Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/scores/giroux2022/locoL_lora64sat_s60000_K1000/summary.json

@@ -0,0 +1,531 @@
+{
+ "schema": "paired_ft_scores.v1",
+ "dataset": "giroux2022",
+ "written_utc": "2026-10-02T21:07:07.035901+00:00",
+ "command": [
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/tools/score_ft.py",
+  "--dataset",
+  "giroux2022",
+  "--threads",
+  "8",
+  "--donors",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/data/giroux2022/usable_samples.txt",
+  "--arm",
+  "locoL_lora64sat_s60000_K1000=/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/generated/giroux2022/hoL_giroux2022/K1000_lora64sat_s60000.npz",
+  "--case-groups",
+  "severe",
+  "--control-groups",
+  "mild",
+  "--out",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/scores/giroux2022/locoL_lora64sat_s60000_K1000"
+ ],
+ "arms": [
+  "locoL_lora64sat_s60000_K1000"
+ ],
+ "donors": [
+  "450905_d0",
+  "450905_d14",
+  "0B943B_d0",
+  "0B943B_d14",
+  "7085CA_d0",
+  "7085CA_d7",
+  "7085CA_d14",
+  "2AD75E_d0",
+  "2AD75E_d7",
+  "2AD75E_d14",
+  "B85D75_d0",
+  "B85D75_d7",
+  "180E1A_d0",
+  "180E1A_d7",
+  "180E1A_d14",
+  "82CCF5_d0",
+  "82CCF5_d7",
+  "82CCF5_d14",
+  "BAAF62_d0",
+  "BAAF62_d7",
+  "BAAF62_d14",
+  "0BF51C_d0",
+  "0BF51C_d7",
+  "0BF51C_d14"
+ ],
+ "n_donors": 24,
+ "reference": {
+  "cache": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/giroux2022",
+  "reference_sha256": "dd0152f7bde41da80e1c71dabf54ee561982f36dc99a336b65fca419880a1593",
+  "features_sha256": "fb41b91617e3ff786d5153c4dc7869d66191ca5947ff28161d24215a24f819e5",
+  "n_r1_cells": 3600,
+  "pca": {
+   "n_components": 50,
+   "explained_variance_ratio_sum": 0.23158283531665802,
+   "n_fit_cells": 3600
+  },
+  "r2_control": {
+   "precision": 0.9980555555555556,
+   "coverage": 1.0,
+   "density": 3.7886666666666664,
+   "n_query": 3600,
+   "depth_matched": {
+    "precision": 0.9980555555555556,
+    "coverage": 1.0,
+    "density": 3.7033611111111115
+   }
+  },
+  "observed_l1_loo_null_mean": 0.30584824369245694
+ },
+ "contrast_definition": {
+  "enabled": true,
+  "rule": "explicit --case-groups / --control-groups",
+  "group_source": "features groups",
+  "case_groups": [
+   "severe"
+  ],
+  "control_groups": [
+   "mild"
+  ],
+  "n_case_donors": 11,
+  "n_control_donors": 13,
+  "n_donors_in_neither": 0,
+  "case": [
+   false,
+   false,
+   false,
+   false,
+   false,
+   false,
+   false,
+   false,
+   false,
+   false,
+   true,
+   true,
+   true,
+   true,
+   true,
+   true,
+   true,
+   true,
+   false,
+   false,
+   false,
+   true,
+   true,
+   true
+  ],
+  "ctrl": [
+   true,
+   true,
+   true,
+   true,
+   true,
+   true,
+   true,
+   true,
+   true,
+   true,
+   false,
+   false,
+   false,
+   false,
+   false,
+   false,
+   false,
+   false,
+   true,
+   true,
+   true,
+   false,
+   false,
+   false
+  ],
+  "groups": [
+   "mild",
+   "mild",
+   "mild",
+   "mild",
+   "mild",
+   "mild",
+   "mild",
+   "mild",
+   "mild",
+   "mild",
+   "severe",
+   "severe",
+   "severe",
+   "severe",
+   "severe",
+   "severe",
+   "severe",
+   "severe",
+   "mild",
+   "mild",
+   "mild",
+   "severe",
+   "severe",
+   "severe"
+  ]
+ },
+ "deg_definition": {
+  "enabled": true,
+  "score": "top-50 DEG overlap per cell type (benchmark total_top50_overlap, Welch backend)",
+  "source": "/home/sj_server_1/PBISC/immun-pbmc-fm/bulk2sc-paper/scripts/evaluate_generated_vs_truth.py",
+  "functions": [
+   "welch_bh_deg",
+   "top_deg_genes",
+   "jaccard"
+  ],
+  "reproduces": "utility_studies_final6k_20260925/sle top-50 DEG 81 (1,006 genes) / 84 (2,279 genes) = sum over the six gated types of top50_overlap_count",
+  "top_n": 50,
+  "fdr": 0.05,
+  "n_genes": 5563,
+  "types": [
+   "T cell",
+   "B cell",
+   "Monocyte",
+   "NK cell",
+   "DC"
+  ],
+  "aggregates_source": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/giroux2022/deg_reference.npz",
+  "aggregates_sha256": "412c44bba7481b0b851c3b12e695f1b7ff662163e1f7849d3a64359dfce6d50e",
+  "aggregates_note": null,
+  "types_not_scored_observed": [],
+  "observed": {
+   "T cell": {
+    "n_significant": 306,
+    "n_tested": 5354,
+    "n_case_donors": 11,
+    "n_control_donors": 13,
+    "n_cells": 145034,
+    "top_list_size": 50
+   },
+   "B cell": {
+    "n_significant": 9,
+    "n_tested": 5142,
+    "n_case_donors": 11,
+    "n_control_donors": 13,
+    "n_cells": 17383,
+    "top_list_size": 9
+   },
+   "Monocyte": {
+    "n_significant": 7,
+    "n_tested": 5295,
+    "n_case_donors": 11,
+    "n_control_donors": 13,
+    "n_cells": 47311,
+    "top_list_size": 7
+   },
+   "NK cell": {
+    "n_significant": 83,
+    "n_tested": 5159,
+    "n_case_donors": 11,
+    "n_control_donors": 13,
+    "n_cells": 29061,
+    "top_list_size": 50
+   },
+   "DC": {
+    "n_significant": 7,
+    "n_tested": 4962,
+    "n_case_donors": 11,
+    "n_control_donors": 13,
+    "n_cells": 4328,
+    "top_list_size": 7
+   }
+  },
+  "broadcast": "welch_bh_deg + top_deg_genes on the arm's condition bulk_counts (scoring genes, case|control donors), one list compared with every type's observed list (sle_broadcast.py broadcast_input_bulk); not part of the published benchmark"
+ },
+ "headline": {
+  "locoL_lora64sat_s60000_K1000": {
+   "comp_l1": 0.42970541878518415,
+   "comp_l1_obs_loo_null": 0.30584824369245694,
+   "precision": 0.9936111111111111,
+   "coverage": 0.8152777777777778,
+   "precision_dm": 0.975,
+   "coverage_dm": 0.8447222222222223,
+   "pb_r_median": 0.8469462885794707,
+   "pb_top1": 0.16666666666666666,
+   "mono_pct_gen": 5.277777777777778,
+   "mono_pct_obs": 19.247866360995445,
+   "median_umi_gen": 501.0,
+   "median_umi_obs": 471.0,
+   "contrast_mean_cp10k_r_gen": 0.3029345793102561,
+   "contrast_pseudobulk_log1p_cpm_r_gen": 0.29264113005668807,
+   "contrast_mean_cp10k_r_broadcast": 0.6381052168564143,
+   "contrast_pseudobulk_log1p_cpm_r_broadcast": 0.794738791087971,
+   "deg_top50_T": 5,
+   "deg_top50_B": 4,
+   "deg_top50_Mono": 2,
+   "deg_top50_NK": 3,
+   "deg_top50_DC": 1,
+   "deg_top50_sum": 15,
+   "deg_top50_max": 123,
+   "deg_top50_sum_4type": 14,
+   "deg_top50_sum_broadcast": 39,
+   "seconds": 9.0
+  }
+ },
+ "per_arm": {
+  "locoL_lora64sat_s60000_K1000": {
+   "arm": "locoL_lora64sat_s60000_K1000",
+   "npz": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/generated/giroux2022/hoL_giroux2022/K1000_lora64sat_s60000.npz",
+   "npz_sha256": "d6ae66bca62483ff3213a4f567ee76bdcf01251da0fc79f76d60fe9315e4a6a3",
+   "receipt_verified": true,
+   "generation": {
+    "checkpoint": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/runs/hoL_giroux2022/K1000/lora64sat/checkpoint_ft_s60000.pt",
+    "checkpoint_sha256": "2766d52ee1068e260f8ae8f2a717f17dcaa652661b6384e5c619f2d85e7076a3",
+    "input_condition_path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/data/giroux2022/test_raw_K1000.npz",
+    "input_condition_sha256": "ac755a41a900fac61df07f3a95be029353cfd4869439fa19612e599527d37665",
+    "seed": 20260927,
+    "steps": 50,
+    "support_readout": "bernoulli",
+    "weights": "ema"
+   },
+   "condition_for_broadcast": {
+    "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_crosscohort_20261002/data/giroux2022/test_raw_K1000.npz",
+    "sha256": "ac755a41a900fac61df07f3a95be029353cfd4869439fa19612e599527d37665"
+   },
+   "poisson": {
+    "seed": 20260916,
+    "rule": "whole NPZ, rows in NPZ order",
+    "nnz_whole_npz": 9059612,
+    "median_library_whole_npz_panel": 1874.0,
+    "n_rows_whole_npz": 36000
+   },
+   "annotation": {
+    "model_pt_sha256": "bf3655899e5abee435340fc008656c4a17342da0f30b70cfd43de8e9223e1930",
+    "zero_padded_model_genes": 0,
+    "label_counts": {
+     "T cell": 25737,
+     "B cell": 3945,
+     "NK cell": 3254,
+     "Monocyte": 1900,
+     "Other": 591,
+     "DC": 573
+    },
+    "mean_l2_max_prob": 0.9226587959180276,
+    "predict_seconds": 4.1
+   },
+   "n_donors": 24,
+   "n_generated_cells": 36000,
+   "composition": {
+    "l1_mean": 0.42970541878518415,
+    "l1_median": 0.34827577456302117,
+    "l1_obs_loo_null_mean": 0.30584824369245694,
+    "l1_generated_loo_mean_vs_observed_mean": 0.4240294634704537,
+    "composition_top1": 0.08333333333333333,
+    "mean_fraction_generated": {
+     "T cell": 0.714917,
+     "B cell": 0.109583,
+     "Monocyte": 0.052778,
+     "NK cell": 0.090389,
+     "DC": 0.015917,
+     "Other": 0.016417
+    },
+    "mean_fraction_observed": {
+     "T cell": 0.566621,
+     "B cell": 0.065423,
+     "Monocyte": 0.192479,
+     "NK cell": 0.114152,
+     "DC": 0.016843,
+     "Other": 0.044482
+    }
+   },
+   "pseudobulk": {
+    "n_genes": 5563,
+    "r_median": 0.8469462885794707,
+    "r_mean": 0.8466876486746496,
+    "top1": 0.16666666666666666,
+    "n_top1": 4,
+    "own_rank_median": 5.0,
+    "chance_top1": 0.041666666666666664
+   },
+   "precision_coverage": {
+    "precision": 0.9936111111111111,
+    "coverage": 0.8152777777777778,
+    "density": 2.0655833333333335,
+    "n_query": 3600,
+    "n_reference": 3600
+   },
+   "precision_coverage_depth_matched": {
+    "precision": 0.975,
+    "coverage": 0.8447222222222223,
+    "density": 1.4498611111111113,
+    "detected_median_before": {
+     "R1": 254.0,
+     "Q": 167.0
+    },
+    "detected_median_after": {
+     "R1": 167.0,
+     "Q": 167.0
+    }
+   },
+   "monocyte_pct": {
+    "generated_mean_over_donors": 5.277777777777778,
+    "observed_mean_over_donors": 19.247866360995445,
+    "generated_pooled_cells": 5.277777777777778
+   },
+   "median_umi_per_cell_scoring_genes": {
+    "generated_pooled": 501.0,
+    "observed_pooled": 471.0,
+    "generated_mean_of_donor_medians": 507.5833435058594,
+    "observed_mean_of_donor_medians": 483.6458333333333
+   },
+   "contrast": {
+    "mean_cp10k": {
+     "per_type": {
+      "T cell": {
+       "r_generated": 0.5098273245011848,
+       "r_broadcast": 0.6665223768938867,
+       "partial_r_generated_given_broadcast": 0.334365304363034,
+       "r_generated_vs_broadcast": 0.42668042906201875,
+       "n_genes": 2908
+      },
+      "B cell": {
+       "r_generated": 0.24706346201199647,
+       "r_broadcast": 0.5741068019580355,
+       "partial_r_generated_given_broadcast": 0.11591473787987704,
+       "r_generated_vs_broadcast": 0.2712253425769915,
+       "n_genes": 3053
+      },
+      "Monocyte": {
+       "r_generated": 0.11890517079980385,
+       "r_broadcast": 0.6998157471451955,
+       "partial_r_generated_given_broadcast": 0.05362235600326642,
+       "r_generated_vs_broadcast": 0.11554184392874314,
+       "n_genes": 3056
+      },
+      "NK cell": {
+       "r_generated": 0.3359423599280391,
+       "r_broadcast": 0.6119759414285395,
+       "partial_r_generated_given_broadcast": 0.2598485138150453,
+       "r_generated_vs_broadcast": 0.22147590804411849,
+       "n_genes": 2955
+      }
+     },
+     "mean_r_generated": 0.3029345793102561,
+     "mean_r_broadcast": 0.6381052168564143,
+     "mean_partial_r": 0.19093772801530567
+    },
+    "pseudobulk_log1p_cpm": {
+     "per_type": {
+      "T cell": {
+       "r_generated": 0.5572861636616837,
+       "r_broadcast": 0.829771536027708,
+       "partial_r_generated_given_broadcast": 0.25730838851033505,
+       "r_generated_vs_broadcast": 0.524236291540207,
+       "n_genes": 2908
+      },
+      "B cell": {
+       "r_generated": 0.2811018823415274,
+       "r_broadcast": 0.7295906969005199,
+       "partial_r_generated_given_broadcast": 0.09090300897766705,
+       "r_generated_vs_broadcast": 0.3041147371912349,
+       "n_genes": 3053
+      },
+      "Monocyte": {
+       "r_generated": 0.05170185072398673,
+       "r_broadcast": 0.8460762046394,
+       "partial_r_generated_given_broadcast": -0.017982034932462792,
+       "r_generated_vs_broadcast": 0.07240745984119075,
+       "n_genes": 3056
+      },
+      "NK cell": {
+       "r_generated": 0.2804746234995544,
+       "r_broadcast": 0.7735167267842559,
+       "partial_r_generated_given_broadcast": 0.17535090167291406,
+       "r_generated_vs_broadcast": 0.22252644978132635,
+       "n_genes": 2955
+      }
+     },
+     "mean_r_generated": 0.29264113005668807,
+     "mean_r_broadcast": 0.794738791087971,
+     "mean_partial_r": 0.12639506605711334
+    }
+   },
+   "deg_top50": {
+    "types_scored": [
+     "T cell",
+     "B cell",
+     "Monocyte",
+     "NK cell",
+     "DC"
+    ],
+    "types_not_scored": {},
+    "per_type": {
+     "T cell": {
+      "deg_scored": true,
+      "top50_overlap": 5,
+      "top50_jaccard": 0.05263157894736842,
+      "obs_top_list_size": 50,
+      "gen_top_list_size": 50,
+      "top50_overlap_broadcast": 10,
+      "top50_jaccard_broadcast": 0.1724137931034483,
+      "n_cells_obs": 145034,
+      "n_cells_gen": 25737
+     },
+     "B cell": {
+      "deg_scored": true,
+      "top50_overlap": 4,
+      "top50_jaccard": 0.4444444444444444,
+      "obs_top_list_size": 9,
+      "gen_top_list_size": 4,
+      "top50_overlap_broadcast": 7,
+      "top50_jaccard_broadcast": 0.35,
+      "n_cells_obs": 17383,
+      "n_cells_gen": 3945
+     },
+     "Monocyte": {
+      "deg_scored": true,
+      "top50_overlap": 2,
+      "top50_jaccard": 0.2222222222222222,
+      "obs_top_list_size": 7,
+      "gen_top_list_size": 4,
+      "top50_overlap_broadcast": 7,
+      "top50_jaccard_broadcast": 0.3888888888888889,
+      "n_cells_obs": 47311,
+      "n_cells_gen": 1900
+     },
+     "NK cell": {
+      "deg_scored": true,
+      "top50_overlap": 3,
+      "top50_jaccard": 0.058823529411764705,
+      "obs_top_list_size": 50,
+      "gen_top_list_size": 4,
+      "top50_overlap_broadcast": 10,
+      "top50_jaccard_broadcast": 0.1724137931034483,
+      "n_cells_obs": 29061,
+      "n_cells_gen": 3254
+     },
+     "DC": {
+      "deg_scored": true,
+      "top50_overlap": 1,
+      "top50_jaccard": 0.14285714285714285,
+      "obs_top_list_size": 7,
+      "gen_top_list_size": 1,
+      "top50_overlap_broadcast": 5,
+      "top50_jaccard_broadcast": 0.25,
+      "n_cells_obs": 4328,
+      "n_cells_gen": 573
+     }
+    },
+    "sum": 15,
+    "sum_4type": 14,
+    "max": 123,
+    "mean": 3.0,
+    "sum_broadcast": 39,
+    "sum_broadcast_4type": 34,
+    "broadcast_n_significant": 18,
+    "seconds": 0.07
+   },
+   "seconds": 9.0
+  }
+ },
+ "code_sha256": {
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/tools/score_ft.py": "ec5f40b4cea6d5062d0138f3b934abb6465e34cbfeac869c8a0906924c033aec",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/tools/ft_common.py": "4a6366b0f22208cd58714fc0261b8840c643b9d4569704d560cfd93534cbc747",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/utility_studies_final6k_20260925/umap/cell_immicro_5arm/tools/stage_b_embed.py": "d056cfdf38fb0e5f478756f370174e993b75d0c973c48e02e21749175a4ca2da",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/bulk2sc-paper/scripts/evaluate_generated_vs_truth.py": "48b12d794782b840838853e93e0fc55fab00532c4b3698f4588e926b985d2881",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/utility_studies_final6k_20260925/immicro/tools/immicro_common.py": "c5c30e917148cc5e664e91ea30c11bbdff9b6d1f53a3e0a13e16b1917eff37e9",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/aifi_scanvi_rebuild_20260916/predict_scanvi.py": "482ee5dcf4d72b54bc36cbb6dcff62e5030bc7afc46cb5fb01eca73a82ee45a5",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/aifi_scanvi_rebuild_20260916/poisson_readout.py": "8bfa080942fbcdc89ae23b276524a1a83266fb8cbc56b9af279e50f82e6cde38"
+ },
+ "seconds": 10.6
+}

paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/scores/giroux2022/locoL_lora64sat_s69250_K1000/README.md

@@ -0,0 +1,73 @@
+# score_ft.py output: giroux2022, 24 scored donors
+
+Arms: locoL_lora64sat_s69250_K1000.  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/giroux2022/reference.npz` (sha dd0152f7bde4), built from datadir:/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/giroux2022;
+  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/giroux2022/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 (5563 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 24 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) (3600 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 0.9981, 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
+   0.9981, 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 ['severe'] (n = 11) vs control groups ['mild'] (n = 13); 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 5563 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/giroux2022/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 11 + 13
+   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/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/scores/giroux2022/locoL_lora64sat_s69250_K1000/contrast.csv

@@ -0,0 +1,14 @@
+arm,celltype,statistic,n_genes,r_generated,r_broadcast,partial_r_generated_given_broadcast,r_generated_vs_broadcast,n_case_donors_obs,n_control_donors_obs,n_case_donors_gen,n_control_donors_gen,n_cells_obs,n_cells_gen,deg_scored,top50_overlap,top50_jaccard,obs_top_list_size,gen_top_list_size,n_sig_obs,n_sig_gen,top50_overlap_broadcast,top50_jaccard_broadcast,broadcast_top_list_size,n_sig_broadcast
+locoL_lora64sat_s69250_K1000,T cell,mean_cp10k,2908,0.5096628991111606,0.6665223768938867,0.3163871417134233,0.4483509968417228,11,13,11,13,145034,25458,,,,,,,,,,,
+locoL_lora64sat_s69250_K1000,T cell,pseudobulk_log1p_cpm,2908,0.5742244021544237,0.829771536027708,0.25128486726112353,0.5509821711983663,11,13,11,13,145034,25458,,,,,,,,,,,
+locoL_lora64sat_s69250_K1000,B cell,mean_cp10k,3053,0.27137785603804643,0.5741068019580355,0.11165691271487725,0.3219303225544919,11,13,11,13,17383,3799,,,,,,,,,,,
+locoL_lora64sat_s69250_K1000,B cell,pseudobulk_log1p_cpm,3053,0.320488072417984,0.7295906969005199,0.11225040574364412,0.3403338290776521,11,13,11,13,17383,3799,,,,,,,,,,,
+locoL_lora64sat_s69250_K1000,Monocyte,mean_cp10k,3056,0.10398568021712025,0.6998157471451955,0.042790864205689706,0.10515428972795672,11,13,11,13,47311,2082,,,,,,,,,,,
+locoL_lora64sat_s69250_K1000,Monocyte,pseudobulk_log1p_cpm,3056,0.051408863068628706,0.8460762046394,-0.03424290504708036,0.08226278611078666,11,13,11,13,47311,2082,,,,,,,,,,,
+locoL_lora64sat_s69250_K1000,NK cell,mean_cp10k,2955,0.37441718643097843,0.6119759414285395,0.2908121155118732,0.2477031487522225,11,13,11,13,29061,3665,,,,,,,,,,,
+locoL_lora64sat_s69250_K1000,NK cell,pseudobulk_log1p_cpm,2955,0.34306725644630226,0.7735167267842559,0.18951129531014732,0.29515922235838155,11,13,11,13,29061,3665,,,,,,,,,,,
+locoL_lora64sat_s69250_K1000,T cell,welch_top50_deg,5563,,,,,11,13,11,13,145034,25458,True,5,0.05263157894736842,50,50,306,178,10,0.1724137931034483,18,18
+locoL_lora64sat_s69250_K1000,B cell,welch_top50_deg,5563,,,,,11,13,11,13,17383,3799,True,6,0.6,9,7,9,7,7,0.35,18,18
+locoL_lora64sat_s69250_K1000,Monocyte,welch_top50_deg,5563,,,,,11,13,11,13,47311,2082,True,2,0.16666666666666666,7,7,7,7,7,0.3888888888888889,18,18
+locoL_lora64sat_s69250_K1000,NK cell,welch_top50_deg,5563,,,,,11,13,11,13,29061,3665,True,4,0.08,50,4,83,4,10,0.1724137931034483,18,18
+locoL_lora64sat_s69250_K1000,DC,welch_top50_deg,5563,,,,,11,13,11,13,4328,664,True,2,0.16666666666666666,7,7,7,7,5,0.25,18,18

paper-releases/pbisc-diffusion/updates/20261003T103040Z-crosscohort-k1000-saturation/diffusion/dataset/server_1/model8/plans/paired_crosscohort_20261002/scores/giroux2022/locoL_lora64sat_s69250_K1000/per_donor.csv

@@ -0,0 +1,25 @@
+arm,donor,group,n_generated_cells,n_observed_cells,n_reference_cells,gen_T cell,obs_T cell,gen_B cell,obs_B cell,gen_Monocyte,obs_Monocyte,gen_NK cell,obs_NK cell,gen_DC,obs_DC,gen_Other,obs_Other,comp_l1,comp_l1_obs_loo_null,pb_r,pb_own_rank,pb_top1,precision,coverage,precision_depth_matched,coverage_depth_matched,mono_pct_gen,mono_pct_obs,median_umi_gen,median_umi_obs
+locoL_lora64sat_s69250_K1000,450905_d0,mild,1500,5914,150,0.7353333333333333,0.30943523841731485,0.09866666666666667,0.036185322962461956,0.03333333333333333,0.4293202570172472,0.06933333333333333,0.12089956036523504,0.048,0.009976327358809604,0.015333333333333332,0.09418329387893135,1.0528062225228272,0.6266504923198378,0.8511764861189245,22,False,0.9733333333333334,0.6733333333333333,0.96,0.7266666666666667,3.3333333333333335,42.93202570172472,527.5,574.5
+locoL_lora64sat_s69250_K1000,450905_d14,mild,1500,7407,150,0.6673333333333333,0.4422843256379101,0.08266666666666667,0.02929661131362225,0.092,0.27420008100445525,0.12733333333333333,0.15782368030241664,0.02666666666666667,0.02983664101525584,0.004,0.06655866072633995,0.5568381260969354,0.36155260046011806,0.8564508892526626,12,False,1.0,0.8066666666666666,1.0,0.9,9.2,27.420008100445525,613.0,558.0
+locoL_lora64sat_s69250_K1000,0B943B_d0,mild,1500,11344,150,0.6646666666666666,0.6220028208744711,0.204,0.08312764456981664,0.02666666666666667,0.12588152327221439,0.07666666666666666,0.11865303244005641,0.023333333333333334,0.023977433004231313,0.004666666666666667,0.026357545839210156,0.32707240244475777,0.1791278384873501,0.8578330476443813,6,False,0.9933333333333333,0.86,0.9933333333333333,0.9,2.666666666666667,12.588152327221438,511.0,539.0
+locoL_lora64sat_s69250_K1000,0B943B_d14,mild,1500,15650,150,0.572,0.5652396166134185,0.11266666666666666,0.11175718849840256,0.21266666666666667,0.17086261980830672,0.07066666666666667,0.09789137380191694,0.012,0.009968051118210863,0.02,0.04428115015974441,0.10301171458998934,0.10063602424453227,0.8788080354552265,1,True,0.9933333333333333,0.84,0.96,0.88,21.266666666666666,17.08626198083067,661.0,461.0
+locoL_lora64sat_s69250_K1000,7085CA_d0,mild,1500,12585,150,0.542,0.21660707191100517,0.12133333333333333,0.024791418355184745,0.182,0.5874453714739769,0.09333333333333334,0.13198251887167262,0.056666666666666664,0.017719507350019866,0.004666666666666667,0.021454112038140644,0.9217640047675806,0.8829287440863147,0.8881438783717525,1,True,0.9666666666666667,0.72,0.9066666666666666,0.78,18.2,58.74453714739769,641.5,640.0
+locoL_lora64sat_s69250_K1000,7085CA_d7,mild,1500,9515,150,0.5233333333333333,0.521492380451918,0.10066666666666667,0.0722017866526537,0.08933333333333333,0.10110352075669994,0.238,0.2725170782974251,0.03666666666666667,0.019863373620599055,0.012,0.012821860220704152,0.09421825188299174,0.3632282731474772,0.8518609672461277,4,False,1.0,0.8866666666666667,0.9933333333333333,0.9133333333333333,8.933333333333334,10.110352075669994,481.5,502.0
+locoL_lora64sat_s69250_K1000,7085CA_d14,mild,1500,11887,150,0.524,0.472533019264743,0.118,0.07133843694792631,0.09466666666666666,0.16210986792294102,0.21466666666666667,0.24270211155043325,0.03866666666666667,0.018591738874400604,0.01,0.032724825439555814,0.2364069431591935,0.2993283707516418,0.8542525543234089,6,False,1.0,0.7533333333333333,0.98,0.8533333333333334,9.466666666666667,16.2109867922941,478.0,536.0
+locoL_lora64sat_s69250_K1000,2AD75E_d0,mild,1500,7693,150,0.65,0.6034056934875861,0.11266666666666666,0.06408423241908229,0.066,0.14064734173924348,0.13066666666666665,0.15949564539191471,0.025333333333333333,0.025347718705316522,0.015333333333333332,0.007019368256856883,0.20698141167294948,0.20681142084856263,0.8597241000817534,1,True,1.0,0.8466666666666667,0.9733333333333334,0.8933333333333333,6.6000000000000005,14.064734173924348,528.0,532.0
+locoL_lora64sat_s69250_K1000,2AD75E_d7,mild,1500,10499,150,0.5126666666666667,0.5808172206876845,0.15666666666666668,0.14372797409277074,0.10066666666666667,0.10277169254214687,0.16466666666666666,0.14715687208305553,0.020666666666666667,0.013715591961139155,0.04466666666666667,0.011810648633203162,0.14051115979299605,0.27874882058987843,0.8389605952183264,2,False,0.9933333333333333,0.9266666666666666,0.9866666666666667,0.96,10.066666666666666,10.277169254214687,395.0,443.0
+locoL_lora64sat_s69250_K1000,2AD75E_d14,mild,1500,13692,150,0.6646666666666666,0.5811422728600643,0.17466666666666666,0.12503651767455448,0.059333333333333335,0.14198071866783524,0.078,0.12496348232544552,0.018,0.016579024247735902,0.005333333333333333,0.010297984224364592,0.2691510371019573,0.19762923258298543,0.8592599298952446,1,True,1.0,0.9,0.98,0.9133333333333333,5.933333333333334,14.198071866783524,483.5,428.0
+locoL_lora64sat_s69250_K1000,B85D75_d0,severe,1500,11441,150,0.8446666666666667,0.7173324010138974,0.07866666666666666,0.031203566121842496,0.0033333333333333335,0.11441307578008915,0.06466666666666666,0.06039681845992483,0.0006666666666666666,0.016519535005681322,0.008,0.06013460361856481,0.3581344288086706,0.35882358586377083,0.8494999691241787,1,True,1.0,0.8733333333333333,0.9933333333333333,0.8733333333333333,0.33333333333333337,11.441307578008916,510.0,347.0
+locoL_lora64sat_s69250_K1000,B85D75_d7,severe,1500,12328,150,0.8633333333333333,0.6435756002595717,0.07066666666666667,0.06448734587929916,0.007333333333333333,0.13741077222582737,0.051333333333333335,0.04201817001946788,0.006,0.027741726151849447,0.0013333333333333333,0.08476638546398442,0.470504434349989,0.28301793033829503,0.8496378841491095,1,True,0.9933333333333333,0.8466666666666667,0.98,0.8133333333333334,0.7333333333333333,13.741077222582737,522.0,363.0
+locoL_lora64sat_s69250_K1000,180E1A_d0,severe,1500,7157,150,0.8,0.7019700992035769,0.10466666666666667,0.04135811094033813,0.008,0.13175911694844208,0.072,0.0655302501047925,0.007333333333333333,0.013273718038284197,0.008,0.04610870476456616,0.3356164128359183,0.31372800769589615,0.8336145537735395,12,False,0.9933333333333333,0.74,1.0,0.78,0.8,13.175911694844208,455.0,433.0
+locoL_lora64sat_s69250_K1000,180E1A_d7,severe,1500,10591,150,0.8173333333333334,0.7348692285903126,0.06666666666666667,0.04909829100179398,0.008,0.09064299877254273,0.09666666666666666,0.08280615617033331,0.004666666666666667,0.010480596733075253,0.006666666666666667,0.03210272873194221,0.2277859818084537,0.37628768415060193,0.8456540996138974,2,False,1.0,0.82,0.9933333333333333,0.8466666666666667,0.8,9.064299877254273,492.0,408.0
+locoL_lora64sat_s69250_K1000,180E1A_d14,severe,1500,12511,150,0.8426666666666667,0.680840860043162,0.04466666666666667,0.05147470226200943,0.012666666666666666,0.12277196067460634,0.08733333333333333,0.0800095915594277,0.008,0.009831348413396212,0.004666666666666667,0.05507153704739829,0.33829909679482056,0.28939519394073066,0.8410749585192591,3,False,0.9933333333333333,0.8133333333333334,0.98,0.7933333333333333,1.2666666666666666,12.277196067460634,473.0,342.0
+locoL_lora64sat_s69250_K1000,82CCF5_d0,severe,1500,10125,150,0.48333333333333334,0.5422222222222223,0.16533333333333333,0.11832098765432099,0.22666666666666666,0.1539753086419753,0.098,0.09303703703703704,0.016666666666666666,0.016691358024691356,0.01,0.07575308641975309,0.24933333333333335,0.18419072822032903,0.8716853831894592,9,False,0.9933333333333333,0.78,0.9733333333333334,0.8733333333333333,22.666666666666664,15.397530864197531,701.5,390.0
+locoL_lora64sat_s69250_K1000,82CCF5_d7,severe,1500,10860,150,0.6706666666666666,0.619060773480663,0.12266666666666666,0.09972375690607735,0.037333333333333336,0.14152854511970533,0.13866666666666666,0.09410681399631676,0.022,0.013627992633517496,0.008666666666666666,0.03195211786372008,0.2549613259668507,0.19419086678828923,0.8495770426786498,2,False,0.9933333333333333,0.8533333333333334,0.9933333333333333,0.8466666666666667,3.733333333333334,14.152854511970533,501.0,497.0
+locoL_lora64sat_s69250_K1000,82CCF5_d14,severe,1500,11056,150,0.7326666666666667,0.6330499276410999,0.10666666666666667,0.10121201157742403,0.035333333333333335,0.11369392185238784,0.104,0.06828871201157742,0.014,0.014562228654124458,0.007333333333333333,0.0691931982633864,0.281565364206464,0.2787071044352101,0.8539696157876856,3,False,1.0,0.7666666666666667,0.9866666666666667,0.8133333333333334,3.5333333333333337,11.369392185238784,538.0,394.0
+locoL_lora64sat_s69250_K1000,BAAF62_d0,mild,1500,14741,150,0.6366666666666667,0.5279153381724442,0.11933333333333333,0.0384641476154942,0.035333333333333335,0.1854690997897022,0.184,0.1717658232141646,0.018,0.023132758971575875,0.006666666666666667,0.053252832236618954,0.4037093819957941,0.14602913946264867,0.8514811618745278,9,False,1.0,0.8933333333333333,0.9866666666666667,0.9066666666666666,3.5333333333333337,18.54690997897022,510.0,449.0
+locoL_lora64sat_s69250_K1000,BAAF62_d7,mild,1500,7297,150,0.7753333333333333,0.6932986158695355,0.12533333333333332,0.04070165821570509,0.01,0.13389063998903658,0.07333333333333333,0.0941482801151158,0.009333333333333334,0.016033986569823216,0.006666666666666667,0.021926819240783883,0.33333278516285225,0.25991971231178185,0.8384453140559326,9,False,1.0,0.78,1.0,0.8533333333333334,1.0,13.389063998903659,474.0,580.0
+locoL_lora64sat_s69250_K1000,BAAF62_d14,mild,1500,11923,150,0.816,0.5431518913025245,0.062,0.019206575526293717,0.010666666666666666,0.2843244149962258,0.09,0.10416841398976767,0.014,0.032793759959741674,0.007333333333333333,0.016354944225446616,0.6312830663423634,0.2201331132920326,0.8417288158175747,9,False,1.0,0.8333333333333334,1.0,0.8466666666666667,1.0666666666666667,28.43244149962258,489.0,513.0
+locoL_lora64sat_s69250_K1000,0BF51C_d0,severe,1500,6510,150,0.896,0.4012288786482335,0.036,0.01674347158218126,0.034666666666666665,0.4609831029185868,0.020666666666666667,0.07726574500768049,0.006666666666666667,0.006144393241167435,0.006,0.03763440860215054,1.029099846390169,0.5556611725994968,0.823197939349846,1,True,0.9866666666666667,0.5933333333333334,0.9466666666666667,0.5266666666666666,3.4666666666666663,46.09831029185868,392.0,628.0
+locoL_lora64sat_s69250_K1000,0BF51C_d7,severe,1500,10105,150,0.8846666666666667,0.612370113805047,0.054,0.05700148441365661,0.0006666666666666666,0.1607125185551707,0.05333333333333334,0.05294408708560119,0.006,0.0073231073725878275,0.0013333333333333333,0.10964868876793667,0.5453715982187036,0.22247826940326798,0.8147818201701309,16,False,0.9933333333333333,0.8,0.9933333333333333,0.8133333333333334,0.06666666666666667,16.071251855517072,505.0,509.0
+locoL_lora64sat_s69250_K1000,0BF51C_d14,severe,1500,11419,150,0.8526666666666667,0.6330676941938874,0.094,0.07960416849111131,0.0013333333333333333,0.151589456169542,0.046,0.07907872843506436,0.0033333333333333335,0.010508801120938787,0.0026666666666666666,0.04615115158945617,0.467989607963336,0.1611535225979176,0.8236048040043449,7,False,1.0,0.8,0.9866666666666667,0.88,0.13333333333333333,15.1589456169542,472.0,541.0