Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261008T022410Z-deconv-scsemi-deck/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores/ye2026_sepsis/scsemi_budget_20261002/fold0/arms/scsemi_n10__raw/deg_top50_lists.csv

@@ -0,0 +1,252 @@
+celltype,source,rank,gene_id
+T cell,observed,1,ENSG00000110077
+T cell,observed,2,ENSG00000163606
+T cell,observed,3,ENSG00000102245
+T cell,observed,4,ENSG00000171115
+T cell,observed,5,ENSG00000262655
+T cell,observed,6,ENSG00000216937
+T cell,observed,7,ENSG00000135426
+T cell,observed,8,ENSG00000145687
+T cell,observed,9,ENSG00000173208
+T cell,observed,10,ENSG00000166750
+T cell,observed,11,ENSG00000167106
+T cell,observed,12,ENSG00000070190
+T cell,observed,13,ENSG00000115415
+T cell,observed,14,ENSG00000082512
+T cell,observed,15,ENSG00000166432
+T cell,observed,16,ENSG00000160593
+T cell,observed,17,ENSG00000116138
+T cell,observed,18,ENSG00000156299
+T cell,observed,19,ENSG00000185621
+T cell,observed,20,ENSG00000133805
+T cell,observed,21,ENSG00000137992
+T cell,observed,22,ENSG00000174123
+T cell,observed,23,ENSG00000178562
+T cell,observed,24,ENSG00000071575
+T cell,observed,25,ENSG00000107863
+T cell,observed,26,ENSG00000116525
+T cell,observed,27,ENSG00000121289
+T cell,observed,28,ENSG00000181722
+T cell,observed,29,ENSG00000188641
+T cell,observed,30,ENSG00000124459
+T cell,observed,31,ENSG00000197548
+T cell,observed,32,ENSG00000138641
+T cell,observed,33,ENSG00000152683
+T cell,observed,34,ENSG00000183337
+T cell,observed,35,ENSG00000172575
+T cell,observed,36,ENSG00000171817
+T cell,observed,37,ENSG00000173065
+T cell,observed,38,ENSG00000100242
+T cell,observed,39,ENSG00000011523
+T cell,observed,40,ENSG00000173214
+T cell,observed,41,ENSG00000177272
+T cell,observed,42,ENSG00000158623
+T cell,observed,43,ENSG00000128513
+T cell,observed,44,ENSG00000100351
+T cell,observed,45,ENSG00000070269
+T cell,observed,46,ENSG00000104689
+T cell,observed,47,ENSG00000133103
+T cell,observed,48,ENSG00000065413
+T cell,observed,49,ENSG00000134146
+T cell,observed,50,ENSG00000170581
+B cell,observed,1,ENSG00000113088
+B cell,observed,2,ENSG00000066923
+B cell,observed,3,ENSG00000102245
+B cell,observed,4,ENSG00000184613
+B cell,observed,5,ENSG00000058866
+B cell,observed,6,ENSG00000215788
+B cell,observed,7,ENSG00000127152
+B cell,observed,8,ENSG00000081059
+B cell,observed,9,ENSG00000171643
+B cell,observed,10,ENSG00000172673
+B cell,observed,11,ENSG00000010610
+B cell,observed,12,ENSG00000182489
+B cell,observed,13,ENSG00000156052
+B cell,observed,14,ENSG00000163689
+B cell,observed,15,ENSG00000248905
+B cell,observed,16,ENSG00000091129
+B cell,observed,17,ENSG00000109943
+B cell,observed,18,ENSG00000066294
+B cell,observed,19,ENSG00000188322
+B cell,observed,20,ENSG00000174123
+B cell,observed,21,ENSG00000162804
+B cell,observed,22,ENSG00000172164
+B cell,observed,23,ENSG00000152495
+B cell,observed,24,ENSG00000153898
+B cell,observed,25,ENSG00000104490
+B cell,observed,26,ENSG00000122223
+B cell,observed,27,ENSG00000196975
+B cell,observed,28,ENSG00000137766
+B cell,observed,29,ENSG00000172575
+B cell,observed,30,ENSG00000126821
+B cell,observed,31,ENSG00000114013
+B cell,observed,32,ENSG00000140853
+B cell,observed,33,ENSG00000203710
+B cell,observed,34,ENSG00000216937
+B cell,observed,35,ENSG00000188906
+B cell,observed,36,ENSG00000147138
+B cell,observed,37,ENSG00000100351
+B cell,observed,38,ENSG00000135426
+B cell,observed,39,ENSG00000151490
+B cell,observed,40,ENSG00000186814
+B cell,observed,41,ENSG00000197548
+B cell,observed,42,ENSG00000140750
+B cell,observed,43,ENSG00000197385
+B cell,observed,44,ENSG00000035403
+B cell,observed,45,ENSG00000115956
+B cell,observed,46,ENSG00000100242
+B cell,observed,47,ENSG00000071575
+B cell,observed,48,ENSG00000071073
+B cell,observed,49,ENSG00000140548
+B cell,observed,50,ENSG00000058091
+Monocyte,observed,1,ENSG00000115590
+Monocyte,observed,2,ENSG00000087116
+Monocyte,observed,3,ENSG00000100079
+Monocyte,observed,4,ENSG00000155659
+Monocyte,observed,5,ENSG00000166428
+Monocyte,observed,6,ENSG00000140545
+Monocyte,observed,7,ENSG00000113088
+Monocyte,observed,8,ENSG00000173369
+Monocyte,observed,9,ENSG00000138821
+Monocyte,observed,10,ENSG00000081059
+Monocyte,observed,11,ENSG00000138795
+Monocyte,observed,12,ENSG00000168685
+Monocyte,observed,13,ENSG00000078053
+Monocyte,observed,14,ENSG00000146122
+Monocyte,observed,15,ENSG00000179344
+Monocyte,observed,16,ENSG00000184613
+Monocyte,observed,17,ENSG00000186810
+Monocyte,observed,18,ENSG00000100385
+Monocyte,observed,19,ENSG00000174705
+Monocyte,observed,20,ENSG00000152495
+Monocyte,observed,21,ENSG00000109321
+Monocyte,observed,22,ENSG00000013725
+Monocyte,observed,23,ENSG00000172575
+Monocyte,observed,24,ENSG00000153823
+Monocyte,observed,25,ENSG00000172543
+Monocyte,observed,26,ENSG00000147138
+Monocyte,observed,27,ENSG00000110448
+Monocyte,observed,28,ENSG00000115085
+Monocyte,observed,29,ENSG00000160856
+Monocyte,observed,30,ENSG00000173114
+Monocyte,observed,31,ENSG00000168209
+Monocyte,observed,32,ENSG00000111796
+Monocyte,observed,33,ENSG00000008517
+Monocyte,observed,34,ENSG00000167286
+Monocyte,observed,35,ENSG00000198851
+Monocyte,observed,36,ENSG00000102245
+Monocyte,observed,37,ENSG00000143184
+Monocyte,observed,38,ENSG00000113263
+Monocyte,observed,39,ENSG00000073861
+Monocyte,observed,40,ENSG00000172673
+Monocyte,observed,41,ENSG00000173372
+Monocyte,observed,42,ENSG00000183918
+Monocyte,observed,43,ENSG00000198821
+Monocyte,observed,44,ENSG00000027869
+Monocyte,observed,45,ENSG00000178562
+Monocyte,observed,46,ENSG00000065675
+Monocyte,observed,47,ENSG00000124181
+Monocyte,observed,48,ENSG00000106178
+Monocyte,observed,49,ENSG00000141384
+Monocyte,observed,50,ENSG00000196935
+NK cell,observed,1,ENSG00000185215
+NK cell,observed,2,ENSG00000100079
+NK cell,observed,3,ENSG00000106066
+NK cell,observed,4,ENSG00000110077
+NK cell,observed,5,ENSG00000171916
+NK cell,observed,6,ENSG00000170298
+NK cell,observed,7,ENSG00000124507
+NK cell,observed,8,ENSG00000010610
+NK cell,observed,9,ENSG00000121552
+NK cell,observed,10,ENSG00000196159
+NK cell,observed,11,ENSG00000197629
+NK cell,observed,12,ENSG00000172243
+NK cell,observed,13,ENSG00000120708
+NK cell,observed,14,ENSG00000126759
+NK cell,observed,15,ENSG00000172322
+NK cell,observed,16,ENSG00000140749
+NK cell,observed,17,ENSG00000120594
+NK cell,observed,18,ENSG00000137393
+NK cell,observed,19,ENSG00000184731
+NK cell,observed,20,ENSG00000197467
+NK cell,observed,21,ENSG00000121807
+NK cell,observed,22,ENSG00000169744
+NK cell,observed,23,ENSG00000104093
+NK cell,observed,24,ENSG00000241399
+NK cell,observed,25,ENSG00000151474
+NK cell,observed,26,ENSG00000087253
+NK cell,observed,27,ENSG00000102349
+NK cell,observed,28,ENSG00000100154
+NK cell,observed,29,ENSG00000101842
+NK cell,observed,30,ENSG00000140543
+NK cell,observed,31,ENSG00000261371
+NK cell,observed,32,ENSG00000198223
+NK cell,observed,33,ENSG00000136205
+NK cell,observed,34,ENSG00000138119
+NK cell,observed,35,ENSG00000101542
+NK cell,observed,36,ENSG00000095637
+NK cell,observed,37,ENSG00000168329
+NK cell,observed,38,ENSG00000219481
+NK cell,observed,39,ENSG00000035403
+NK cell,observed,40,ENSG00000161638
+NK cell,observed,41,ENSG00000147576
+NK cell,observed,42,ENSG00000135426
+NK cell,observed,43,ENSG00000170989
+NK cell,observed,44,ENSG00000123636
+NK cell,observed,45,ENSG00000185697
+NK cell,observed,46,ENSG00000100385
+NK cell,observed,47,ENSG00000160991
+NK cell,observed,48,ENSG00000172575
+NK cell,observed,49,ENSG00000006530
+NK cell,observed,50,ENSG00000185989
+DC,observed,1,ENSG00000179639
+DC,observed,2,ENSG00000153823
+DC,observed,3,ENSG00000163606
+DC,observed,4,ENSG00000198561
+DC,observed,5,ENSG00000081059
+DC,observed,6,ENSG00000118922
+DC,observed,7,ENSG00000150054
+DC,observed,8,ENSG00000106034
+DC,observed,9,ENSG00000112182
+DC,observed,10,ENSG00000154229
+DC,observed,11,ENSG00000074964
+DC,observed,12,ENSG00000123607
+DC,observed,13,ENSG00000165983
+DC,observed,14,ENSG00000013725
+DC,observed,15,ENSG00000136111
+DC,observed,16,ENSG00000171408
+DC,observed,17,ENSG00000104361
+DC,observed,18,ENSG00000134852
+DC,observed,19,ENSG00000100105
+DC,observed,20,ENSG00000120162
+DC,observed,21,ENSG00000150636
+DC,observed,22,ENSG00000186260
+DC,observed,23,ENSG00000111846
+DC,observed,24,ENSG00000133059
+DC,observed,25,ENSG00000104228
+DC,observed,26,ENSG00000186854
+DC,observed,27,ENSG00000114982
+DC,observed,28,ENSG00000149292
+DC,observed,29,ENSG00000171843
+DC,observed,30,ENSG00000135362
+DC,observed,31,ENSG00000109099
+DC,observed,32,ENSG00000100890
+DC,observed,33,ENSG00000187912
+DC,observed,34,ENSG00000130023
+DC,observed,35,ENSG00000116852
+DC,observed,36,ENSG00000012174
+DC,observed,37,ENSG00000269713
+DC,observed,38,ENSG00000113761
+DC,observed,39,ENSG00000137942
+DC,observed,40,ENSG00000171617
+DC,observed,41,ENSG00000135040
+DC,observed,42,ENSG00000066651
+DC,observed,43,ENSG00000120262
+DC,observed,44,ENSG00000146373
+DC,observed,45,ENSG00000115594
+DC,observed,46,ENSG00000151422
+DC,observed,47,ENSG00000151014
+DC,observed,48,ENSG00000141447
+DC,observed,49,ENSG00000115998
+DC,observed,50,ENSG00000152315
+DC,generated,1,ENSG00000198755

paper-releases/pbisc-diffusion/updates/20261008T022410Z-deconv-scsemi-deck/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores/ye2026_sepsis/scsemi_cv_20261002/fold3/README.md

@@ -0,0 +1,73 @@
+# score_ft.py output: ye2026_sepsis, 16 scored donors
+
+Arms: scsemi__raw.  Every arm is scored on the same donors against the same observed reference.
+
+## Inputs and fixed contracts
+- Observed reference: `/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ye2026_sepsis/reference.npz` (sha bac12da8a241), built from datadir:/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/ye2026_sepsis;
+  observed cells sum exactly to the features pseudobulk of every donor. Observed labels: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ye2026_sepsis/observed_labels/labels/labels.parquet.
+- Generated cells: `generated_counts` read out once with Poisson(seed 20260916) over the **whole NPZ in row order**
+  (`poisson_readout.poisson_integerise`), then the scored donors' rows. These integer counts are used for labels, embedding,
+  depth and the contrast. Pseudobulk uses the summed `generated_counts` (rates), as `score_generated.py`.
+- Labels: frozen f01common scANVI (`predict_scanvi.py`, CPU, zero query epochs), six-type gate ['T cell', 'B cell', 'Monocyte', 'NK cell', 'DC', 'Other']; query genes =
+  the 1,006-gene annotation list minus genes the dataset's sc assay did not measure in every sample (1006 genes).
+- Scoring genes: 6k panel genes measured by both assays in every sample of the dataset (5975 genes).
+
+## Definitions
+1. **Composition**: fraction of each donor's cells per six-type label (all generated cells; all observed QC cells).
+   `comp_l1` = sum over the six labels of |generated - observed| (immicro_common.l1). `comp_l1_obs_loo_null` = L1 of the
+   donor's observed composition to the mean of the *other scored donors'* observed compositions. `composition_top1` =
+   fraction of donors whose generated composition is closest (L1) to their own observed composition.
+2. **Pseudobulk**: `pb_r` = Pearson r of CPM-log1p(summed generated_counts) vs CPM-log1p(observed pseudobulk) on the scoring
+   genes (evaluate_realbulk.cpm_log1p / paired_r). `pb_top1` = the donor's generated pseudobulk is closest (Pearson) to its own
+   observed pseudobulk among the 16 scored donors (immicro_common.retrieval = evaluate_realbulk.top1 arithmetic).
+3. **Precision / coverage / density** (stage_b_embed.py functions): per-cell CP10K over the scoring genes + log1p; PCA-50
+   (randomized, random_state 20260925) and k = 10 real-real k-NN balls fitted on **R1 of the scored donors only**.
+   Reference rule (independent of fold and arm): per donor, cells in canonical order (library_id+barcode), one
+   permutation from `default_rng([20260925, uint64(sha256(donor)[:8])])`; R1 = first n, R2 = next n,
+   n = min(150, n_cells // 2) (2400 R1 cells here). Generated query: per donor n cells (the donor's R1 size),
+   without replacement, `default_rng([20260926, same key])`. Precision = fraction of generated cells inside any R1 ball
+   (Kynkaanniemi 2019); coverage = fraction of R1 cells whose ball holds a generated cell; density = Naeem 2020.
+   Per-donor precision = that donor's generated cells; per-donor coverage = that donor's R1 cells (the pooled values are
+   their n-weighted means; plain means when every donor has 150 reference cells). R2 (held-out real cells of the same donors) is the
+   perfect-generator control: precision 0.9967, coverage 1.0000.
+   **Depth matched**: stage_b_embed depth matching run pairwise (R1 vs this arm): per donor and within-donor rank of detected
+   genes, both cells are thinned by UMI sampling without replacement to the smaller detected-gene count; PCA kept from
+   un-thinned R1, radii recomputed on thinned R1; fresh rng(20260929) per arm. R2 control: precision
+   0.9971, coverage 1.0000.
+4. **Monocyte %** = 100 x Monocyte fraction (all cells) per donor; **median UMI per cell** = median over cells of the
+   Poisson-read (generated) / observed counts summed over the scoring genes (per donor, and pooled).
+5. **Group contrast** (T, B, Monocyte, NK): case groups ['adult_sepsis', 'ped_sepsis'] (n = 13) vs control groups ['adult_HC', 'adult_ICU_ctrl', 'ped_HC', 'ped_ICU_ctrl'] (n = 3); rule: explicit --case-groups / --control-groups; groups from features groups.
+   Genes per type = scoring genes detected (count > 0) in >= 1 % of the scored donors' observed
+   cells of that type. Primary statistic `mean_cp10k`: per-cell CP10K over the scoring genes; per donor the mean over its
+   cells of the type (donors with >= 1 such cell); group mean = mean over the group's donors (donor-balanced);
+   logFC = log2(case mean + 1) - log2(control mean + 1). Secondary `pseudobulk_log1p_cpm`: the SLE benchmark evaluator's
+   statistic (donor x type summed counts, log1p CPM over the scoring genes, mean case - mean control;
+   evaluate_generated_vs_truth.group_log_fold_change). Both are computed on the generated cells (their frozen labels) and on
+   the observed cells; `r_generated` = Pearson r of the two logFC vectors. Broadcast baseline `r_broadcast`: the same
+   observed vector against the logFC of the input bulk (condition NPZ `bulk_counts` of the arm, scoring genes, per-donor
+   CP10K resp. log1p CPM, group means). `partial_r_generated_given_broadcast`: r of the OLS residuals of observed and
+   generated logFC on the bulk logFC (residual_specificity.py).
+6. **Top-50 DEG overlap** (contrast.csv rows `statistic = welch_top50_deg`; summary `deg_top50`): the SLE benchmark's score (`total_top50_overlap` of
+   `bulk2sc-paper/scripts/evaluate_generated_vs_truth.py`, Welch backend: the "top-50 DEG 81 / 84" of the final model), run
+   with the evaluator's own functions on the same donor x type pseudobulks as the contrast. Per cell type, observed and
+   generated separately: donor x type summed counts on all 5975 scoring genes (donors of the two groups with
+   >= 1 cell of the type) -> `welch_bh_deg` (log1p CPM, two-sided Welch t-test case vs control, Benjamini-Hochberg over the
+   genes with a defined p-value; needs >= 2 donors per group) -> `top_deg_genes` (FDR < 0.05, descending |logFC|,
+   first 50; shorter when fewer genes pass). `top50_overlap` = |observed list & generated list|; `top50_jaccard`
+   the evaluator's Jaccard. Summary `sum` = total over the scored types (the quantity reported as 81), `max` = the sum of
+   the observed list lengths (the "of 262"), `sum_4type` = T + B + Monocyte + NK. A type is scored when its observed test
+   is defined; an arm whose generated test is undefined for a scored type gets overlap 0. Types: T cell, B cell, Monocyte, NK cell, DC
+   (aggregates: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ye2026_sepsis/deg_reference.npz).
+   **Broadcast analogue** (`top50_overlap_broadcast`, as `sle_broadcast.py` `broadcast_input_bulk`; not part of the
+   published benchmark): the same test on the arm's condition bulk_counts (scoring genes, the 13 + 3
+   donors), one whole-PBMC top-50 list, compared with every type's observed list.
+
+## Files
+- `per_donor.csv`: one row per arm x donor (all per-donor quantities above).
+- `summary.json`: per arm the donor means / pooled values, provenance (NPZ sha, receipt check, generation metadata,
+  Poisson nnz of the whole NPZ, annotation contract), and the reference block (PCA, R2 control).
+- `contrast.csv`: per arm x cell type x statistic (`mean_cp10k`, `pseudobulk_log1p_cpm`: logFC r; `welch_top50_deg`:
+  top-50 DEG overlap, T / B / Monocyte / NK / DC; columns of the other kind are empty).
+- `arms/<arm>/labels/` (predict_scanvi.py labels + manifest), `arms/<arm>/generated_cells.parquet` (NPZ row, donor, label,
+  UMI on the scoring genes, P/C subsample flag), `arms/<arm>/deg_top50_lists.csv` (observed, generated and broadcast
+  top-50 gene lists, Ensembl ids, rank order).

paper-releases/pbisc-diffusion/updates/20261008T022410Z-deconv-scsemi-deck/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores/ye2026_sepsis/scsemi_cv_20261002/fold4/README.md

@@ -0,0 +1,73 @@
+# score_ft.py output: ye2026_sepsis, 16 scored donors
+
+Arms: scsemi__raw.  Every arm is scored on the same donors against the same observed reference.
+
+## Inputs and fixed contracts
+- Observed reference: `/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ye2026_sepsis/reference.npz` (sha bac12da8a241), built from datadir:/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/ye2026_sepsis;
+  observed cells sum exactly to the features pseudobulk of every donor. Observed labels: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ye2026_sepsis/observed_labels/labels/labels.parquet.
+- Generated cells: `generated_counts` read out once with Poisson(seed 20260916) over the **whole NPZ in row order**
+  (`poisson_readout.poisson_integerise`), then the scored donors' rows. These integer counts are used for labels, embedding,
+  depth and the contrast. Pseudobulk uses the summed `generated_counts` (rates), as `score_generated.py`.
+- Labels: frozen f01common scANVI (`predict_scanvi.py`, CPU, zero query epochs), six-type gate ['T cell', 'B cell', 'Monocyte', 'NK cell', 'DC', 'Other']; query genes =
+  the 1,006-gene annotation list minus genes the dataset's sc assay did not measure in every sample (1006 genes).
+- Scoring genes: 6k panel genes measured by both assays in every sample of the dataset (5975 genes).
+
+## Definitions
+1. **Composition**: fraction of each donor's cells per six-type label (all generated cells; all observed QC cells).
+   `comp_l1` = sum over the six labels of |generated - observed| (immicro_common.l1). `comp_l1_obs_loo_null` = L1 of the
+   donor's observed composition to the mean of the *other scored donors'* observed compositions. `composition_top1` =
+   fraction of donors whose generated composition is closest (L1) to their own observed composition.
+2. **Pseudobulk**: `pb_r` = Pearson r of CPM-log1p(summed generated_counts) vs CPM-log1p(observed pseudobulk) on the scoring
+   genes (evaluate_realbulk.cpm_log1p / paired_r). `pb_top1` = the donor's generated pseudobulk is closest (Pearson) to its own
+   observed pseudobulk among the 16 scored donors (immicro_common.retrieval = evaluate_realbulk.top1 arithmetic).
+3. **Precision / coverage / density** (stage_b_embed.py functions): per-cell CP10K over the scoring genes + log1p; PCA-50
+   (randomized, random_state 20260925) and k = 10 real-real k-NN balls fitted on **R1 of the scored donors only**.
+   Reference rule (independent of fold and arm): per donor, cells in canonical order (library_id+barcode), one
+   permutation from `default_rng([20260925, uint64(sha256(donor)[:8])])`; R1 = first n, R2 = next n,
+   n = min(150, n_cells // 2) (2400 R1 cells here). Generated query: per donor n cells (the donor's R1 size),
+   without replacement, `default_rng([20260926, same key])`. Precision = fraction of generated cells inside any R1 ball
+   (Kynkaanniemi 2019); coverage = fraction of R1 cells whose ball holds a generated cell; density = Naeem 2020.
+   Per-donor precision = that donor's generated cells; per-donor coverage = that donor's R1 cells (the pooled values are
+   their n-weighted means; plain means when every donor has 150 reference cells). R2 (held-out real cells of the same donors) is the
+   perfect-generator control: precision 0.9975, coverage 1.0000.
+   **Depth matched**: stage_b_embed depth matching run pairwise (R1 vs this arm): per donor and within-donor rank of detected
+   genes, both cells are thinned by UMI sampling without replacement to the smaller detected-gene count; PCA kept from
+   un-thinned R1, radii recomputed on thinned R1; fresh rng(20260929) per arm. R2 control: precision
+   0.9958, coverage 1.0000.
+4. **Monocyte %** = 100 x Monocyte fraction (all cells) per donor; **median UMI per cell** = median over cells of the
+   Poisson-read (generated) / observed counts summed over the scoring genes (per donor, and pooled).
+5. **Group contrast** (T, B, Monocyte, NK): case groups ['adult_sepsis', 'ped_sepsis'] (n = 12) vs control groups ['adult_HC', 'adult_ICU_ctrl', 'ped_HC', 'ped_ICU_ctrl'] (n = 4); rule: explicit --case-groups / --control-groups; groups from features groups.
+   Genes per type = scoring genes detected (count > 0) in >= 1 % of the scored donors' observed
+   cells of that type. Primary statistic `mean_cp10k`: per-cell CP10K over the scoring genes; per donor the mean over its
+   cells of the type (donors with >= 1 such cell); group mean = mean over the group's donors (donor-balanced);
+   logFC = log2(case mean + 1) - log2(control mean + 1). Secondary `pseudobulk_log1p_cpm`: the SLE benchmark evaluator's
+   statistic (donor x type summed counts, log1p CPM over the scoring genes, mean case - mean control;
+   evaluate_generated_vs_truth.group_log_fold_change). Both are computed on the generated cells (their frozen labels) and on
+   the observed cells; `r_generated` = Pearson r of the two logFC vectors. Broadcast baseline `r_broadcast`: the same
+   observed vector against the logFC of the input bulk (condition NPZ `bulk_counts` of the arm, scoring genes, per-donor
+   CP10K resp. log1p CPM, group means). `partial_r_generated_given_broadcast`: r of the OLS residuals of observed and
+   generated logFC on the bulk logFC (residual_specificity.py).
+6. **Top-50 DEG overlap** (contrast.csv rows `statistic = welch_top50_deg`; summary `deg_top50`): the SLE benchmark's score (`total_top50_overlap` of
+   `bulk2sc-paper/scripts/evaluate_generated_vs_truth.py`, Welch backend: the "top-50 DEG 81 / 84" of the final model), run
+   with the evaluator's own functions on the same donor x type pseudobulks as the contrast. Per cell type, observed and
+   generated separately: donor x type summed counts on all 5975 scoring genes (donors of the two groups with
+   >= 1 cell of the type) -> `welch_bh_deg` (log1p CPM, two-sided Welch t-test case vs control, Benjamini-Hochberg over the
+   genes with a defined p-value; needs >= 2 donors per group) -> `top_deg_genes` (FDR < 0.05, descending |logFC|,
+   first 50; shorter when fewer genes pass). `top50_overlap` = |observed list & generated list|; `top50_jaccard`
+   the evaluator's Jaccard. Summary `sum` = total over the scored types (the quantity reported as 81), `max` = the sum of
+   the observed list lengths (the "of 262"), `sum_4type` = T + B + Monocyte + NK. A type is scored when its observed test
+   is defined; an arm whose generated test is undefined for a scored type gets overlap 0. Types: T cell, B cell, Monocyte, NK cell, DC
+   (aggregates: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ye2026_sepsis/deg_reference.npz).
+   **Broadcast analogue** (`top50_overlap_broadcast`, as `sle_broadcast.py` `broadcast_input_bulk`; not part of the
+   published benchmark): the same test on the arm's condition bulk_counts (scoring genes, the 12 + 4
+   donors), one whole-PBMC top-50 list, compared with every type's observed list.
+
+## Files
+- `per_donor.csv`: one row per arm x donor (all per-donor quantities above).
+- `summary.json`: per arm the donor means / pooled values, provenance (NPZ sha, receipt check, generation metadata,
+  Poisson nnz of the whole NPZ, annotation contract), and the reference block (PCA, R2 control).
+- `contrast.csv`: per arm x cell type x statistic (`mean_cp10k`, `pseudobulk_log1p_cpm`: logFC r; `welch_top50_deg`:
+  top-50 DEG overlap, T / B / Monocyte / NK / DC; columns of the other kind are empty).
+- `arms/<arm>/labels/` (predict_scanvi.py labels + manifest), `arms/<arm>/generated_cells.parquet` (NPZ row, donor, label,
+  UMI on the scoring genes, P/C subsample flag), `arms/<arm>/deg_top50_lists.csv` (observed, generated and broadcast
+  top-50 gene lists, Ensembl ids, rank order).

paper-releases/pbisc-diffusion/updates/20261008T022410Z-deconv-scsemi-deck/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores/ye2026_sepsis/scsemi_cv_20261002/pooled/README.md

@@ -0,0 +1,73 @@
+# score_ft.py output: ye2026_sepsis, 82 scored donors
+
+Arms: scsemi__raw.  Every arm is scored on the same donors against the same observed reference.
+
+## Inputs and fixed contracts
+- Observed reference: `/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ye2026_sepsis/reference.npz` (sha bac12da8a241), built from datadir:/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/ye2026_sepsis;
+  observed cells sum exactly to the features pseudobulk of every donor. Observed labels: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ye2026_sepsis/observed_labels/labels/labels.parquet.
+- Generated cells: `generated_counts` read out once with Poisson(seed 20260916) over the **whole NPZ in row order**
+  (`poisson_readout.poisson_integerise`), then the scored donors' rows. These integer counts are used for labels, embedding,
+  depth and the contrast. Pseudobulk uses the summed `generated_counts` (rates), as `score_generated.py`.
+- Labels: frozen f01common scANVI (`predict_scanvi.py`, CPU, zero query epochs), six-type gate ['T cell', 'B cell', 'Monocyte', 'NK cell', 'DC', 'Other']; query genes =
+  the 1,006-gene annotation list minus genes the dataset's sc assay did not measure in every sample (1006 genes).
+- Scoring genes: 6k panel genes measured by both assays in every sample of the dataset (5975 genes).
+
+## Definitions
+1. **Composition**: fraction of each donor's cells per six-type label (all generated cells; all observed QC cells).
+   `comp_l1` = sum over the six labels of |generated - observed| (immicro_common.l1). `comp_l1_obs_loo_null` = L1 of the
+   donor's observed composition to the mean of the *other scored donors'* observed compositions. `composition_top1` =
+   fraction of donors whose generated composition is closest (L1) to their own observed composition.
+2. **Pseudobulk**: `pb_r` = Pearson r of CPM-log1p(summed generated_counts) vs CPM-log1p(observed pseudobulk) on the scoring
+   genes (evaluate_realbulk.cpm_log1p / paired_r). `pb_top1` = the donor's generated pseudobulk is closest (Pearson) to its own
+   observed pseudobulk among the 82 scored donors (immicro_common.retrieval = evaluate_realbulk.top1 arithmetic).
+3. **Precision / coverage / density** (stage_b_embed.py functions): per-cell CP10K over the scoring genes + log1p; PCA-50
+   (randomized, random_state 20260925) and k = 10 real-real k-NN balls fitted on **R1 of the scored donors only**.
+   Reference rule (independent of fold and arm): per donor, cells in canonical order (library_id+barcode), one
+   permutation from `default_rng([20260925, uint64(sha256(donor)[:8])])`; R1 = first n, R2 = next n,
+   n = min(150, n_cells // 2) (12300 R1 cells here). Generated query: per donor n cells (the donor's R1 size),
+   without replacement, `default_rng([20260926, same key])`. Precision = fraction of generated cells inside any R1 ball
+   (Kynkaanniemi 2019); coverage = fraction of R1 cells whose ball holds a generated cell; density = Naeem 2020.
+   Per-donor precision = that donor's generated cells; per-donor coverage = that donor's R1 cells (the pooled values are
+   their n-weighted means; plain means when every donor has 150 reference cells). R2 (held-out real cells of the same donors) is the
+   perfect-generator control: precision 0.9805, coverage 0.9997.
+   **Depth matched**: stage_b_embed depth matching run pairwise (R1 vs this arm): per donor and within-donor rank of detected
+   genes, both cells are thinned by UMI sampling without replacement to the smaller detected-gene count; PCA kept from
+   un-thinned R1, radii recomputed on thinned R1; fresh rng(20260929) per arm. R2 control: precision
+   0.9796, coverage 0.9998.
+4. **Monocyte %** = 100 x Monocyte fraction (all cells) per donor; **median UMI per cell** = median over cells of the
+   Poisson-read (generated) / observed counts summed over the scoring genes (per donor, and pooled).
+5. **Group contrast** (T, B, Monocyte, NK): case groups ['adult_sepsis', 'ped_sepsis'] (n = 61) vs control groups ['adult_HC', 'adult_ICU_ctrl', 'ped_HC', 'ped_ICU_ctrl'] (n = 21); rule: explicit --case-groups / --control-groups; groups from features groups.
+   Genes per type = scoring genes detected (count > 0) in >= 1 % of the scored donors' observed
+   cells of that type. Primary statistic `mean_cp10k`: per-cell CP10K over the scoring genes; per donor the mean over its
+   cells of the type (donors with >= 1 such cell); group mean = mean over the group's donors (donor-balanced);
+   logFC = log2(case mean + 1) - log2(control mean + 1). Secondary `pseudobulk_log1p_cpm`: the SLE benchmark evaluator's
+   statistic (donor x type summed counts, log1p CPM over the scoring genes, mean case - mean control;
+   evaluate_generated_vs_truth.group_log_fold_change). Both are computed on the generated cells (their frozen labels) and on
+   the observed cells; `r_generated` = Pearson r of the two logFC vectors. Broadcast baseline `r_broadcast`: the same
+   observed vector against the logFC of the input bulk (condition NPZ `bulk_counts` of the arm, scoring genes, per-donor
+   CP10K resp. log1p CPM, group means). `partial_r_generated_given_broadcast`: r of the OLS residuals of observed and
+   generated logFC on the bulk logFC (residual_specificity.py).
+6. **Top-50 DEG overlap** (contrast.csv rows `statistic = welch_top50_deg`; summary `deg_top50`): the SLE benchmark's score (`total_top50_overlap` of
+   `bulk2sc-paper/scripts/evaluate_generated_vs_truth.py`, Welch backend: the "top-50 DEG 81 / 84" of the final model), run
+   with the evaluator's own functions on the same donor x type pseudobulks as the contrast. Per cell type, observed and
+   generated separately: donor x type summed counts on all 5975 scoring genes (donors of the two groups with
+   >= 1 cell of the type) -> `welch_bh_deg` (log1p CPM, two-sided Welch t-test case vs control, Benjamini-Hochberg over the
+   genes with a defined p-value; needs >= 2 donors per group) -> `top_deg_genes` (FDR < 0.05, descending |logFC|,
+   first 50; shorter when fewer genes pass). `top50_overlap` = |observed list & generated list|; `top50_jaccard`
+   the evaluator's Jaccard. Summary `sum` = total over the scored types (the quantity reported as 81), `max` = the sum of
+   the observed list lengths (the "of 262"), `sum_4type` = T + B + Monocyte + NK. A type is scored when its observed test
+   is defined; an arm whose generated test is undefined for a scored type gets overlap 0. Types: T cell, B cell, Monocyte, NK cell, DC
+   (aggregates: /home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ye2026_sepsis/deg_reference.npz).
+   **Broadcast analogue** (`top50_overlap_broadcast`, as `sle_broadcast.py` `broadcast_input_bulk`; not part of the
+   published benchmark): the same test on the arm's condition bulk_counts (scoring genes, the 61 + 21
+   donors), one whole-PBMC top-50 list, compared with every type's observed list.
+
+## Files
+- `per_donor.csv`: one row per arm x donor (all per-donor quantities above).
+- `summary.json`: per arm the donor means / pooled values, provenance (NPZ sha, receipt check, generation metadata,
+  Poisson nnz of the whole NPZ, annotation contract), and the reference block (PCA, R2 control).
+- `contrast.csv`: per arm x cell type x statistic (`mean_cp10k`, `pseudobulk_log1p_cpm`: logFC r; `welch_top50_deg`:
+  top-50 DEG overlap, T / B / Monocyte / NK / DC; columns of the other kind are empty).
+- `arms/<arm>/labels/` (predict_scanvi.py labels + manifest), `arms/<arm>/generated_cells.parquet` (NPZ row, donor, label,
+  UMI on the scoring genes, P/C subsample flag), `arms/<arm>/deg_top50_lists.csv` (observed, generated and broadcast
+  top-50 gene lists, Ensembl ids, rank order).