marker support audit.
+
+Reuses the chunk/row algorithm in corpus_macrophage_markers_20261015.
+Selection is fixed by source L1 annotation and registered split, never expression.
+"""
+from __future__ import annotations
+
+import argparse
+from collections import defaultdict
+import hashlib
+import json
+from pathlib import Path
+import resource
+import time
+
+import h5py
+import numpy as np
+import pandas as pd
+import pyarrow.parquet as pq
+import zarr
+
+ROOT = Path(__file__).resolve().parent
+PLAN = ROOT.parents[1]
+M = PLAN.parents[1]
+M7 = M.parent / 'model_7_no_annotation_sh_hwang/data'
+E2 = M7 / 'e2_full_source'
+BLOCK = Path('/home/sj_server_1/PBISC/PBISC-data/single-cell-data-block-blood')
+KEY = ['dataset_id', 'donor_id', 'sample_id']
+SOURCE_CT = Path('/home/sj_server_1/PBISC/paper-release-20260928/manuscript/source_data/corpus/corpus_celltypes.csv')
+zarr.config.set({'async.concurrency': 2, 'threading.max_workers': 2})
+
+
+def sha(path: Path) -> str:
+    h = hashlib.sha256()
+    with Path(path).open('rb') as f:
+        for b in iter(lambda: f.read(1 << 20), b''): h.update(b)
+    return h.hexdigest()
+
+
+def receipt(path: Path) -> dict:
+    p = Path(path)
+    return {'path': str(p), 'bytes': p.stat().st_size, 'sha256': sha(p)}
+
+
+def dump(name: str, value: dict | list) -> None:
+    (ROOT / name).write_text(json.dumps(value, indent=2) + '\n')
+
+
+def targets() -> pd.DataFrame:
+    p = ROOT / 'locked_target_genes.csv'
+    t = pd.read_csv(p)
+    assert len(t) == 48 and not t.gene_id.duplicated().any()
+    return t
+
+
+def prepare() -> None:
+    start = time.time()
+    target_source = PLAN / 'ilc_targeted_audit_genes.csv'
+    t = pd.read_csv(target_source)
+    t.to_csv(ROOT / 'locked_target_genes.csv', index=False)
+    t = targets()
+    cols = t.raw_panel_position.to_numpy(dtype=int)
+    var = pd.read_parquet(E2 / 'metadata/var.parquet')
+    assert len(var) == 20097
+    assert var.iloc[cols].gene_id.tolist() == t.gene_id.tolist()
+    assert var.iloc[cols].gene_symbol.tolist() == t.gene_symbol.tolist()
+    assert var.iloc[cols].panel_index.tolist() == cols.tolist()
+    cfgp = M / 'runs/f01_6k_1x_seed20260627/effective_config.json'
+    cfg = json.loads(cfgp.read_text())['data']
+    assert cfg['n_genes'] == 6000
+    manifest_path = Path(cfg['gene_selection_manifest'])
+    position_path = Path(cfg['gene_positions_path'])
+    assert sha(manifest_path) == cfg['gene_selection_manifest_sha256']
+    assert sha(position_path) == cfg['gene_positions_sha256']
+    manifest = json.loads(manifest_path.read_text())
+    selected_path = Path(manifest['selected_genes_path'])
+    assert sha(selected_path) == manifest['selected_genes_sha256']
+    selected = pd.read_csv(selected_path)
+    assert len(selected) == 6000 and not selected.gene_id.duplicated().any()
+    assert np.array_equal(np.load(position_path), selected.panel_position)
+    assert var.iloc[selected.panel_position].gene_id.tolist() == selected.gene_id.tolist()
+    assert np.array_equal(t.in_final6k, t.gene_id.isin(selected.gene_id))
+    records = [receipt(p) for p in [target_source, ROOT / 'locked_target_genes.csv',
+               E2 / 'metadata/var.parquet', SOURCE_CT, cfgp, manifest_path, position_path, selected_path]]
+    si_path = M7 / 'e2_full_index/metadata/sample_index.parquet'
+    si = pd.read_parquet(si_path)
+    for k in KEY: si[k] = si[k].astype(str)
+    assert not si.duplicated(KEY).any()
+    records.append(receipt(si_path))
+    ct = pd.read_csv(SOURCE_CT)
+    expected = ct[ct.celltype_l1 == 'ILC'].groupby('dataset_prefix').n_cells.sum().to_dict()
+    assert sum(expected.values()) == 19329
+    files = [p for p in sorted((E2 / 'metadata/obs_sc.parquet').glob('*.parquet'))
+             if any(p.stem.startswith(k + '_') for k in expected)
+             and not p.name.endswith('__strict_overlay.parquet')]
+    selected_rows, checks, all_ilc = [], [], defaultdict(int)
+    for fi, ep in enumerate(files):
+        bp = BLOCK / 'metadata/obs_sc.parquet' / ep.name
+        eq, bq = pq.ParquetFile(ep), pq.ParquetFile(bp)
+        assert eq.metadata.num_rows == bq.metadata.num_rows
+        bcols = ['source_row_index', 'row_index', 'celltype_l1']
+        extra = [x for x in ['source_celltype_l1', 'label_original', 'source_file'] if x in bq.schema.names]
+        bi = iter(bq.iter_batches(batch_size=32768, columns=bcols + extra, use_threads=False))
+        offset = n_ilc = n_train = 0
+        for eb in eq.iter_batches(batch_size=32768, columns=KEY + ['source_row_index', 'row_index'], use_threads=False):
+            e, b = eb.to_pandas(), next(bi).to_pandas()
+            assert len(e) == len(b) and np.array_equal(e.source_row_index, b.source_row_index)
+            assert np.array_equal(e.row_index, np.arange(offset, offset + len(e)))
+            keep = b.celltype_l1.astype(str).eq('ILC').to_numpy()
+            if keep.any():
+                v = e.loc[keep].copy()
+                for k in KEY: v[k] = v[k].astype(str)
+                v['shard_row'] = v.row_index
+                v['block_pre_qc_row_index'] = b.loc[keep, 'row_index'].to_numpy()
+                v['source_label'] = b.loc[keep, 'celltype_l1'].astype(str).to_numpy()
+                for c in extra: v[c] = b.loc[keep, c].astype(str).to_numpy()
+                v = v.merge(si[KEY + ['split', 'sample_key']], on=KEY, how='left', validate='many_to_one')
+                assert v.split.notna().all()
+                prefix = next(k for k in expected if ep.stem.startswith(k + '_'))
+                all_ilc[prefix] += len(v)
+                n_ilc += len(v)
+                v = v[v.split == 'train'].copy()
+                v['shard_id'] = ep.stem
+                n_train += len(v)
+                if len(v): selected_rows.append(v)
+            offset += len(e)
+        assert next(bi, None) is None and offset == eq.metadata.num_rows
+        checks.append({'shard_id': ep.stem, 'metadata_rows': offset, 'source_ilc': n_ilc,
+                       'train_ilc': n_train, 'source_identity_and_physical_order_match': True})
+        records.extend([receipt(ep), receipt(bp)])
+        if fi % 20 == 0 or fi + 1 == len(files): print(f'metadata {fi+1}/{len(files)}', flush=True)
+    assert dict(all_ilc) == expected, (dict(all_ilc), expected)
+    v = pd.concat(selected_rows, ignore_index=True)
+    assert len(v) == 16439 and not v.duplicated(['shard_id', 'shard_row']).any()
+    v.to_parquet(ROOT / 'selected_training_ilc.parquet', index=False)
+    rows_by_sid = {x['shard_id']: x['metadata_rows'] for x in checks}
+    plans = []
+    for sid, d in v.groupby('shard_id', sort=True):
+        zp = BLOCK / 'counts/sc_counts' / (sid + '.zarr')
+        g = zarr.open_group(str(zp), mode='r')
+        for p in sorted(zp.glob('**/zarr.json')): records.append(receipt(p))
+        assert g.attrs['shape'] == [rows_by_sid[sid], 20097] and g.attrs['shard_id'] == sid
+        mask = np.asarray(g['measurement_mask'][:])[cols].astype(bool)
+        ptr = np.asarray(g['indptr'][:], dtype=np.int64)
+        for a in ['indptr', 'measurement_mask']:
+            for p in sorted((zp / a / 'c').glob('*')):
+                if p.is_file(): records.append(receipt(p))
+        r = d.shard_row.to_numpy()
+        size = g['indices'].chunks[0]
+        assert size == g['data'].chunks[0] == 1000000
+        chunks = sorted(set(c for rr in r for c in range(int(ptr[rr]) // size, (int(ptr[rr+1]) - 1) // size + 1)))
+        paths = [zp / a / 'c' / str(c) for c in chunks for a in ['indices', 'data']]
+        plans.append({'shard_id': sid, 'cells': len(d), 'chunks': chunks,
+                      'compressed_bytes': sum(p.stat().st_size for p in paths if p.exists()),
+                      'decoded_bytes_upper_bound': len(chunks) * size * 12,
+                      'selected_nnz': int((ptr[r+1]-ptr[r]).sum()), 'measured': mask.tolist(),
+                      'source_count_matrix': g.attrs['source_count_matrix'], 'source_path': g.attrs['source_path']})
+    plan = {'status': 'prepared', 'source_ilc': sum(all_ilc.values()), 'train_ilc': len(v),
+            'source_ilc_by_dataset_prefix': all_ilc, 'metadata_shards_scanned': len(files),
+            'metadata_rows_scanned': sum(x['metadata_rows'] for x in checks),
+            'selected_shards': len(plans), 'compressed_bytes': sum(x['compressed_bytes'] for x in plans),
+            'decoded_bytes_upper_bound': sum(x['decoded_bytes_upper_bound'] for x in plans),
+            'plans': plans, 'metadata_checks': checks, 'input_sources': records,
+            'elapsed_seconds': time.time()-start}
+    dump('extraction_plan.json', plan)
+    print(json.dumps({k: plan[k] for k in ['source_ilc', 'train_ilc', 'selected_shards', 'compressed_bytes', 'decoded_bytes_upper_bound', 'elapsed_seconds']}), flush=True)
+
+
+def extract() -> None:
+    start = time.time()
+    plan = json.loads((ROOT / 'extraction_plan.json').read_text())
+    # Coordinator reviewed the 12.40 GiB cost gate and authorized this exact local subset.
+    assert plan['compressed_bytes'] <= 13 * 2**30, 'Cost exceeds coordinator-reviewed 12.40 GiB subset'
+    t = targets(); cols = t.raw_panel_position.to_numpy(dtype=int)
+    v = pd.read_parquet(ROOT / 'selected_training_ilc.parquet')
+    out = np.zeros((len(v), len(cols)), dtype=np.int64)
+    measured = np.zeros_like(out, dtype=bool)
+    totals = np.zeros(len(v), dtype=np.int64)
+    chunks_read, shard_checks = [], []
+    dataset_timings = []; ds_started = time.time(); ds_bytes = 0; current_ds = None
+    for pi, pp in enumerate(plan['plans']):
+        sid = pp['shard_id']; d = v[v.shard_id == sid].sort_values('shard_row')
+        ds = str(d.dataset_id.iloc[0])
+        if current_ds is None: current_ds = ds
+        elif ds != current_ds:
+            timing = {'dataset_id': current_ds, 'seconds': time.time()-ds_started, 'compressed_bytes': ds_bytes,
+                      'process_peak_rss_kib': resource.getrusage(resource.RUSAGE_SELF).ru_maxrss}
+            dataset_timings.append(timing); print(json.dumps({'dataset_done': timing}), flush=True)
+            current_ds = ds; ds_started = time.time(); ds_bytes = 0
+        zp = BLOCK / 'counts/sc_counts' / (sid + '.zarr')
+        g = zarr.open_group(str(zp), mode='r')
+        mask = np.asarray(g['measurement_mask'][:])[cols].astype(bool)
+        assert mask.tolist() == pp['measured']
+        measured[d.index] = mask
+        ptr = np.asarray(g['indptr'][:], dtype=np.int64)
+        segs = defaultdict(list); size = 1000000
+        for oi, r in zip(d.index, d.shard_row):
+            a, b = int(ptr[r]), int(ptr[r+1])
+            while a < b:
+                c = a // size; end = min(b, (c+1)*size)
+                segs[c].append((int(oi), a-c*size, end-c*size)); a = end
+        assert sorted(segs) == pp['chunks']
+        nnz = 0
+        for c, pieces in sorted(segs.items()):
+            a, b = c*size, min((c+1)*size, g['indices'].shape[0])
+            ix = np.asarray(g['indices'][a:b], dtype=np.int32)
+            val = np.asarray(g['data'][a:b])
+            assert (val >= 0).all() and np.array_equal(val, val.astype(np.int64))
+            val = val.astype(np.int64)
+            for oi, lo, hi in pieces:
+                ids, values = ix[lo:hi], val[lo:hi]
+                totals[oi] += values.sum(); nnz += hi-lo
+                for j, col in enumerate(cols): out[oi, j] += values[ids == col].sum()
+            for arr in ['indices', 'data']:
+                p = zp / arr / 'c' / str(c)
+                if p.exists(): chunks_read.append(receipt(p))
+        assert nnz == pp['selected_nnz']
+        shard_checks.append({'shard_id': sid, 'cells': len(d), 'selected_nnz': nnz, 'chunks': len(segs)})
+        ds_bytes += pp['compressed_bytes']
+        print(f'expression {pi+1}/{len(plan["plans"])} {sid[:6]} cells={len(d)}', flush=True)
+    assert (totals > 0).all() and (out[~measured] == 0).all()
+    for j, gid in enumerate(t.gene_id):
+        v[gid] = out[:, j]
+        v[gid + '__measured'] = measured[:, j]
+    v['total_counts_20097'] = totals
+    v.to_parquet(ROOT / 'training_ilc_marker_counts.parquet', index=False)
+    dump('expression_chunk_receipts.json', chunks_read)
+    dataset_timings.append({'dataset_id': current_ds, 'seconds': time.time()-ds_started, 'compressed_bytes': ds_bytes,
+                            'process_peak_rss_kib': resource.getrusage(resource.RUSAGE_SELF).ru_maxrss})
+    dump('extraction_result.json', {'status': 'extracted', 'cells': len(v), 'genes': len(t),
+         'elapsed_seconds': time.time()-start, 'shard_checks': shard_checks,
+         'process_peak_rss_kib': resource.getrusage(resource.RUSAGE_SELF).ru_maxrss, 'dataset_timings': dataset_timings,
+         'chunk_receipts': receipt(ROOT / 'expression_chunk_receipts.json')})
+    summarize(v, t)
+
+
+def summarize(v: pd.DataFrame, t: pd.DataFrame) -> None:
+    tables = {'overall': [], 'by_dataset': [], 'by_donor': []}
+    grouped = [('overall', {}, v)]
+    grouped += [('by_dataset', {'dataset_id': k}, d) for k, d in v.groupby('dataset_id', sort=True)]
+    grouped += [('by_donor', dict(zip(KEY[:2], k)), d) for k, d in v.groupby(KEY[:2], sort=True)]
+    for level, labels, d in grouped:
+        for r in t.itertuples(index=False):
+            eligible = d[r.gene_id + '__measured']
+            det = eligible & d[r.gene_id].gt(0)
+            e, p = d[eligible], d[det]
+            n = len(e)
+            tables[level].append({**labels, 'gene_id': r.gene_id, 'gene_symbol': r.gene_symbol,
+                'raw_panel_position': int(r.raw_panel_position), 'source_labelled_ilc_cells': len(d),
+                'measured_eligible_cells': n, 'unmeasured_cells': len(d)-n, 'detected_cells': len(p),
+                'detection_fraction_among_measured': len(p)/n if n else None,
+                'total_raw_counts': int(e[r.gene_id].sum()),
+                'mean_raw_counts_among_measured': float(e[r.gene_id].mean()) if n else None,
+                'measured_dataset_scoped_donors': len(e[KEY[:2]].drop_duplicates()),
+                'detected_dataset_scoped_donors': len(p[KEY[:2]].drop_duplicates()),
+                'measured_samples': e.sample_key.nunique(), 'detected_samples': p.sample_key.nunique(),
+                'in_final6k': bool(r.in_final6k), 'in_7k': bool(r.in_7k), 'in_8k': bool(r.in_8k),
+                'sentinel': bool(r.sentinel)})
+    for name, rows in tables.items(): pd.DataFrame(rows).to_csv(ROOT / f'marker_support_{name}.csv', index=False)
+    depths = []
+    for name, d in [('ALL', v)] + list(v.groupby('dataset_id', sort=True)):
+        q = d.total_counts_20097.quantile([0, .1, .25, .5, .75, .9, 1]).to_dict()
+        depths.append({'dataset_id': name, 'cells': len(d), 'source_donor_ids': d.donor_id.nunique(),
+                       'samples': d.sample_key.nunique(), **{f'canonical_count_q{k}': x for k, x in q.items()}})
+    pd.DataFrame(depths).to_csv(ROOT / 'native_depth_support.csv', index=False)
+    v.groupby(['dataset_id', 'source_label', 'label_original'], dropna=False).size().rename('cells').reset_index().to_csv(ROOT / 'source_label_support.csv', index=False)
+
+
+def verify() -> None:
+    t = targets(); cols = t.raw_panel_position.to_numpy(dtype=int); ids = t.gene_id.tolist()
+    var = pd.read_parquet(E2 / 'metadata/var.parquet')
+    v = pd.read_parquet(ROOT / 'training_ilc_marker_counts.parquet')
+    physical, direct, original, original_gaps = [], [], [], []
+    first_ds, boundary_ds = set(), set()
+    for sid, d in v.groupby('shard_id', sort=True):
+        zp = BLOCK / 'counts/sc_counts' / (sid + '.zarr')
+        g = zarr.open_group(str(zp), mode='r')
+        ep = E2 / 'metadata/obs_sc.parquet' / (sid + '.parquet')
+        bp = BLOCK / 'metadata/obs_sc.parquet' / (sid + '.parquet')
+        e = pq.read_table(ep, columns=['row_index', 'source_row_index']).to_pandas()
+        b = pq.read_table(bp, columns=['row_index', 'source_row_index', 'celltype_l1']).to_pandas()
+        rows = d.shard_row.to_numpy()
+        assert np.array_equal(e.iloc[rows].row_index, rows)
+        assert np.array_equal(e.iloc[rows].source_row_index, d.source_row_index)
+        assert np.array_equal(b.iloc[rows].source_row_index, d.source_row_index)
+        assert b.iloc[rows].celltype_l1.astype(str).eq('ILC').all()
+        assert g.attrs['shape'] == [len(e), len(var)]
+        physical.append({'shard_id': sid, 'cells': len(d), 'physical_identity_pass': True,
+                         'block_pre_qc_row_different': int((b.iloc[rows].row_index.to_numpy() != rows).sum())})
+        ds = str(d.dataset_id.iloc[0]); chosen = set()
+        if ds not in first_ds: chosen.update([d.index[0], d.index[-1]])
+        ptr = np.asarray(g['indptr'][:])
+        boundary = ptr[rows] // 1000000 != (ptr[rows+1]-1) // 1000000
+        if ds not in boundary_ds and boundary.any():
+            chosen.add(d.index[np.flatnonzero(boundary)[0]]); boundary_ds.add(ds)
+        for ix in sorted(chosen):
+            r = v.loc[ix]; a, stop = map(int, ptr[int(r.shard_row):int(r.shard_row)+2])
+            ri, rv = np.asarray(g['indices'][a:stop]), np.asarray(g['data'][a:stop])
+            observed = np.array([rv[ri == c].sum() for c in cols])
+            assert np.array_equal(observed, r[ids].to_numpy(dtype=np.int64))
+            assert int(rv.sum()) == int(r.total_counts_20097)
+            rec = {'dataset_id': ds, 'shard_id': sid, 'shard_row': int(r.shard_row),
+                   'counts': observed.tolist(), 'row_sum': int(rv.sum()),
+                   'crosses_chunk_boundary': a // 1000000 != (stop-1) // 1000000}
+            direct.append(rec)
+            if ds not in first_ds:
+                hp = Path('/home/sj_server_1/gblee') / g.attrs['source_path']
+                if not hp.is_file():
+                    original_gaps.append({'dataset_id': ds, 'recorded_path': str(hp),
+                        'reason': 'Historical original H5AD path is absent locally; canonical CSR replay remains available and verified.'})
+                    first_ds.add(ds)
+                    continue
+                with h5py.File(hp, 'r') as h:
+                    key = g.attrs['source_count_matrix']; vg = h['raw/var' if key.startswith('raw/') else 'var']
+                    rawids = [x.decode() if isinstance(x, bytes) else str(x) for x in vg[vg.attrs['_index']][:]]
+                    assert len(rawids) == len(set(rawids))
+                    rawpos = {gid: j for j, gid in enumerate(rawids)}; mx = h[key]
+                    assert mx.attrs['encoding-type'] == 'csr_matrix'
+                    aa, zz = map(int, mx['indptr'][int(r.source_row_index):int(r.source_row_index)+2])
+                    oi, ov = mx['indices'][aa:zz], mx['data'][aa:zz]
+                    raw_counts = np.array([ov[oi == rawpos[gid]].sum() if gid in rawpos else 0 for gid in ids])
+                    raw_mask = np.array([gid in rawpos for gid in ids])
+                    assert np.array_equal(raw_counts, observed)
+                    assert np.array_equal(raw_mask, r[[gid + '__measured' for gid in ids]].to_numpy(dtype=bool))
+                    canonical = np.array([rawpos[gid] for gid in var.gene_id if gid in rawpos])
+                    assert ov[np.isin(oi, canonical)].sum() == r.total_counts_20097
+                    selected_bytes_hash = hashlib.sha256(oi.tobytes() + ov.tobytes()).hexdigest()
+                original.append({'dataset_id': ds, 'path': str(hp), 'bytes': hp.stat().st_size,
+                    'mtime_ns': hp.stat().st_mtime_ns, 'matrix': key, 'source_row_index': int(r.source_row_index),
+                    'shard_id': sid, 'shard_row': int(r.shard_row), '48_counts_match': True,
+                    '48_measurement_mask_matches_original_gene_axis': True, 'canonical_count_sum_matches': True,
+                    'original_selected_row_indices_values_sha256': selected_bytes_hash,
+                    'original_gene_axis_sha256': hashlib.sha256('\n'.join(rawids).encode()).hexdigest()})
+                first_ds.add(ds)
+    assert len(original) + len(original_gaps) == v.dataset_id.nunique() == 12
+    assert len(original) >= 4
+    assert sum(x['cells'] for x in physical) == 16439
+    dump('verification.json', {'status': 'PASS', 'physical_rows': physical, 'direct_csr_rows': direct,
+         'original_h5ad_rows': original, 'raw_positions': cols.tolist(), 'gene_ids': ids,
+         'original_source_gaps': original_gaps,
+         'source_label_is_not_independently_verified_biological_identity': True})
+    print(json.dumps({'status': 'PASS', 'cells': len(v), 'shards': len(physical),
+         'direct_csr_rows': len(direct), 'original_h5ad_rows': len(original)}), flush=True)
+
+
+if __name__ == '__main__':
+    ap = argparse.ArgumentParser(); ap.add_argument('stage', choices=['prepare', 'extract', 'verify'])
+    globals()[ap.parse_args().stage]()

paper-releases/pbisc-diffusion/updates/20261015T141849Z-rare-cell-7k-8k-panels/diffusion/dataset/server_1/model8/plans/rare_gene_context_expansion_20261015/coordination/ilc_training_support/cgroup_resource_observation.txt

@@ -0,0 +1,10 @@
+OOMPolicy=kill
+MemoryCurrent=2147188736
+MemoryPeak=2149576704
+CPUUsageNSec=79003112000
+CPUQuotaPerSecUSec=2s
+MemoryHigh=2147483648
+MemoryMax=3221225472
+MemorySwapMax=0
+OOMScoreAdjust=500
+ActiveState=active

paper-releases/pbisc-diffusion/updates/20261015T141849Z-rare-cell-7k-8k-panels/diffusion/dataset/server_1/model8/plans/rare_gene_context_expansion_20261015/coordination/ilc_training_support/codetection_diagnostics.csv

@@ -0,0 +1,92 @@
+dataset_id,pattern,eligible_cells,positive_cells,fraction,detection_rule,interpretation
+ALL,IL7R_and_KLRB1,16439,6936,0.421923474663909,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+ALL,CD3D_or_CD3E,16439,2188,0.13309812032362064,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+ALL,NKG7_and_GNLY,16439,1401,0.0852241620536529,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+ALL,IL7R_and_KLRB1_with_CD3D_or_CD3E,16439,1381,0.08400754303789769,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+ALL,IL7R_and_KLRB1_with_NKG7_and_GNLY,16439,1106,0.06727903157126346,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+ALL,CD3D_or_CD3E_with_NKG7_and_GNLY,16439,515,0.031327939655696815,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+ALL,all_three_patterns,16439,402,0.024454042216679846,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+005_A_collection_dde06e0f-ab3b-46be-96a2-a8082383c4_OneK1K_Yazar_et_al._single-cell_eQTL_mapping,IL7R_and_KLRB1,350,202,0.5771428571428572,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+005_A_collection_dde06e0f-ab3b-46be-96a2-a8082383c4_OneK1K_Yazar_et_al._single-cell_eQTL_mapping,CD3D_or_CD3E,350,92,0.26285714285714284,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+005_A_collection_dde06e0f-ab3b-46be-96a2-a8082383c4_OneK1K_Yazar_et_al._single-cell_eQTL_mapping,NKG7_and_GNLY,350,109,0.31142857142857144,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+005_A_collection_dde06e0f-ab3b-46be-96a2-a8082383c4_OneK1K_Yazar_et_al._single-cell_eQTL_mapping,IL7R_and_KLRB1_with_CD3D_or_CD3E,350,57,0.16285714285714287,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+005_A_collection_dde06e0f-ab3b-46be-96a2-a8082383c4_OneK1K_Yazar_et_al._single-cell_eQTL_mapping,IL7R_and_KLRB1_with_NKG7_and_GNLY,350,65,0.18571428571428572,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+005_A_collection_dde06e0f-ab3b-46be-96a2-a8082383c4_OneK1K_Yazar_et_al._single-cell_eQTL_mapping,CD3D_or_CD3E_with_NKG7_and_GNLY,350,47,0.13428571428571429,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+005_A_collection_dde06e0f-ab3b-46be-96a2-a8082383c4_OneK1K_Yazar_et_al._single-cell_eQTL_mapping,all_three_patterns,350,32,0.09142857142857143,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+006_A_collection_ced320a1-29f3-47c1-a735-513c7084d5_AIDA_Asian_Immune_Diversity_Atlas,IL7R_and_KLRB1,556,453,0.814748201438849,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+006_A_collection_ced320a1-29f3-47c1-a735-513c7084d5_AIDA_Asian_Immune_Diversity_Atlas,CD3D_or_CD3E,556,141,0.25359712230215825,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+006_A_collection_ced320a1-29f3-47c1-a735-513c7084d5_AIDA_Asian_Immune_Diversity_Atlas,NKG7_and_GNLY,556,58,0.10431654676258993,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+006_A_collection_ced320a1-29f3-47c1-a735-513c7084d5_AIDA_Asian_Immune_Diversity_Atlas,IL7R_and_KLRB1_with_CD3D_or_CD3E,556,103,0.18525179856115107,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+006_A_collection_ced320a1-29f3-47c1-a735-513c7084d5_AIDA_Asian_Immune_Diversity_Atlas,IL7R_and_KLRB1_with_NKG7_and_GNLY,556,47,0.08453237410071943,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+006_A_collection_ced320a1-29f3-47c1-a735-513c7084d5_AIDA_Asian_Immune_Diversity_Atlas,CD3D_or_CD3E_with_NKG7_and_GNLY,556,29,0.052158273381294966,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+006_A_collection_ced320a1-29f3-47c1-a735-513c7084d5_AIDA_Asian_Immune_Diversity_Atlas,all_three_patterns,556,21,0.03776978417266187,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+009_A_Human_Immune_Health_Atlas_Human_Immune_Health_Atlas_AIFI_immune_dynamics_across_healthy_age,IL7R_and_KLRB1,661,550,0.8320726172465961,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+009_A_Human_Immune_Health_Atlas_Human_Immune_Health_Atlas_AIFI_immune_dynamics_across_healthy_age,CD3D_or_CD3E,661,138,0.2087745839636914,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+009_A_Human_Immune_Health_Atlas_Human_Immune_Health_Atlas_AIFI_immune_dynamics_across_healthy_age,NKG7_and_GNLY,661,116,0.17549167927382753,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+009_A_Human_Immune_Health_Atlas_Human_Immune_Health_Atlas_AIFI_immune_dynamics_across_healthy_age,IL7R_and_KLRB1_with_CD3D_or_CD3E,661,123,0.18608169440242056,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+009_A_Human_Immune_Health_Atlas_Human_Immune_Health_Atlas_AIFI_immune_dynamics_across_healthy_age,IL7R_and_KLRB1_with_NKG7_and_GNLY,661,99,0.14977307110438728,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+009_A_Human_Immune_Health_Atlas_Human_Immune_Health_Atlas_AIFI_immune_dynamics_across_healthy_age,CD3D_or_CD3E_with_NKG7_and_GNLY,661,37,0.05597579425113464,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+009_A_Human_Immune_Health_Atlas_Human_Immune_Health_Atlas_AIFI_immune_dynamics_across_healthy_age,all_three_patterns,661,33,0.049924357034795766,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+010_A_collection_4a9fd4d7-d870-4265-89a5-ad51ab811d_ScaleBio_single-cell_RNA_sequencing_of_human_PBMCs,IL7R_and_KLRB1,631,180,0.28526148969889065,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+010_A_collection_4a9fd4d7-d870-4265-89a5-ad51ab811d_ScaleBio_single-cell_RNA_sequencing_of_human_PBMCs,CD3D_or_CD3E,631,137,0.21711568938193343,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+010_A_collection_4a9fd4d7-d870-4265-89a5-ad51ab811d_ScaleBio_single-cell_RNA_sequencing_of_human_PBMCs,NKG7_and_GNLY,631,33,0.05229793977812995,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+010_A_collection_4a9fd4d7-d870-4265-89a5-ad51ab811d_ScaleBio_single-cell_RNA_sequencing_of_human_PBMCs,IL7R_and_KLRB1_with_CD3D_or_CD3E,631,31,0.0491283676703645,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+010_A_collection_4a9fd4d7-d870-4265-89a5-ad51ab811d_ScaleBio_single-cell_RNA_sequencing_of_human_PBMCs,IL7R_and_KLRB1_with_NKG7_and_GNLY,631,9,0.014263074484944533,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+010_A_collection_4a9fd4d7-d870-4265-89a5-ad51ab811d_ScaleBio_single-cell_RNA_sequencing_of_human_PBMCs,CD3D_or_CD3E_with_NKG7_and_GNLY,631,11,0.017432646592709985,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+010_A_collection_4a9fd4d7-d870-4265-89a5-ad51ab811d_ScaleBio_single-cell_RNA_sequencing_of_human_PBMCs,all_three_patterns,631,1,0.001584786053882726,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+053_B_collection_b0cf0afa-ec40-4d65-b570-ed4ceacc68_Integrated_analysis_of_multimodal_single-cell_data_CITE-seq_PBMC,IL7R_and_KLRB1,85,77,0.9058823529411765,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+053_B_collection_b0cf0afa-ec40-4d65-b570-ed4ceacc68_Integrated_analysis_of_multimodal_single-cell_data_CITE-seq_PBMC,CD3D_or_CD3E,85,27,0.3176470588235294,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+053_B_collection_b0cf0afa-ec40-4d65-b570-ed4ceacc68_Integrated_analysis_of_multimodal_single-cell_data_CITE-seq_PBMC,NKG7_and_GNLY,85,9,0.10588235294117647,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+053_B_collection_b0cf0afa-ec40-4d65-b570-ed4ceacc68_Integrated_analysis_of_multimodal_single-cell_data_CITE-seq_PBMC,IL7R_and_KLRB1_with_CD3D_or_CD3E,85,25,0.29411764705882354,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+053_B_collection_b0cf0afa-ec40-4d65-b570-ed4ceacc68_Integrated_analysis_of_multimodal_single-cell_data_CITE-seq_PBMC,IL7R_and_KLRB1_with_NKG7_and_GNLY,85,8,0.09411764705882353,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+053_B_collection_b0cf0afa-ec40-4d65-b570-ed4ceacc68_Integrated_analysis_of_multimodal_single-cell_data_CITE-seq_PBMC,CD3D_or_CD3E_with_NKG7_and_GNLY,85,4,0.047058823529411764,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+053_B_collection_b0cf0afa-ec40-4d65-b570-ed4ceacc68_Integrated_analysis_of_multimodal_single-cell_data_CITE-seq_PBMC,all_three_patterns,85,3,0.03529411764705882,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+065_A_GSE271896_GSE271896_Sound_Life_deep_longitudinal_immune_profiling,IL7R_and_KLRB1,4616,3833,0.8303726169844021,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+065_A_GSE271896_GSE271896_Sound_Life_deep_longitudinal_immune_profiling,CD3D_or_CD3E,4616,791,0.17136048526863085,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+065_A_GSE271896_GSE271896_Sound_Life_deep_longitudinal_immune_profiling,NKG7_and_GNLY,4616,688,0.14904679376083188,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+065_A_GSE271896_GSE271896_Sound_Life_deep_longitudinal_immune_profiling,IL7R_and_KLRB1_with_CD3D_or_CD3E,4616,687,0.14883015597920277,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+065_A_GSE271896_GSE271896_Sound_Life_deep_longitudinal_immune_profiling,IL7R_and_KLRB1_with_NKG7_and_GNLY,4616,586,0.12694974003466206,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+065_A_GSE271896_GSE271896_Sound_Life_deep_longitudinal_immune_profiling,CD3D_or_CD3E_with_NKG7_and_GNLY,4616,216,0.04679376083188908,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+065_A_GSE271896_GSE271896_Sound_Life_deep_longitudinal_immune_profiling,all_three_patterns,4616,190,0.04116117850953206,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+093_C_Local_and_systemic_SARS-CoV-2_response_CELLxG_Local_and_systemic_SARS-CoV-2_response_CELLxGENE,IL7R_and_KLRB1,310,201,0.6483870967741936,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+093_C_Local_and_systemic_SARS-CoV-2_response_CELLxG_Local_and_systemic_SARS-CoV-2_response_CELLxGENE,CD3D_or_CD3E,310,134,0.432258064516129,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+093_C_Local_and_systemic_SARS-CoV-2_response_CELLxG_Local_and_systemic_SARS-CoV-2_response_CELLxGENE,NKG7_and_GNLY,310,167,0.5387096774193548,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+093_C_Local_and_systemic_SARS-CoV-2_response_CELLxG_Local_and_systemic_SARS-CoV-2_response_CELLxGENE,IL7R_and_KLRB1_with_CD3D_or_CD3E,310,95,0.3064516129032258,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+093_C_Local_and_systemic_SARS-CoV-2_response_CELLxG_Local_and_systemic_SARS-CoV-2_response_CELLxGENE,IL7R_and_KLRB1_with_NKG7_and_GNLY,310,113,0.36451612903225805,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+093_C_Local_and_systemic_SARS-CoV-2_response_CELLxG_Local_and_systemic_SARS-CoV-2_response_CELLxGENE,CD3D_or_CD3E_with_NKG7_and_GNLY,310,99,0.3193548387096774,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+093_C_Local_and_systemic_SARS-CoV-2_response_CELLxG_Local_and_systemic_SARS-CoV-2_response_CELLxGENE,all_three_patterns,310,67,0.2161290322580645,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+200_A_GSE275067_Immunobiology_of_Aging_Cohort_PBMC_Profiling,IL7R_and_KLRB1,1671,1345,0.8049072411729503,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+200_A_GSE275067_Immunobiology_of_Aging_Cohort_PBMC_Profiling,CD3D_or_CD3E,1671,283,0.16935966487133453,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+200_A_GSE275067_Immunobiology_of_Aging_Cohort_PBMC_Profiling,NKG7_and_GNLY,1671,198,0.118491921005386,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+200_A_GSE275067_Immunobiology_of_Aging_Cohort_PBMC_Profiling,IL7R_and_KLRB1_with_CD3D_or_CD3E,1671,232,0.13883901855176542,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+200_A_GSE275067_Immunobiology_of_Aging_Cohort_PBMC_Profiling,IL7R_and_KLRB1_with_NKG7_and_GNLY,1671,166,0.09934171154997008,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+200_A_GSE275067_Immunobiology_of_Aging_Cohort_PBMC_Profiling,CD3D_or_CD3E_with_NKG7_and_GNLY,1671,59,0.0353081986834231,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+200_A_GSE275067_Immunobiology_of_Aging_Cohort_PBMC_Profiling,all_three_patterns,1671,49,0.029323758228605626,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+300b_A_GSE132802_DRESS_DiHS_PBMC_only_scRNAseq,IL7R_and_KLRB1,9,4,0.4444444444444444,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+300b_A_GSE132802_DRESS_DiHS_PBMC_only_scRNAseq,CD3D_or_CD3E,9,8,0.8888888888888888,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+300b_A_GSE132802_DRESS_DiHS_PBMC_only_scRNAseq,NKG7_and_GNLY,9,3,0.3333333333333333,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+300b_A_GSE132802_DRESS_DiHS_PBMC_only_scRNAseq,IL7R_and_KLRB1_with_CD3D_or_CD3E,9,4,0.4444444444444444,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+300b_A_GSE132802_DRESS_DiHS_PBMC_only_scRNAseq,IL7R_and_KLRB1_with_NKG7_and_GNLY,9,3,0.3333333333333333,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+300b_A_GSE132802_DRESS_DiHS_PBMC_only_scRNAseq,CD3D_or_CD3E_with_NKG7_and_GNLY,9,3,0.3333333333333333,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+300b_A_GSE132802_DRESS_DiHS_PBMC_only_scRNAseq,all_three_patterns,9,3,0.3333333333333333,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+301_A_GSE163121_SLE_healthy_Bcell_PBMC_scRNAseq,IL7R_and_KLRB1,21,11,0.5238095238095238,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+301_A_GSE163121_SLE_healthy_Bcell_PBMC_scRNAseq,CD3D_or_CD3E,21,6,0.2857142857142857,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+301_A_GSE163121_SLE_healthy_Bcell_PBMC_scRNAseq,NKG7_and_GNLY,21,3,0.14285714285714285,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+301_A_GSE163121_SLE_healthy_Bcell_PBMC_scRNAseq,IL7R_and_KLRB1_with_CD3D_or_CD3E,21,3,0.14285714285714285,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+301_A_GSE163121_SLE_healthy_Bcell_PBMC_scRNAseq,IL7R_and_KLRB1_with_NKG7_and_GNLY,21,1,0.047619047619047616,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+301_A_GSE163121_SLE_healthy_Bcell_PBMC_scRNAseq,CD3D_or_CD3E_with_NKG7_and_GNLY,21,1,0.047619047619047616,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+301_A_GSE163121_SLE_healthy_Bcell_PBMC_scRNAseq,all_three_patterns,21,0,0.0,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+302_A_GSE314010_SLE_IL36_stim_PBMC_scRNAseq,IL7R_and_KLRB1,44,31,0.7045454545454546,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+302_A_GSE314010_SLE_IL36_stim_PBMC_scRNAseq,CD3D_or_CD3E,44,12,0.2727272727272727,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+302_A_GSE314010_SLE_IL36_stim_PBMC_scRNAseq,NKG7_and_GNLY,44,12,0.2727272727272727,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+302_A_GSE314010_SLE_IL36_stim_PBMC_scRNAseq,IL7R_and_KLRB1_with_CD3D_or_CD3E,44,8,0.18181818181818182,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+302_A_GSE314010_SLE_IL36_stim_PBMC_scRNAseq,IL7R_and_KLRB1_with_NKG7_and_GNLY,44,9,0.20454545454545456,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+302_A_GSE314010_SLE_IL36_stim_PBMC_scRNAseq,CD3D_or_CD3E_with_NKG7_and_GNLY,44,5,0.11363636363636363,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+302_A_GSE314010_SLE_IL36_stim_PBMC_scRNAseq,all_three_patterns,44,3,0.06818181818181818,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+304_A_Parse_10M_PBMC_12donor_90cytokines,IL7R_and_KLRB1,7485,49,0.0065464261857047425,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+304_A_Parse_10M_PBMC_12donor_90cytokines,CD3D_or_CD3E,7485,419,0.05597862391449566,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+304_A_Parse_10M_PBMC_12donor_90cytokines,NKG7_and_GNLY,7485,5,0.0006680026720106881,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+304_A_Parse_10M_PBMC_12donor_90cytokines,IL7R_and_KLRB1_with_CD3D_or_CD3E,7485,13,0.0017368069472277889,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+304_A_Parse_10M_PBMC_12donor_90cytokines,IL7R_and_KLRB1_with_NKG7_and_GNLY,7485,0,0.0,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+304_A_Parse_10M_PBMC_12donor_90cytokines,CD3D_or_CD3E_with_NKG7_and_GNLY,7485,4,0.0005344021376085504,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"
+304_A_Parse_10M_PBMC_12donor_90cytokines,all_three_patterns,7485,0,0.0,Each named gene raw count > 0; all six genes measured in all selected cells,"Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity"

paper-releases/pbisc-diffusion/updates/20261015T141849Z-rare-cell-7k-8k-panels/diffusion/dataset/server_1/model8/plans/rare_gene_context_expansion_20261015/coordination/ilc_training_support/codetection_diagnostics.py

@@ -0,0 +1,34 @@
+#!/usr/bin/env python3
+"""Descriptive positive-count overlaps; these are not annotation rules."""
+from __future__ import annotations
+
+import pandas as pd
+from audit_training_ilc import ROOT
+
+v = pd.read_parquet(ROOT / 'training_ilc_marker_counts.parquet')
+t = pd.read_csv(ROOT / 'locked_target_genes.csv')
+genes = ['IL7R', 'KLRB1', 'CD3D', 'CD3E', 'NKG7', 'GNLY']
+gid = t[t.gene_symbol.isin(genes)].set_index('gene_symbol').gene_id.to_dict()
+assert len(gid) == 6
+assert all(v[gid[g] + '__measured'].all() for g in genes)
+positive = {g: v[gid[g]].gt(0) for g in genes}
+both_il = positive['IL7R'] & positive['KLRB1']
+any_cd3 = positive['CD3D'] | positive['CD3E']
+both_nk = positive['NKG7'] & positive['GNLY']
+patterns = {'IL7R_and_KLRB1': both_il,
+            'CD3D_or_CD3E': any_cd3,
+            'NKG7_and_GNLY': both_nk,
+            'IL7R_and_KLRB1_with_CD3D_or_CD3E': both_il & any_cd3,
+            'IL7R_and_KLRB1_with_NKG7_and_GNLY': both_il & both_nk,
+            'CD3D_or_CD3E_with_NKG7_and_GNLY': any_cd3 & both_nk,
+            'all_three_patterns': both_il & any_cd3 & both_nk}
+rows = []
+for ds, d in [('ALL', v)] + list(v.groupby('dataset_id', sort=True)):
+    for name, mask in patterns.items():
+        n = int(mask.loc[d.index].sum())
+        rows.append({'dataset_id': ds, 'pattern': name, 'eligible_cells': len(d),
+                     'positive_cells': n, 'fraction': n/len(d),
+                     'detection_rule': 'Each named gene raw count > 0; all six genes measured in all selected cells',
+                     'interpretation': 'Descriptive overlap within source-labelled ILC records only; not classifier, relabelling or validated identity'})
+pd.DataFrame(rows).to_csv(ROOT / 'codetection_diagnostics.csv', index=False)
+print(pd.DataFrame(rows).query('dataset_id == "ALL"')[['pattern', 'positive_cells', 'fraction']].to_string(index=False))