Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261005T121346Z-scsemi-cv-budget-interim/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores/ji2025_chili/scsemi_cv_20261002/fold2/per_donor.csv

@@ -0,0 +1,7 @@
+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
+scsemi__raw,N055_AC,AC,1500,4308,150,0.36133333333333334,0.340297121634169,0.078,0.07799442896935933,0.43,0.32474466109563604,0.07933333333333334,0.2051996285979573,0.027333333333333334,0.0060352831940575676,0.024,0.0457288765088208,0.29519034354688956,0.28753046275160377,0.6515429782713951,1,True,0.98,0.42,0.6533333333333333,0.58,43.0,32.47446610956361,1063.0,1139.5
+scsemi__raw,N150_BC,BC,1500,2433,150,0.7166666666666667,0.802301685162351,0.009333333333333334,0.02301685162351007,0.11733333333333333,0.1385121249486231,0.086,0.02178380600082203,0.043333333333333335,0.005754212905877518,0.027333333333333334,0.008631319358816275,0.2409946568023017,0.8495800601407223,0.6445675186094368,1,True,0.9866666666666667,0.14,0.72,0.6733333333333333,11.733333333333333,13.851212494862311,1063.0,1629.0
+scsemi__raw,N001_TOL,TOL,1500,5658,150,0.6226666666666667,0.51926475786497,0.0026666666666666666,0.03375751148815836,0.20466666666666666,0.21067515022976316,0.018,0.2163308589607635,0.12333333333333334,0.003711558854718982,0.028666666666666667,0.016260162601626018,0.4708603746907034,0.3001465278392606,0.46695160485388026,4,False,0.9266666666666666,0.46,0.9333333333333333,0.16666666666666666,20.466666666666665,21.067515022976316,1061.0,787.5
+scsemi__raw,N001_FU1,FU1,1500,4547,150,0.33666666666666667,0.29601935341983726,0.013333333333333334,0.06575764240158347,0.38466666666666666,0.48823400043985044,0.09933333333333333,0.13591378931163403,0.08466666666666667,0.0010996261271167802,0.08133333333333333,0.012975588299978007,0.3851441976394693,0.3817288628232209,0.5143407286343378,3,False,0.7066666666666667,0.31333333333333335,0.6666666666666666,0.22,38.46666666666667,48.823400043985046,1062.5,1165.0
+scsemi__raw,N002_TOL,TOL,1500,6318,150,0.3473333333333333,0.41579613801836024,0.028666666666666667,0.05777144666033555,0.21266666666666667,0.3779677113010446,0.26,0.1179170623615068,0.12533333333333332,0.020101297879075657,0.026,0.010446343779677113,0.5257372586261475,0.22203438554178426,0.5118157613212175,4,False,0.9733333333333334,0.37333333333333335,0.78,0.26,21.266666666666666,37.79677113010446,1063.0,933.5
+scsemi__raw,N002_FU1,FU1,1500,4975,150,0.4126666666666667,0.31618090452261305,0.008,0.07276381909547738,0.242,0.45587939698492463,0.218,0.12261306532663317,0.09066666666666667,0.02271356783919598,0.028666666666666667,0.00984924623115578,0.5572864321608042,0.4130670417000685,0.49955333612466496,6,False,0.96,0.36666666666666664,0.7133333333333334,0.15333333333333332,24.2,45.58793969849246,1062.0,979.0

paper-releases/pbisc-diffusion/updates/20261005T121346Z-scsemi-cv-budget-interim/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores/ji2025_chili/scsemi_cv_20261002/fold2/summary.json

@@ -0,0 +1,459 @@
+{
+ "schema": "paired_ft_scores.v1",
+ "dataset": "ji2025_chili",
+ "written_utc": "2026-10-02T08:58:17.163818+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",
+  "ji2025_chili",
+  "--out",
+  "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/ji2025_chili/scsemi_cv_20261002/fold2",
+  "--threads",
+  "8",
+  "--arm",
+  "scsemi__raw=/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/generated/ji2025_chili/fold2/scsemi__raw.npz",
+  "--case-groups",
+  "TOL,FU1",
+  "--control-groups",
+  "BC,AC"
+ ],
+ "arms": [
+  "scsemi__raw"
+ ],
+ "donors": [
+  "N055_AC",
+  "N150_BC",
+  "N001_TOL",
+  "N001_FU1",
+  "N002_TOL",
+  "N002_FU1"
+ ],
+ "n_donors": 6,
+ "reference": {
+  "cache": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scoring_cache/ji2025_chili",
+  "reference_sha256": "19c30f56db699eb4f5a642927c1cd6e173212b877c29ff20d84e60a5ecccef6d",
+  "features_sha256": "049795e3a2470a223347015c0c965f56ca90c21143aad6f72d12e39f3e95e58b",
+  "n_r1_cells": 900,
+  "pca": {
+   "n_components": 50,
+   "explained_variance_ratio_sum": 0.3404575288295746,
+   "n_fit_cells": 900
+  },
+  "r2_control": {
+   "precision": 1.0,
+   "coverage": 1.0,
+   "density": 4.421555555555555,
+   "n_query": 900,
+   "depth_matched": {
+    "precision": 0.9988888888888889,
+    "coverage": 1.0,
+    "density": 4.357333333333333
+   }
+  },
+  "observed_l1_loo_null_mean": 0.40901455679944343
+ },
+ "contrast_definition": {
+  "enabled": true,
+  "rule": "explicit --case-groups / --control-groups",
+  "group_source": "features groups",
+  "case_groups": [
+   "TOL",
+   "FU1"
+  ],
+  "control_groups": [
+   "BC",
+   "AC"
+  ],
+  "n_case_donors": 4,
+  "n_control_donors": 2,
+  "n_donors_in_neither": 0,
+  "case": [
+   false,
+   false,
+   true,
+   true,
+   true,
+   true
+  ],
+  "ctrl": [
+   true,
+   true,
+   false,
+   false,
+   false,
+   false
+  ],
+  "groups": [
+   "AC",
+   "BC",
+   "TOL",
+   "FU1",
+   "TOL",
+   "FU1"
+  ]
+ },
+ "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": 5752,
+  "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/ji2025_chili/deg_reference.npz",
+  "aggregates_sha256": "4f5eb8c8389ef1308ce3ec44be8be024cedb97de37351d109d9070c818df5a9e",
+  "aggregates_note": null,
+  "types_not_scored_observed": [],
+  "observed": {
+   "T cell": {
+    "n_significant": 142,
+    "n_tested": 5455,
+    "n_case_donors": 4,
+    "n_control_donors": 2,
+    "n_cells": 11902,
+    "top_list_size": 50
+   },
+   "B cell": {
+    "n_significant": 0,
+    "n_tested": 4935,
+    "n_case_donors": 4,
+    "n_control_donors": 2,
+    "n_cells": 1609,
+    "top_list_size": 0
+   },
+   "Monocyte": {
+    "n_significant": 0,
+    "n_tested": 5563,
+    "n_case_donors": 4,
+    "n_control_donors": 2,
+    "n_cells": 9804,
+    "top_list_size": 0
+   },
+   "NK cell": {
+    "n_significant": 3,
+    "n_tested": 5060,
+    "n_case_donors": 4,
+    "n_control_donors": 2,
+    "n_cells": 4134,
+    "top_list_size": 3
+   },
+   "DC": {
+    "n_significant": 3,
+    "n_tested": 4460,
+    "n_case_donors": 4,
+    "n_control_donors": 2,
+    "n_cells": 306,
+    "top_list_size": 3
+   }
+  },
+  "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": {
+  "scsemi__raw": {
+   "comp_l1": 0.41253554391105257,
+   "comp_l1_obs_loo_null": 0.40901455679944343,
+   "precision": 0.9222222222222223,
+   "coverage": 0.34555555555555556,
+   "precision_dm": 0.7444444444444445,
+   "coverage_dm": 0.3422222222222222,
+   "pb_r_median": 0.5130782449777777,
+   "pb_top1": 0.3333333333333333,
+   "mono_pct_gen": 26.522222222222226,
+   "mono_pct_obs": 33.2668840833307,
+   "median_umi_gen": 1063.0,
+   "median_umi_obs": 1030.0,
+   "contrast_mean_cp10k_r_gen": -0.25399219264866946,
+   "contrast_pseudobulk_log1p_cpm_r_gen": -0.02929690578119841,
+   "contrast_mean_cp10k_r_broadcast": -0.10111878626433048,
+   "contrast_pseudobulk_log1p_cpm_r_broadcast": 0.010537158270212697,
+   "deg_top50_T": 0,
+   "deg_top50_B": 0,
+   "deg_top50_Mono": 0,
+   "deg_top50_NK": 0,
+   "deg_top50_DC": 0,
+   "deg_top50_sum": 0,
+   "deg_top50_max": 56,
+   "deg_top50_sum_4type": 0,
+   "deg_top50_sum_broadcast": 0,
+   "seconds": 4.7
+  }
+ },
+ "per_arm": {
+  "scsemi__raw": {
+   "arm": "scsemi__raw",
+   "npz": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/generated/ji2025_chili/fold2/scsemi__raw.npz",
+   "npz_sha256": "e9b47daf36a07b5dbbf53623ddb0e850ff00e56178842acac7ac8dae8c316cfd",
+   "receipt_verified": true,
+   "generation": {
+    "checkpoint": null,
+    "checkpoint_sha256": null,
+    "input_condition_path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/ji2025_chili/fold2/test_raw.npz",
+    "input_condition_sha256": "d9a1eb82edf9f6a2ccd110049a18740554ca1d2b02f4ff331a12e7eb7207a7c6",
+    "seed": 20260927,
+    "steps": null,
+    "support_readout": "upstream (values <= 10 normalized units zeroed by fast_semi)",
+    "weights": null
+   },
+   "condition_for_broadcast": {
+    "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/ji2025_chili/fold2/test_raw.npz",
+    "sha256": "d9a1eb82edf9f6a2ccd110049a18740554ca1d2b02f4ff331a12e7eb7207a7c6"
+   },
+   "poisson": {
+    "seed": 20260916,
+    "rule": "whole NPZ, rows in NPZ order",
+    "nnz_whole_npz": 1081824,
+    "median_library_whole_npz_panel": 1063.0,
+    "n_rows_whole_npz": 9000
+   },
+   "annotation": {
+    "model_pt_sha256": "bf3655899e5abee435340fc008656c4a17342da0f30b70cfd43de8e9223e1930",
+    "zero_padded_model_genes": 1,
+    "label_counts": {
+     "T cell": 4196,
+     "Monocyte": 2387,
+     "NK cell": 1141,
+     "DC": 742,
+     "Other": 324,
+     "B cell": 210
+    },
+    "mean_l2_max_prob": 0.8843977564159367,
+    "predict_seconds": 3.6
+   },
+   "n_donors": 6,
+   "n_generated_cells": 9000,
+   "composition": {
+    "l1_mean": 0.41253554391105257,
+    "l1_median": 0.42800228616508634,
+    "l1_obs_loo_null_mean": 0.40901455679944343,
+    "l1_generated_loo_mean_vs_observed_mean": 0.48486440662867974,
+    "composition_top1": 0.16666666666666666,
+    "mean_fraction_generated": {
+     "T cell": 0.466222,
+     "B cell": 0.023333,
+     "Monocyte": 0.265222,
+     "NK cell": 0.126778,
+     "DC": 0.082444,
+     "Other": 0.036
+    },
+    "mean_fraction_observed": {
+     "T cell": 0.44831,
+     "B cell": 0.055177,
+     "Monocyte": 0.332669,
+     "NK cell": 0.136626,
+     "DC": 0.009903,
+     "Other": 0.017315
+    }
+   },
+   "pseudobulk": {
+    "n_genes": 5752,
+    "r_median": 0.5130782449777777,
+    "r_mean": 0.548128654635822,
+    "top1": 0.3333333333333333,
+    "n_top1": 2,
+    "own_rank_median": 3.5,
+    "chance_top1": 0.16666666666666666
+   },
+   "precision_coverage": {
+    "precision": 0.9222222222222223,
+    "coverage": 0.34555555555555556,
+    "density": 0.6941111111111111,
+    "n_query": 900,
+    "n_reference": 900
+   },
+   "precision_coverage_depth_matched": {
+    "precision": 0.7444444444444445,
+    "coverage": 0.3422222222222222,
+    "density": 0.4286666666666667,
+    "detected_median_before": {
+     "R1": 367.0,
+     "Q": 121.0
+    },
+    "detected_median_after": {
+     "R1": 121.0,
+     "Q": 121.0
+    }
+   },
+   "monocyte_pct": {
+    "generated_mean_over_donors": 26.522222222222226,
+    "observed_mean_over_donors": 33.2668840833307,
+    "generated_pooled_cells": 26.522222222222226
+   },
+   "median_umi_per_cell_scoring_genes": {
+    "generated_pooled": 1063.0,
+    "observed_pooled": 1030.0,
+    "generated_mean_of_donor_medians": 1062.4166259765625,
+    "observed_mean_of_donor_medians": 1105.5833333333333
+   },
+   "contrast": {
+    "mean_cp10k": {
+     "per_type": {
+      "T cell": {
+       "r_generated": -0.2777122091986991,
+       "r_broadcast": -0.136748154494899,
+       "partial_r_generated_given_broadcast": -0.2557643715600354,
+       "r_generated_vs_broadcast": 0.22600763074019473,
+       "n_genes": 3071
+      },
+      "B cell": {
+       "r_generated": -0.22205414532503903,
+       "r_broadcast": -0.09943855024335586,
+       "partial_r_generated_given_broadcast": -0.2086913743655284,
+       "r_generated_vs_broadcast": 0.17821384363942416,
+       "n_genes": 3275
+      },
+      "Monocyte": {
+       "r_generated": -0.2609013703210285,
+       "r_broadcast": -0.03619225031477173,
+       "partial_r_generated_given_broadcast": -0.2622021959594069,
+       "r_generated_vs_broadcast": 0.2982017868479014,
+       "n_genes": 3221
+      },
+      "NK cell": {
+       "r_generated": -0.2553010457499112,
+       "r_broadcast": -0.1320961900042953,
+       "partial_r_generated_given_broadcast": -0.23298568701329497,
+       "r_generated_vs_broadcast": 0.23215383893331812,
+       "n_genes": 3022
+      }
+     },
+     "mean_r_generated": -0.25399219264866946,
+     "mean_r_broadcast": -0.10111878626433048,
+     "mean_partial_r": -0.23991090722456643
+    },
+    "pseudobulk_log1p_cpm": {
+     "per_type": {
+      "T cell": {
+       "r_generated": -0.0225337891733434,
+       "r_broadcast": 0.016838325109610475,
+       "partial_r_generated_given_broadcast": -0.025853786304497754,
+       "r_generated_vs_broadcast": 0.17363255938106922,
+       "n_genes": 3071
+      },
+      "B cell": {
+       "r_generated": -0.0658745919471903,
+       "r_broadcast": -0.017254112686454763,
+       "partial_r_generated_given_broadcast": -0.06440568435397624,
+       "r_generated_vs_broadcast": 0.10719402242061334,
+       "n_genes": 3275
+      },
+      "Monocyte": {
+       "r_generated": -0.07235614767470153,
+       "r_broadcast": 0.05487205087610296,
+       "partial_r_generated_given_broadcast": -0.08594972242456485,
+       "r_generated_vs_broadcast": 0.21037204874300022,
+       "n_genes": 3221
+      },
+      "NK cell": {
+       "r_generated": 0.04357690567044159,
+       "r_broadcast": -0.012307630218407884,
+       "partial_r_generated_given_broadcast": 0.04656672117931589,
+       "r_generated_vs_broadcast": 0.1804980827789389,
+       "n_genes": 3022
+      }
+     },
+     "mean_r_generated": -0.02929690578119841,
+     "mean_r_broadcast": 0.010537158270212697,
+     "mean_partial_r": -0.03241061797593074
+    }
+   },
+   "deg_top50": {
+    "types_scored": [
+     "T cell",
+     "B cell",
+     "Monocyte",
+     "NK cell",
+     "DC"
+    ],
+    "types_not_scored": {},
+    "per_type": {
+     "T cell": {
+      "deg_scored": true,
+      "top50_overlap": 0,
+      "top50_jaccard": 0.0,
+      "obs_top_list_size": 50,
+      "gen_top_list_size": 0,
+      "top50_overlap_broadcast": 0,
+      "top50_jaccard_broadcast": 0.0,
+      "n_cells_obs": 11902,
+      "n_cells_gen": 4196
+     },
+     "B cell": {
+      "deg_scored": true,
+      "top50_overlap": 0,
+      "top50_jaccard": null,
+      "obs_top_list_size": 0,
+      "gen_top_list_size": 0,
+      "top50_overlap_broadcast": 0,
+      "top50_jaccard_broadcast": null,
+      "n_cells_obs": 1609,
+      "n_cells_gen": 210
+     },
+     "Monocyte": {
+      "deg_scored": true,
+      "top50_overlap": 0,
+      "top50_jaccard": null,
+      "obs_top_list_size": 0,
+      "gen_top_list_size": 0,
+      "top50_overlap_broadcast": 0,
+      "top50_jaccard_broadcast": null,
+      "n_cells_obs": 9804,
+      "n_cells_gen": 2387
+     },
+     "NK cell": {
+      "deg_scored": true,
+      "top50_overlap": 0,
+      "top50_jaccard": 0.0,
+      "obs_top_list_size": 3,
+      "gen_top_list_size": 0,
+      "top50_overlap_broadcast": 0,
+      "top50_jaccard_broadcast": 0.0,
+      "n_cells_obs": 4134,
+      "n_cells_gen": 1141
+     },
+     "DC": {
+      "deg_scored": true,
+      "top50_overlap": 0,
+      "top50_jaccard": 0.0,
+      "obs_top_list_size": 3,
+      "gen_top_list_size": 0,
+      "top50_overlap_broadcast": 0,
+      "top50_jaccard_broadcast": 0.0,
+      "n_cells_obs": 306,
+      "n_cells_gen": 742
+     }
+    },
+    "sum": 0,
+    "sum_4type": 0,
+    "max": 56,
+    "mean": 0.0,
+    "sum_broadcast": 0,
+    "sum_broadcast_4type": 0,
+    "broadcast_n_significant": 0,
+    "seconds": 0.02
+   },
+   "seconds": 4.7
+  }
+ },
+ "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": 5.2
+}

paper-releases/pbisc-diffusion/updates/20261005T121346Z-scsemi-cv-budget-interim/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores/ji2025_chili/scsemi_cv_20261002/fold3/README.md

@@ -0,0 +1,73 @@
+# score_ft.py output: ji2025_chili, 8 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/ji2025_chili/reference.npz` (sha 19c30f56db69), built from datadir:/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/ji2025_chili;
+  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/ji2025_chili/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 (1004 genes).
+- Scoring genes: 6k panel genes measured by both assays in every sample of the dataset (5752 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 8 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) (1200 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 1.0000, 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
+   1.0000, 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 ['TOL', 'FU1'] (n = 4) vs control groups ['BC', 'AC'] (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 5752 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/ji2025_chili/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 4 + 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/20261005T121346Z-scsemi-cv-budget-interim/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/scores/ji2025_chili/scsemi_cv_20261002/fold3/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
+scsemi__raw,T cell,mean_cp10k,3055,-0.17673093620560432,0.2089663375687252,-0.2393690269706704,0.241348141411102,4,4,4,4,14945,7221,,,,,,,,,,,
+scsemi__raw,T cell,pseudobulk_log1p_cpm,3055,0.017402038214953345,0.32657181877736113,-0.027197232346612817,0.1313202614955506,4,4,4,4,14945,7221,,,,,,,,,,,
+scsemi__raw,B cell,mean_cp10k,3282,-0.09141088929451731,0.05112929523901414,-0.10487913614253079,0.21349913776434892,4,4,4,4,747,455,,,,,,,,,,,
+scsemi__raw,B cell,pseudobulk_log1p_cpm,3282,0.10243086226919114,0.0952205726272668,0.09036568312454724,0.14037370766695262,4,4,4,4,747,455,,,,,,,,,,,
+scsemi__raw,Monocyte,mean_cp10k,3146,0.25645442491766546,0.31084121428948736,0.20790138944097755,0.20250226667716112,4,4,4,4,4157,2586,,,,,,,,,,,
+scsemi__raw,Monocyte,pseudobulk_log1p_cpm,3146,0.14623274261278793,0.34577096497019094,0.09867788125586849,0.1585216629984345,4,4,4,4,4157,2586,,,,,,,,,,,
+scsemi__raw,NK cell,mean_cp10k,3049,-0.1125465907246297,0.1175675951173203,-0.13874303389558498,0.1926767750843385,4,4,4,4,5655,1103,,,,,,,,,,,
+scsemi__raw,NK cell,pseudobulk_log1p_cpm,3049,-0.004437019441596001,0.14908893470964257,-0.02340633234580535,0.12427677095549852,4,4,4,4,5655,1103,,,,,,,,,,,
+scsemi__raw,T cell,welch_top50_deg,5752,,,,,4,4,4,4,14945,7221,True,0,0.0,50,0,83,0,0,0.0,1,1
+scsemi__raw,B cell,welch_top50_deg,5752,,,,,4,4,4,4,747,455,True,0,0.0,2,0,2,0,0,0.0,1,1
+scsemi__raw,Monocyte,welch_top50_deg,5752,,,,,4,4,4,4,4157,2586,True,0,,0,0,0,0,0,0.0,1,1
+scsemi__raw,NK cell,welch_top50_deg,5752,,,,,4,4,4,4,5655,1103,True,0,0.0,1,0,1,0,0,0.0,1,1
+scsemi__raw,DC,welch_top50_deg,5752,,,,,2,4,4,4,597,500,True,0,0.0,50,0,299,0,0,0.0,1,1