Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261008T125804Z-deck-fig2-3-healthy-fig5-rare/diffusion/dataset/server_1/model8/plans/rare_states_paired_20261008/tools/rare_paired.py

@@ -0,0 +1,295 @@
+#!/usr/bin/env python3
+"""Rare cell states in the four paired disease cohorts: generated vs observed (CPU only; 2026-10-08).
+
+User decision of 2026-10-08 (Fig. 5): show rare-cell fractions and how well rare states are found, on the paired
+disease cohorts. Main comparison = 5 donors (LoRA-5 vs scSemiProfiler with the same 5 donors as representatives, plus
+the zero-shot PBISC versions); 5-fold CV is reported separately. No new generation.
+
+Generalises plans/rare_states_screen_20261004/tools/sepsis_states.py (sepsis 5-fold only) to every cohort and design:
+  cohorts  ye2026_sepsis, imperial_eskd, giroux2022, ji2025_chili (plans/paired_finetune_intradataset_20260927)
+  units    fewpair draw j = 0..4: zs_raw (fewshot_n5/zs__raw_all.npz), zs_cal, lora5 (draw<j>/lora__raw.npz),
+           scsemi5 (draw<j>/scsemi5__raw.npz); fivefold: zs__raw, zs__cal, real_lora__raw (pooled/), scsemi__raw
+           (pooled_scsemi_cv_20261002/)
+  labels   the benchmark's own frozen f01common scANVI labels of each scored call (scores/<ds>/.../arms/<arm>/labels);
+           so the generated counts must be the benchmark's own Poisson read-out (whole NPZ, seed 20260916, rows in NPZ
+           order), checked cell by cell against the recorded umi_scoring_genes.
+  states   plans/rare_states_screen_20261004/tools/states.json (literature marker rules; label states as they are).
+  depth    for marker states, observed parent cells are binomially thinned (seed 20261004) by one constant per unit,
+           arm and state to the generated parent median library on the measured panel genes (screen.py rule).
+Commands
+  extract DS fewpair J ARM | extract DS fivefold - ARM
+      NPZ read locally, or streamed from spark-0110 when it lives there; sha256 must equal its receipt. Writes
+      work/<ds>/<unit>/<arm>.npz (marker counts per labelled cell, measured-panel library, donor, AIFI L1/L2).
+  analyze
+      out/per_donor.csv     per unit, arm, state, donor: parent / state cells, observed (native, thinned), generated
+      out/summary.csv       per unit, arm, state: pooled % (generated, observed thinned), ratio, Spearman over donors,
+                            detection AUROC (generated % separating donors in the top vs bottom half of observed %)
+      out/contrasts.csv     per unit, arm, state: case-vs-control difference of donor means, observed vs generated
+"""
+from __future__ import annotations
+
+import hashlib
+import importlib.util
+import io
+import json
+import os
+import subprocess
+import sys
+from pathlib import Path
+
+import numpy as np
+import pandas as pd
+import scipy.sparse as sp
+from scipy.stats import mannwhitneyu, spearmanr
+
+PLAN = Path(__file__).resolve().parents[1]
+M8 = Path("/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion")
+FT = M8 / "plans/paired_finetune_intradataset_20260927"
+SCREEN = M8 / "plans/rare_states_screen_20261004/tools"
+sys.path.insert(0, str(SCREEN))
+from screen import call_state, parent_mask, state_cols  # noqa: E402
+
+CANON = Path("/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/strict_qc_v1_20260913/reference/"
+             "canonical_genes.tsv")
+POISSON_PY = M8 / "plans/aifi_scanvi_rebuild_20260916/poisson_readout.py"
+POISSON_SEED, THIN_SEED = 20260916, 20261004
+REMOTE = "spark-0110"
+COHORTS = ["ye2026_sepsis", "imperial_eskd", "giroux2022", "ji2025_chili"]
+CASE_CTRL = {"ye2026_sepsis": (["adult_sepsis", "ped_sepsis"], ["adult_HC", "adult_ICU_ctrl", "ped_HC", "ped_ICU_ctrl"]),
+             "imperial_eskd": (["covid_positive"], ["covid_negative"]),
+             "giroux2022": (["severe"], ["mild"]),
+             "ji2025_chili": (["TOL", "FU1"], ["BC", "AC"])}           # = tools/contrast_args.sh of the FT plan
+FEW_ARMS = ["zs_raw", "zs_cal", "scsemi5", "lora5"]
+POOL_ARMS = ["zs__raw", "zs__cal", "scsemi__raw", "real_lora__raw"]
+
+
+def unit_name(design: str, j) -> str:
+    return f"draw{j}" if design == "fewpair" else "fivefold"
+
+
+def locate(ds: str, design: str, j, arm: str) -> tuple[Path, Path]:
+    """(NPZ stem without suffix, scored-arm directory)"""
+    g, s = FT / "generated" / ds, FT / "scores" / ds
+    if design == "fewpair":
+        stem = {"zs_raw": g / "fewshot_n5/zs__raw_all", "zs_cal": g / f"fewshot_n5/draw{j}/zs__cal",
+                "lora5": g / f"fewshot_n5/draw{j}/lora__raw", "scsemi5": g / f"fewshot_n5/draw{j}/scsemi5__raw"}[arm]
+        sdir = s / f"fewshot_n5/draw{j}_scsemi5/arms/scsemi5" if arm == "scsemi5" else s / f"fewshot_n5/draw{j}/arms/{arm}"
+    else:
+        stem = g / "pooled_scsemi_cv_20261002/scsemi__raw" if arm == "scsemi__raw" else g / f"pooled/{arm}"
+        sdir = s / "scsemi_cv_20261002/pooled/arms/scsemi__raw" if arm == "scsemi__raw" else s / f"pooled/arms/{arm}"
+    return stem, sdir
+
+
+def fetch(path: Path) -> tuple[bytes, str]:
+    """file bytes from server_1, else streamed from spark-0110 (same absolute path); returns (bytes, host)"""
+    if path.is_file():
+        return path.read_bytes(), "server_1"
+    r = subprocess.run(["ssh", "-n", "-o", "BatchMode=yes", REMOTE, "cat", str(path)], stdin=subprocess.DEVNULL,
+                       capture_output=True, check=False)   # -n: never read the caller's stdin (job lists)
+    if r.returncode != 0:
+        raise SystemExit(f"{path}: not on server_1 and not readable on {REMOTE}")
+    return r.stdout, "server_2"
+
+
+def load_poisson():
+    spec = importlib.util.spec_from_file_location("poisson_readout", POISSON_PY)
+    mod = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(mod)
+    return mod
+
+
+def marker_genes(states: list[dict]) -> list[str]:
+    out = []
+    for st in states:
+        for g in st.get("genes", []) + [x for grp in st.get("require_any", []) for x in grp] + st.get("exclude", []):
+            if g not in out:
+                out.append(g)
+    return out
+
+
+def panel_setup(ds: str, states: list[dict]):
+    d = np.load(FT / "data" / ds / "cells_panel6k.npz", allow_pickle=False)
+    panel = d["panel_gene_ids"].astype(str)
+    measured = d["panel_measured"].any(0)
+    canon = pd.read_csv(CANON, sep="\t", dtype=str)
+    sym2id = dict(zip(canon.gene_symbol, canon.gene_id))
+    pos = {g: i for i, g in enumerate(panel)}
+    syms = [g for g in marker_genes(states) if sym2id.get(g) in pos and measured[pos[sym2id[g]]]]
+    return panel, measured, syms, np.asarray([pos[sym2id[g]] for g in syms], dtype=np.int64)
+
+
+def cmd_extract(ds: str, design: str, j, arm: str) -> None:
+    states = json.loads((SCREEN / "states.json").read_text())
+    stem, sdir = locate(ds, design, j, arm)
+    out = PLAN / "work" / ds / unit_name(design, j) / f"{arm}.npz"
+    if out.exists():
+        print(json.dumps({"skip": str(out)}))
+        return
+    if not (sdir / "generated_cells.parquet").exists():
+        raise SystemExit(f"{ds} {design} {j} {arm}: no scored labels at {sdir}")
+    panel, measured, syms, cols = panel_setup(ds, states)
+    rec_raw, _ = fetch(stem.with_suffix(".log"))
+    rec = json.loads(rec_raw)
+    raw, host = fetch(stem.with_suffix(".npz"))
+    sha = hashlib.sha256(raw).hexdigest()
+    if sha != rec["output_sha256"]:
+        raise SystemExit(f"{stem}.npz: sha256 {sha[:12]} != receipt {rec['output_sha256'][:12]}")
+    d = np.load(io.BytesIO(raw), allow_pickle=False)
+    del raw
+    if not np.array_equal(d["gene_ids"].astype(str), panel):
+        raise SystemExit(f"{stem}.npz: gene_ids differ from the 6k panel")
+    rates = np.ascontiguousarray(d["generated_counts"], dtype=np.float32)
+    n_rows = rates.shape[0]
+    full = load_poisson().poisson_integerise(rates, seed=POISSON_SEED, chunk_rows=4096)
+    del rates, d
+    gc = pd.read_parquet(sdir / "generated_cells.parquet")
+    lab = pd.read_parquet(sdir / "labels/labels.parquet")
+    if len(gc) != len(lab) or not (lab.sample_id.astype(str).str.removesuffix("|generated").values
+                                   == gc.donor.astype(str).values).all():
+        raise SystemExit(f"{sdir}: generated_cells and labels are not row-aligned")
+    X = full[gc.npz_row.to_numpy()].tocsr()
+    del full
+    ref = np.load(FT / "scoring_cache" / ds / "reference.npz", allow_pickle=False)
+    umi = np.asarray(X[:, ref["score_idx"]].sum(1)).ravel()
+    if not np.array_equal(umi, gc.umi_scoring_genes.to_numpy()):
+        raise SystemExit(f"{sdir}: Poisson read-out does not reproduce the benchmark "
+                         f"({int((umi != gc.umi_scoring_genes.to_numpy()).sum())} cells differ)")
+    lib = np.asarray(X[:, np.flatnonzero(measured)].sum(1)).ravel()
+    out.parent.mkdir(parents=True, exist_ok=True)
+    np.savez_compressed(out, markers=X[:, cols].toarray().astype(np.int32), marker_symbols=np.asarray(syms),
+                        lib_measured=lib, donor=np.asarray(gc.donor, dtype=str),
+                        aifi_l1=np.asarray(lab.aifi_l1, dtype=str), aifi_l2=np.asarray(lab.aifi_l2, dtype=str),
+                        receipt=json.dumps({"dataset": ds, "design": design, "draw": j, "arm": arm,
+                                            "npz": str(stem) + ".npz", "host": host, "npz_sha256": sha,
+                                            "npz_rows": int(n_rows), "cells": int(len(gc)), "labels": str(sdir),
+                                            "poisson_seed": POISSON_SEED, "umi_scoring_genes_reproduced": True}))
+    print(json.dumps({"dataset": ds, "unit": unit_name(design, j), "arm": arm, "host": host, "cells": int(len(gc)),
+                      "donors": int(gc.donor.nunique()), "median_lib_measured": float(np.median(lib))}))
+
+
+def observed(ds: str, states: list[dict]) -> dict:
+    _, measured, syms, cols = panel_setup(ds, states)
+    d = np.load(FT / "data" / ds / "cells_panel6k.npz", allow_pickle=False)
+    X = sp.csr_matrix((d["data"], d["indices"], d["indptr"]), shape=tuple(d["shape"]))
+    obs = pd.read_parquet(FT / "data" / ds / "cells_obs.parquet", columns=["sample_id", "group"])
+    lab = pd.read_parquet(FT / "scoring_cache" / ds / "observed_labels/labels/labels.parquet")
+    if len(lab) != X.shape[0] or (lab.sample_id.astype(str).values != obs.sample_id.astype(str).values).any():
+        raise SystemExit(f"{ds}: observed labels not aligned with cells_panel6k.npz")
+    return {"markers": X[:, cols].toarray().astype(np.int32), "symbols": syms,
+            "lib": np.asarray(X[:, np.flatnonzero(measured)].sum(1)).ravel(),
+            "donor": obs.sample_id.astype(str).to_numpy(), "lab": lab[["aifi_l1", "aifi_l2"]].reset_index(drop=True),
+            "groups": obs.drop_duplicates("sample_id").set_index("sample_id").group}
+
+
+def auroc_high_low(gen: pd.Series, obs: pd.Series) -> float:
+    """AUROC of the generated % for donors in the top vs bottom half of the observed % (ties at the median dropped)"""
+    ok = gen.notna() & obs.notna()
+    g, o = gen[ok], obs[ok]
+    if o.nunique() < 2:
+        return np.nan
+    med = o.median()
+    hi, lo = g[o > med], g[o < med]
+    if len(hi) < 2 or len(lo) < 2:
+        return np.nan
+    return float(mannwhitneyu(hi, lo, alternative="two-sided").statistic / (len(hi) * len(lo)))
+
+
+def cmd_analyze() -> None:
+    states = json.loads((SCREEN / "states.json").read_text())
+    per, summ, con = [], [], []
+    for ds in COHORTS:
+        ob = observed(ds, states)
+        case, ctrl = CASE_CTRL[ds]
+        units = [("fewpair", j) for j in range(5)] + [("fivefold", None)]
+        for design, j in units:
+            unit = unit_name(design, j)
+            arms = FEW_ARMS if design == "fewpair" else POOL_ARMS
+            gens = {a: np.load(PLAN / "work" / ds / unit / f"{a}.npz", allow_pickle=False) for a in arms}
+            donors = sorted(set(gens[arms[0]]["donor"]))
+            for a in arms:
+                if sorted(set(gens[a]["donor"])) != donors:
+                    raise SystemExit(f"{ds} {unit} {a}: scored donors differ from {arms[0]}")
+            keep = np.isin(ob["donor"], donors)
+            om, ol, od = ob["markers"][keep], ob["lab"][keep].reset_index(drop=True), ob["donor"][keep]
+            olib = ob["lib"][keep]
+            rng = np.random.default_rng([THIN_SEED, COHORTS.index(ds), (j if j is not None else 9)])
+            for arm in arms:
+                g = gens[arm]
+                syms = list(g["marker_symbols"].astype(str))
+                if syms != ob["symbols"]:
+                    raise SystemExit(f"{ds} {unit} {arm}: marker columns differ from the observed side")
+                s2c = {s: i for i, s in enumerate(syms)}
+                lg = pd.DataFrame({"aifi_l1": g["aifi_l1"], "aifi_l2": g["aifi_l2"]})
+                for st in states:
+                    po, pg = parent_mask(ol, st["parent"]), parent_mask(lg, st["parent"])
+                    if st["kind"] == "label":
+                        ho = ol.aifi_l2.isin(st["labels"]).to_numpy() & po
+                        hg = lg.aifi_l2.isin(st["labels"]).to_numpy() & pg
+                        ht, thin = ho, 1.0
+                    else:
+                        _, pcols, req, exc = state_cols(st, {s: s for s in syms}, s2c)
+                        c = pcols + [x for r in req for x in r] + exc
+                        thin = min(1.0, float(np.median(g["lib_measured"][pg])) / float(np.median(olib[po])))
+                        co = om[:, c].astype(np.int64)
+                        ho = call_state(co, st, len(pcols), req, exc) & po
+                        ct = rng.binomial(co, thin) if thin < 1.0 else co
+                        ht = call_state(ct, st, len(pcols), req, exc) & po
+                        hg = call_state(g["markers"][:, c], st, len(pcols), req, exc) & pg
+                    rows = []
+                    for s in donors:
+                        mo, mg = od == s, g["donor"] == s
+                        rows.append({"dataset": ds, "design": design, "unit": unit, "arm": arm, "state": st["name"],
+                                     "donor": s, "group": ob["groups"].get(s),
+                                     "obs_parent": int((po & mo).sum()), "obs_state": int((ho & mo).sum()),
+                                     "obsthin_state": int((ht & mo).sum()), "gen_parent": int((pg & mg).sum()),
+                                     "gen_state": int((hg & mg).sum())})
+                    t = pd.DataFrame(rows)
+                    t["obsthin_pct"] = 100 * t.obsthin_state / t.obs_parent.where(t.obs_parent > 0)
+                    t["gen_pct"] = 100 * t.gen_state / t.gen_parent.where(t.gen_parent > 0)
+                    per.append(t)
+                    ok = t[["gen_pct", "obsthin_pct"]].dropna()
+                    rho = float(spearmanr(ok.gen_pct, ok.obsthin_pct).correlation) if len(ok) > 2 and \
+                        ok.gen_pct.nunique() > 1 and ok.obsthin_pct.nunique() > 1 else np.nan
+                    ro = 100 * t.obsthin_state.sum() / max(t.obs_parent.sum(), 1)
+                    rg = 100 * t.gen_state.sum() / max(t.gen_parent.sum(), 1)
+                    summ.append({"dataset": ds, "design": design, "unit": unit, "arm": arm, "state": st["name"],
+                                 "thin": round(thin, 3), "gen_state": int(t.gen_state.sum()),
+                                 "gen_parent": int(t.gen_parent.sum()), "obs_state_thinned": int(t.obsthin_state.sum()),
+                                 "obs_parent": int(t.obs_parent.sum()), "obs_state_native": int(t.obs_state.sum()),
+                                 "gen_pct": rg, "obs_pct": ro, "gen_over_obs": rg / ro if ro else np.nan,
+                                 "spearman_donors": rho,
+                                 "auroc_high_low": auroc_high_low(t.gen_pct, t.obsthin_pct),
+                                 "donors_with_state_obs": int((t.obsthin_state > 0).sum()),
+                                 "of_those_with_state_gen": int(((t.obsthin_state > 0) & (t.gen_state > 0)).sum()),
+                                 "n_donors": len(t)})
+                    ga, gb = t[t.group.isin(case)], t[t.group.isin(ctrl)]
+                    con.append({"dataset": ds, "design": design, "unit": unit, "arm": arm, "state": st["name"],
+                                "n_case": len(ga), "n_ctrl": len(gb),
+                                "obs_diff": ga.obsthin_pct.mean() - gb.obsthin_pct.mean(),
+                                "gen_diff": ga.gen_pct.mean() - gb.gen_pct.mean()})
+            print(json.dumps({"dataset": ds, "unit": unit, "arms": arms, "donors": len(donors)}), flush=True)
+    C = pd.DataFrame(con)
+    C["same_direction"] = np.sign(C.obs_diff) == np.sign(C.gen_diff)
+    out = PLAN / "out"
+    out.mkdir(exist_ok=True)
+    pd.concat(per).to_csv(out / "per_donor.csv", index=False)
+    pd.DataFrame(summ).to_csv(out / "summary.csv", index=False)
+    C.to_csv(out / "contrasts.csv", index=False)
+    print(json.dumps({"rows_summary": len(summ), "rows_contrasts": len(C)}))
+
+
+def main() -> int:
+    if os.environ.get("CUDA_VISIBLE_DEVICES") != "":
+        raise SystemExit("run with CUDA_VISIBLE_DEVICES=''")
+    a = sys.argv[1:]
+    if len(a) == 5 and a[0] == "extract":
+        cmd_extract(a[1], a[2], None if a[3] == "-" else int(a[3]), a[4])
+    elif a == ["analyze"]:
+        cmd_analyze()
+    else:
+        raise SystemExit(__doc__)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

paper-releases/pbisc-diffusion/updates/20261008T125804Z-deck-fig2-3-healthy-fig5-rare/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/figure_plan.json

@@ -0,0 +1,396 @@
+{
+ "source": "Nature Style Manuscript Writing.pdf (paper-release-20260928/reference-papers): six main figures, up to 17 supplementary. User decisions of 2026-10-06: no PBISC-VAE; scSemiProfiler, a VAE-based method, is the comparison. User decisions of 2026-10-08 (supersede the earlier figure split): Fig. 2 shows no disease metrics; Fig. 3 compares with scSemiProfiler on healthy samples, with time as a main result; Figs 4-6 use the disease cohorts: Fig. 4 disease benchmarks (PBISC versions, scSemiProfiler, and BayesPrism / CIBERSORTx labelled as cell-type deconvolution), Fig. 5 rare-cell fractions and detection, Fig. 6 new findings from the bulk-only sepsis and COVID samples (K = 1,000 and 4,000).",
+ "status_legend": {
+  "Ready": "slide exists and fits the panel",
+  "Adapt": "slide exists; change its scope or style",
+  "Draft": "new draft exists, not yet approved",
+  "Build": "results exist; no slide yet",
+  "Run": "needs a new analysis or experiment",
+  "Plan": "not planned yet"
+ },
+ "figures": [
+  {
+   "n": 1,
+   "title": "Design and performance of PBISC",
+   "claim": "Diffusion on each gene's value and detection, conditioned on bulk, writes realistic blood cells.",
+   "panels": [
+    {
+     "p": "a",
+     "content": "Backbones compared: scDiffusion, SDD, PBISC-Diffusion",
+     "evidence": [
+      "Backbone: SDD → PBISC-Diffusion"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "b",
+     "content": "Overview: pretraining, diffusion, generation",
+     "evidence": [
+      "Overview draft (Nature style)"
+     ],
+     "status": "Draft"
+    },
+    {
+     "p": "c",
+     "content": "What the bulk condition contains",
+     "evidence": [
+      "Bulk conditioning"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "d",
+     "content": "Pretraining corpus: assays, conditions, cell types",
+     "evidence": [
+      "Pretraining corpus"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "e",
+     "content": "Width and context: lower loss, same biology",
+     "evidence": [
+      "Width & context ablation: loss saturation",
+      "Width & context ablation: benchmark"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "f",
+     "content": "Gene panel: 2k, 3k, 6k, 15k",
+     "evidence": [
+      "Gene context: 2k / 3k / 6k / 15k"
+     ],
+     "status": "Ready"
+    }
+   ],
+   "candidates": [],
+   "details": [
+    "PBISC-Diffusion: bulk in, cells out",
+    "Denoiser: joint value + support diffusion"
+   ]
+  },
+  {
+   "n": 2,
+   "title": "Fine-tuning PBISC with LoRA",
+   "claim": "On a healthy cohort, a few pairs and a LoRA of minutes give cells close to the real ones. No disease metrics.",
+   "panels": [
+    {
+     "p": "a",
+     "content": "Real bulk differs from the training input",
+     "evidence": [
+      "Real bulk ≠ pseudobulk (model input)",
+      "Same shift in three paired cohorts"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "b",
+     "content": "Without fine-tuning, cells fall off the real manifold",
+     "evidence": [
+      "Consequence: cells off-manifold"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "c",
+     "content": "LoRA on paired bulk and cells; how it is tested",
+     "evidence": [
+      "Fine-tuning with paired bulk + sc",
+      "Evaluation design"
+     ],
+     "status": "Ready",
+     "note": "design layout; example cohort sepsis"
+    },
+    {
+     "p": "d",
+     "content": "How generated cells are scored (cells and composition)",
+     "evidence": [
+      "How we score generated cells"
+     ],
+     "status": "Ready",
+     "note": "metric table for Figs 2-4"
+    },
+    {
+     "p": "e",
+     "content": "Healthy cohort: precision, coverage, composition",
+     "evidence": [
+      "Healthy cohort: LoRA cells cover the real ones"
+     ],
+     "status": "Ready",
+     "note": "ImMicro only"
+    },
+    {
+     "p": "f",
+     "content": "Healthy cohort: classifier AUROC, MMD, E_N",
+     "evidence": [
+      "Healthy cohort: realism beyond precision/coverage"
+     ],
+     "status": "Ready",
+     "note": "ImMicro only"
+    },
+    {
+     "p": "g",
+     "content": "Full, LoRA and condition-encoder fine-tuning",
+     "evidence": [
+      "Healthy cohort: full vs LoRA vs condition encoder"
+     ],
+     "status": "Ready",
+     "note": "ImMicro only"
+    },
+    {
+     "p": "h",
+     "content": "How many pairs are needed",
+     "evidence": [
+      "Healthy cohort: how many pairs"
+     ],
+     "status": "Ready",
+     "note": "ImMicro only"
+    },
+    {
+     "p": "i",
+     "content": "LoRA rank and K choices",
+     "evidence": [
+      "Healthy cohort: LoRA rank and K"
+     ],
+     "status": "Ready",
+     "note": "ImMicro only; supplementary candidate"
+    }
+   ],
+   "candidates": [],
+   "details": []
+  },
+  {
+   "n": 3,
+   "title": "PBISC vs scSemiProfiler on healthy donors",
+   "claim": "On healthy donors, LoRA cells are closer to the real ones than scSemiProfiler's, in minutes instead of days.",
+   "panels": [
+    {
+     "p": "a",
+     "content": "Five donors: generated cells and realism",
+     "evidence": [
+      "Healthy donors, 5 pairs: LoRA-5 vs scSemiProfiler-5"
+     ],
+     "status": "Ready",
+     "note": "ImMicro only"
+    },
+    {
+     "p": "b",
+     "content": "Same training pairs (5-fold) and equal pair budget (5-40)",
+     "evidence": [
+      "Healthy donors: the same pairs, 5 to all"
+     ],
+     "status": "Ready",
+     "note": "ImMicro only"
+    },
+    {
+     "p": "c",
+     "content": "scSemiProfiler copies its representative",
+     "evidence": [
+      "Healthy donors: scSemiProfiler copies its representative"
+     ],
+     "status": "Ready",
+     "note": "ImMicro only"
+    },
+    {
+     "p": "d",
+     "content": "Time: GPU hours and days per experiment",
+     "evidence": [
+      "Time: days for scSemiProfiler, minutes for LoRA",
+      "5 donors: time to apply"
+     ],
+     "status": "Ready",
+     "note": "all cohorts shown: time is a property of the method"
+    }
+   ],
+   "candidates": [],
+   "details": []
+  },
+  {
+   "n": 4,
+   "title": "Disease signal on paired cohorts",
+   "claim": "LoRA cells recover the paired single cells' disease signal; compared with other PBISC versions, scSemiProfiler and deconvolution.",
+   "panels": [
+    {
+     "p": "a",
+     "content": "Five paired cohorts",
+     "evidence": [
+      "Five paired cohorts"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "b",
+     "content": "With 5 pairs: LoRA-5 vs PBISC versions and scSemiProfiler-5",
+     "evidence": [
+      "Disease signal with 5 pairs: LoRA-5 vs others"
+     ],
+     "status": "Ready",
+     "note": "main comparison"
+    },
+    {
+     "p": "c",
+     "content": "5-fold CV, reported separately",
+     "evidence": [
+      "5-fold CV: disease signal"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "d",
+     "content": "Disease genes at every list length (CAT)",
+     "evidence": [
+      "Disease genes at every list length (CAT)"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "e",
+     "content": "vs BayesPrism and CIBERSORTx (cell-type deconvolution)",
+     "evidence": [
+      "Cell-type deconvolution, same pairs: disease signal"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "f",
+     "content": "Lupus benchmark (sc-pseudobulk input, zero-shot)",
+     "evidence": [
+      "SLE benchmark: PBISC vs scSemiProfiler",
+      "SLE benchmark: per cell type"
+     ],
+     "status": "Ready"
+    }
+   ],
+   "candidates": [],
+   "details": [
+    "Sepsis: fine-tuning variants",
+    "Pairs needed: donor and disease signal",
+    "Cell-type deconvolution, same pairs: composition"
+   ]
+  },
+  {
+   "n": 5,
+   "title": "Rare cell states in disease cohorts",
+   "claim": "LoRA cells carry rare disease states near their observed fractions, in the donors that have them.",
+   "panels": [
+    {
+     "p": "a",
+     "content": "Rare-state fractions with 5 pairs, four paired cohorts",
+     "evidence": [
+      "Rare states with 5 pairs: how many cells"
+     ],
+     "status": "Ready",
+     "note": "LoRA-5 vs scSemiProfiler-5"
+    },
+    {
+     "p": "b",
+     "content": "Finding rare states: right donors, disease direction",
+     "evidence": [
+      "Rare states with 5 pairs: in the right donors"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "c",
+     "content": "5-fold CV, reported separately",
+     "evidence": [
+      "5-fold CV: rare states"
+     ],
+     "status": "Ready"
+    },
+    {
+     "p": "d",
+     "content": "Behçet: C1Q-high monocytes from real bulk",
+     "evidence": [
+      "Consequence: disease direction flips"
+     ],
+     "status": "Adapt",
+     "note": "add the LoRA-3 result (C1Q-high BD > HC)"
+    },
+    {
+     "p": "e",
+     "content": "Diseases absent from the corpus: Kawasaki MS1, SLE ISG-high T",
+     "evidence": [],
+     "status": "Build",
+     "note": "rare_states_screen_20261004 results"
+    }
+   ],
+   "candidates": [],
+   "details": []
+  },
+  {
+   "n": 6,
+   "title": "Discovery from public bulk-only data",
+   "claim": "New findings from LoRA cells of bulk-only sepsis and COVID dialysis samples (K = 1,000 and 4,000).",
+   "panels": [
+    {
+     "p": "a",
+     "content": "Bulk-only samples generated: sepsis 122, COVID dialysis 164",
+     "evidence": [
+      "Next: cohorts with a paired subset (planned)"
+     ],
+     "status": "Adapt",
+     "note": "update: K = 1,000 and 4,000 done"
+    },
+    {
+     "p": "b",
+     "content": "New findings",
+     "evidence": [],
+     "status": "Run",
+     "note": "earlier screen: 0 confirmed beyond the bulk"
+    }
+   ],
+   "candidates": [],
+   "details": []
+  }
+ ],
+ "supplementary": {
+  "slides": [
+   "5 donors: LoRA vs scSemiProfiler — cells",
+   "5 donors: realism beyond precision/coverage",
+   "5 donors: precision/coverage at k = 3, 5, 10",
+   "5-fold: realism beyond precision/coverage",
+   "Same training pairs: LoRA vs scSemiProfiler",
+   "scSemiProfiler copies its representative",
+   "Full vs LoRA vs cond.-encoder-only",
+   "Pairs needed: composition, coverage, training steps",
+   "LoRA rank: r = 4, 16, 64",
+   "K diversification: no gain",
+   "Pairs from other cohorts do not transfer",
+   "Longer training on other cohorts: gap stays",
+   "No pretraining: same cohort, from scratch",
+   "No pretraining: other cohorts, from scratch",
+   "Rare states: from scratch vs pretrained"
+  ],
+  "planned": [
+   {
+    "p": "S",
+    "content": "Influenza exposure from raw clinical bulk (Randolph)",
+    "evidence": [],
+    "status": "Build",
+    "note": "manuscript result; placement open"
+   }
+  ],
+  "notes": {
+   "Pairs from other cohorts do not transfer": "moved from Fig. 5; placement open",
+   "Longer training on other cohorts: gap stays": "moved from Fig. 5; placement open",
+   "5 donors: LoRA vs scSemiProfiler — cells": "all five cohorts",
+   "5 donors: realism beyond precision/coverage": "all five cohorts",
+   "5 donors: precision/coverage at k = 3, 5, 10": "all five cohorts",
+   "5-fold: realism beyond precision/coverage": "all five cohorts",
+   "Same training pairs: LoRA vs scSemiProfiler": "all five cohorts",
+   "scSemiProfiler copies its representative": "all five cohorts",
+   "Full vs LoRA vs cond.-encoder-only": "all five cohorts",
+   "Pairs needed: composition, coverage, training steps": "all five cohorts",
+   "LoRA rank: r = 4, 16, 64": "all five cohorts",
+   "K diversification: no gain": "all five cohorts"
+  }
+ },
+ "appendix": [
+  "Appendix: scSemiProfiler without output threshold"
+ ],
+ "discussion": [
+  "Next: other tissues from a few dozen pairs"
+ ]
+}

paper-releases/pbisc-diffusion/updates/20261008T125804Z-deck-fig2-3-healthy-fig5-rare/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/figure_plan_slide_text.json

@@ -0,0 +1,132 @@
+{
+ "section": "figure-plan",
+ "scripts": [
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/tools/figure_plan.py"
+ ],
+ "generated_from": "tools/figure_plan.py (rerun to refresh; do not edit by hand)",
+ "slides": [
+  {
+   "title": "Manuscript plan: six main figures",
+   "layout": "full",
+   "figures": [
+    "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/figplan_overview.png"
+   ],
+   "keywords": [],
+   "footnote_short": "Nature style: short abstract and introduction, descriptive results, short discussion, detailed methods; up to 17 supplementary figures.",
+   "footnote": "Nature Style Manuscript Writing.pdf (paper-release-20260928/reference-papers): six main figures, up to 17 supplementary. User decisions of 2026-10-06: no PBISC-VAE; scSemiProfiler, a VAE-based method, is the comparison. User decisions of 2026-10-08 (supersede the earlier figure split): Fig. 2 shows no disease metrics; Fig. 3 compares with scSemiProfiler on healthy samples, with time as a main result; Figs 4-6 use the disease cohorts: Fig. 4 disease benchmarks (PBISC versions, scSemiProfiler, and BayesPrism / CIBERSORTx labelled as cell-type deconvolution), Fig. 5 rare-cell fractions and detection, Fig. 6 new findings from the bulk-only sepsis and COVID samples (K = 1,000 and 4,000). Status: Ready = slide exists and fits the panel; Adapt = slide exists; change its scope or style; Draft = new draft exists, not yet approved; Build = results exist; no slide yet; Run = needs a new analysis or experiment; Plan = not planned yet.",
+   "numbers": []
+  },
+  {
+   "title": "Figure 1 plan: Design and performance of PBISC",
+   "layout": "full",
+   "figures": [
+    "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/figplan_fig1.png"
+   ],
+   "keywords": [],
+   "footnote_short": "Diffusion on each gene's value and detection, conditioned on bulk, writes realistic blood cells. 5 of 6 panels ready.",
+   "footnote": "Panels: a) Backbones compared: scDiffusion, SDD, PBISC-Diffusion [Ready]. b) Overview: pretraining, diffusion, generation [Draft]. c) What the bulk condition contains [Ready]. d) Pretraining corpus: assays, conditions, cell types [Ready]. e) Width and context: lower loss, same biology [Ready]. f) Gene panel: 2k, 3k, 6k, 15k [Ready]. Details in this section: PBISC-Diffusion: bulk in, cells out; Denoiser: joint value + support diffusion.",
+   "numbers": []
+  },
+  {
+   "title": "Figure 2 plan: Fine-tuning PBISC with LoRA",
+   "layout": "full",
+   "figures": [
+    "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/figplan_fig2.png"
+   ],
+   "keywords": [],
+   "footnote_short": "On a healthy cohort, a few pairs and a LoRA of minutes give cells close to the real ones. No disease metrics. 9 of 9 panels ready.",
+   "footnote": "Panels: a) Real bulk differs from the training input [Ready]. b) Without fine-tuning, cells fall off the real manifold [Ready]. c) LoRA on paired bulk and cells; how it is tested [Ready; design layout; example cohort sepsis]. d) How generated cells are scored (cells and composition) [Ready; metric table for Figs 2-4]. e) Healthy cohort: precision, coverage, composition [Ready; ImMicro only]. f) Healthy cohort: classifier AUROC, MMD, E_N [Ready; ImMicro only]. g) Full, LoRA and condition-encoder fine-tuning [Ready; ImMicro only]. h) How many pairs are needed [Ready; ImMicro only]. i) LoRA rank and K choices [Ready; ImMicro only; supplementary candidate].",
+   "numbers": []
+  },
+  {
+   "title": "Figure 3 plan: PBISC vs scSemiProfiler on healthy donors",
+   "layout": "full",
+   "figures": [
+    "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/figplan_fig3.png"
+   ],
+   "keywords": [],
+   "footnote_short": "On healthy donors, LoRA cells are closer to the real ones than scSemiProfiler's, in minutes instead of days. 4 of 4 panels ready.",
+   "footnote": "Panels: a) Five donors: generated cells and realism [Ready; ImMicro only]. b) Same training pairs (5-fold) and equal pair budget (5-40) [Ready; ImMicro only]. c) scSemiProfiler copies its representative [Ready; ImMicro only]. d) Time: GPU hours and days per experiment [Ready; all cohorts shown: time is a property of the method].",
+   "numbers": []
+  },
+  {
+   "title": "Figure 4 plan: Disease signal on paired cohorts",
+   "layout": "full",
+   "figures": [
+    "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/figplan_fig4.png"
+   ],
+   "keywords": [],
+   "footnote_short": "LoRA cells recover the paired single cells' disease signal; compared with other PBISC versions, scSemiProfiler and deconvolution. 6 of 6 panels ready.",
+   "footnote": "Panels: a) Five paired cohorts [Ready]. b) With 5 pairs: LoRA-5 vs PBISC versions and scSemiProfiler-5 [Ready; main comparison]. c) 5-fold CV, reported separately [Ready]. d) Disease genes at every list length (CAT) [Ready]. e) vs BayesPrism and CIBERSORTx (cell-type deconvolution) [Ready]. f) Lupus benchmark (sc-pseudobulk input, zero-shot) [Ready]. Details in this section: Sepsis: fine-tuning variants; Pairs needed: donor and disease signal; Cell-type deconvolution, same pairs: composition.",
+   "numbers": []
+  },
+  {
+   "title": "Figure 5 plan: Rare cell states in disease cohorts",
+   "layout": "full",
+   "figures": [
+    "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/figplan_fig5.png"
+   ],
+   "keywords": [],
+   "footnote_short": "LoRA cells carry rare disease states near their observed fractions, in the donors that have them. 3 of 5 panels ready.",
+   "footnote": "Panels: a) Rare-state fractions with 5 pairs, four paired cohorts [Ready; LoRA-5 vs scSemiProfiler-5]. b) Finding rare states: right donors, disease direction [Ready]. c) 5-fold CV, reported separately [Ready]. d) Behçet: C1Q-high monocytes from real bulk [Adapt; add the LoRA-3 result (C1Q-high BD > HC)]. e) Diseases absent from the corpus: Kawasaki MS1, SLE ISG-high T [Build; rare_states_screen_20261004 results].",
+   "numbers": []
+  },
+  {
+   "title": "Figure 6 plan: Discovery from public bulk-only data",
+   "layout": "full",
+   "figures": [
+    "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/figplan_fig6.png"
+   ],
+   "keywords": [],
+   "footnote_short": "New findings from LoRA cells of bulk-only sepsis and COVID dialysis samples (K = 1,000 and 4,000). 0 of 2 panels ready.",
+   "footnote": "Panels: a) Bulk-only samples generated: sepsis 122, COVID dialysis 164 [Adapt; update: K = 1,000 and 4,000 done]. b) New findings [Run; earlier screen: 0 confirmed beyond the bulk].",
+   "numbers": []
+  },
+  {
+   "title": "Supplementary plan",
+   "layout": "full",
+   "figures": [
+    "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/figplan_supp.png"
+   ],
+   "keywords": [],
+   "footnote_short": "Bulk deconvolution methods appear here only: PBISC generates single cells, a different task.",
+   "footnote": "Supplementary: 5 donors: LoRA vs scSemiProfiler — cells; 5 donors: realism beyond precision/coverage; 5 donors: precision/coverage at k = 3, 5, 10; 5-fold: realism beyond precision/coverage; Same training pairs: LoRA vs scSemiProfiler; scSemiProfiler copies its representative; Full vs LoRA vs cond.-encoder-only; Pairs needed: composition, coverage, training steps; LoRA rank: r = 4, 16, 64; K diversification: no gain; Pairs from other cohorts do not transfer; Longer training on other cohorts: gap stays; No pretraining: same cohort, from scratch; No pretraining: other cohorts, from scratch; Rare states: from scratch vs pretrained. Planned: Influenza exposure from raw clinical bulk (Randolph) [Build; manuscript result; placement open]. Appendix: Appendix: scSemiProfiler without output threshold.",
+   "numbers": []
+  },
+  {
+   "title": "Overview draft (Nature style)",
+   "layout": "full",
+   "figures": [
+    "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/fig1b_overview_nature.png"
+   ],
+   "keywords": [
+    "Each cell is learned with its own bulk",
+    "Diffusion on each gene's value and detection",
+    "One bulk in, K cells out; optional LoRA"
+   ],
+   "footnote_short": "Draft for Fig. 1b; schematic, no data. Plan: manuscript editorial/fig1_nature_redesign_plan.",
+   "footnote": "Candidate replacement for the engineering diagram (kept in this section as details and for Methods). Source: paper-release-20260928/manuscript/figures/fig01_architecture_nature.svg (scripts/build_figure1_architecture_nature.py).",
+   "numbers": []
+  },
+  {
+   "title": "Pretraining corpus",
+   "layout": "full",
+   "figures": [
+    "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/fig1d_corpus.png"
+   ],
+   "keywords": [
+    "41.4 M cells, 22 datasets, 3,199 donors",
+    "88.5% healthy donors; 23.6% cultured in vitro",
+    "Labels describe the corpus; never model input"
+   ],
+   "footnote_short": "Detailed version: Supplementary Fig. S1 of the manuscript.",
+   "footnote": "Source: paper-release-20260928/manuscript/source_data/corpus/ (build_corpus_tables.py) and figures/fig01_corpus_nature.png (build_figure1_corpus.py).",
+   "numbers": [
+    {
+     "value": "41,395,184 cells; 22 datasets; 3,199 donors; healthy 88.5%; in vitro 23.6%",
+     "source": "manuscript/source_data/corpus/corpus_summary.json"
+    }
+   ]
+  }
+ ]
+}