Archive 4 checkpoints (1-4)

paper-releases/pbisc-diffusion/updates/20261002T124820Z-crosscohort-liu-default/diffusion/dataset/server_1/dependency/eca339fe1e9e0d4e/FINDINGS.json

@@ -0,0 +1,1161 @@
+{
+ "bulk_material_and_protocol": {
+  "geo_source_name": [
+   "Peripheral Blood Mononuclear Cells"
+  ],
+  "geo_unsorted_or_sorted": [
+   "Unsorted"
+  ],
+  "geo_molecule": [
+   "total RNA"
+  ],
+  "geo_description": [
+   "For each sample, a sample of 100,000 cells was processed in Trizol using the miRNAeasy micro kit (Qiagen, Germantown, MD; catalogue number 217084) and standard RNA sequencing libraries were generated using Illumina Truseq library (Illuimna, San Diego, CA; 20020595) preparation kits. Sequencing results were demultiplexed and converted to FASTQ format using Illumina bcl2fastq software. The sequencing reads were adapter and quality trimmed and then aligned to the human genome using the splice-aware STAR aligner and SNP calls were generated using the previously published protocol: Nucleic Acids Research, Volume 47, Issue 21, 02 December 2019, Page e136. These SNP calls were then used along with the demuxlet software (https://github.com/statgen/demuxlet) to demultiplex the single cell multimodal RNAseq data (see supplemental demux.best files). Counts per gene were generated using htseq count (version 0.9.1) and the UCSC hg19 genome annotation."
+  ],
+  "conclusion": "PBMC bulk (not whole blood): every one of the 73 bulk GSMs has source 'Peripheral Blood Mononuclear Cells' and 'Unsorted'; 100,000 PBMCs per sample in Trizol, miRNeasy micro, Illumina TruSeq library, STAR, htseq-count 0.9.1 with the UCSC hg19 annotation. The bulk RNA of each sample was also the SNP reference used by demuxlet to assign single cells to subjects."
+ },
+ "sc_chemistry": {
+  "geo_statement": [
+   "10x 5\u2019 Chromium Single Cell Immune Profiling Next GEM v1.1 chemistry"
+  ],
+  "geo_processing": [
+   "Genome_build: hg19 (cellranger version 1.2.0, https://support.10xgenomics.com/single-cell-gene-expression/software/release-notes/build#hg19_1.2.0)",
+   "mapping, CellRanger 3.1.0 (10x Genomics)",
+   "sample demultiplexing: popscle demuxlet (https://github.com/statgen/popscle)",
+   "secondary analysis: Seurat 3.1.0 (R package)"
+  ],
+  "first_sorted_barcode_by_batch": {
+   "B1": "AAACCTGAGAAACCTA-1",
+   "B2": "AAACCTGAGAAACCTA-1",
+   "B3": "AAACCTGAGAAACCAT-1"
+  },
+  "barcode_suffixes": [
+   "1"
+  ],
+  "conclusion": "10x 5' Next GEM v1.1 CITE-seq (GEX + 192 ADT + HTO + TCR); Cell Ranger 3.1.0 with the hg19-1.2.0 reference (32,738 genes; the Seurat RNA rownames equal this table, see reference_crosswalk). Only the RNA counts are used."
+ },
+ "author_filtering": {
+  "object_cells": 411902,
+  "object_min_nFeature_RNA": 201,
+  "object_max_nFeature_RNA": 3991,
+  "object_min_nCount_RNA": 244.0,
+  "object_max_percent_mito": 29.982046678635598,
+  "selected_cells": 224919,
+  "selected_min_nFeature_RNA": 201,
+  "selected_min_nCount_RNA": 244.0,
+  "selected_max_percent_mito": 29.982046678635598,
+  "selected_failing_conservative_rule": {
+   "genes_lt_200": 0,
+   "umis_lt_500": 2374,
+   "mito_gt_20pct": 3932
+  },
+  "author_cluster_doublets_in_selected": {
+   "Dblt": 4802
+  },
+  "conclusion": "The author object is already filtered (every cell has >= 200 genes, author percent.mito is a fraction, max ~0.3); it still holds HTO doublets/negatives (all demuxlet singlets) and the sorted fractions. The ImMicro rule (>=200 genes, >=500 UMIs, <=20 % mito) is applied on top. Author cluster-level 'Dblt' cells are not an HTO call and are kept (counted here)."
+ },
+ "cell_selection": {
+  "by_batchsort_and_hto": {
+   "Doublet": {
+    "B1Sort": 10732,
+    "B1UnSort": 358,
+    "B2Sort": 596,
+    "B2UnSort": 98,
+    "B3Sort": 0,
+    "B3UnSort": 1426
+   },
+   "Negative": {
+    "B1Sort": 0,
+    "B1UnSort": 0,
+    "B2Sort": 0,
+    "B2UnSort": 15723,
+    "B3Sort": 148,
+    "B3UnSort": 17932
+   },
+   "Singlet": {
+    "B1Sort": 23605,
+    "B1UnSort": 6982,
+    "B2Sort": 49582,
+    "B2UnSort": 109828,
+    "B3Sort": 66783,
+    "B3UnSort": 108109
+   }
+  },
+  "why_unsorted_only": "The bulk is unsorted PBMC. Batch 1 lanes pool FACS-sorted B, T, CD14+ monocyte and CD14- innate cells with unsorted PBMC 1:1:1:1:1 (sorted and unsorted fractions carry different hashtags), and batches 2-3 have separate lanes of sorted non-naive T and B cells. Sorted cells would distort the composition, so only Sorted == 'N' cells are used. Consequence: batch-1 samples keep only ~120-490 cells each.",
+  "hto_negative_composition_b2_b3_unsorted": {
+   "total_variation_singlet_vs_negative": 0.23350874938883667,
+   "cell_types_more_frequent_among_negatives": {
+    "NK_CD16hi": 0.0126,
+    "CD8_Mem": 0.017,
+    "CD8_Naive": 0.0228,
+    "CD4_Mem": 0.0699,
+    "CD4_Naive": 0.089
+   },
+   "cell_types_less_frequent_among_negatives": {
+    "Mono_Classical": -0.2013,
+    "B_Mem": -0.0074,
+    "MAIT": -0.0056,
+    "gated_out": -0.0043,
+    "B_Naive": -0.0035
+   },
+   "note": "HTO 'Negative' cells concentrate in weakly stained pools (B2 HTO3, B3 HTO5: 30-70 % of a sample). They are dropped per the task rule (their timepoint was assigned by the authors from the max, below-threshold hashtag); this total-variation distance measures how much their author cell-type mix differs from the singlets'."
+  },
+  "paired_rows_dropped": {
+   "sorted_cells": 135870,
+   "hto_doublets": 1748,
+   "hto_negatives": 31147,
+   "hto_negative_fraction_of_unsorted_max": 0.689406430338005,
+   "rows_with_negative_fraction_gt_0.3": 14
+  }
+ },
+ "sc_source_choice": {
+  "used": "author Seurat object GSE161918_AllBatches_SeuratObj.rds.gz (RNA counts slot), exported unchanged in R",
+  "cellxgene_h5ad": {
+   "73024e1c-c5e4-48d2-9ca4-49d2f845c8b9.h5ad": {
+    "cells": 125117,
+    "genes": 30000,
+    "cells_in_author_object": 125117,
+    "of_which_sorted": 36246,
+    "of_which_hto_not_singlet": 15273,
+    "obs_has_sorted_or_hto_columns": false
+   },
+   "e6ef9f09-bf7f-49bf-900c-457c02675411.h5ad": {
+    "cells": 246964,
+    "genes": 30000,
+    "cells_in_author_object": 246964,
+    "of_which_sorted": 100742,
+    "of_which_hto_not_singlet": 27042,
+    "obs_has_sorted_or_hto_columns": false
+   }
+  },
+  "corpus_obs": {
+   "202_B_GSE161918_Time-resolved_Systems_Immunology_Reveals_a_Late_Juncture_Linked_to_Fatal_COVID-1__73024e1c-c5e4-48d2-9ca4-49d2f845c8b9__dec6d48a1403.parquet": {
+    "rows": 125117,
+    "has_barcode_column": false,
+    "qc_status": {
+     "pass": 125117
+    },
+    "donor_timepoints": 67
+   },
+   "202_B_GSE161918_Time-resolved_Systems_Immunology_Reveals_a_Late_Juncture_Linked_to_Fatal_COVID-1__e6ef9f09-bf7f-49bf-900c-457c02675411__e0d44f5435c7.parquet": {
+    "rows": 246790,
+    "has_barcode_column": false,
+    "qc_status": {
+     "pass": 246790
+    },
+    "donor_timepoints": 67
+   }
+  },
+  "spot_check_cellxgene_raw_vs_our_counts": [
+   {
+    "file": "73024e1c-c5e4-48d2-9ca4-49d2f845c8b9.h5ad",
+    "cell": "TAGTTGGTCAGTGCAT-5_1",
+    "our_umis": 3613,
+    "our_genes_in_cxg_var": 1466,
+    "equal_on_shared_ids": true,
+    "cxg_ids_not_in_hg19_table": [],
+    "equal_after_symbol_match": true,
+    "cxg_umis": 3519.0,
+    "our_umis_on_cxg_genes": 3519
+   },
+   {
+    "file": "73024e1c-c5e4-48d2-9ca4-49d2f845c8b9.h5ad",
+    "cell": "AGTTGGTGTGTAATGA-7_1",
+    "our_umis": 4158,
+    "our_genes_in_cxg_var": 1467,
+    "equal_on_shared_ids": true,
+    "cxg_ids_not_in_hg19_table": [
+     {
+      "cxg_id": "ENSG00000291237",
+      "cxg_feature_name": "SOD2",
+      "cxg_count": 1.0,
+      "our_count_same_symbol": 1
+     }
+    ],
+    "equal_after_symbol_match": true,
+    "cxg_umis": 4061.0,
+    "our_umis_on_cxg_genes": 4060
+   },
+   {
+    "file": "73024e1c-c5e4-48d2-9ca4-49d2f845c8b9.h5ad",
+    "cell": "GACCAATAGGCGTACA-16_4",
+    "our_umis": 1806,
+    "our_genes_in_cxg_var": 898,
+    "equal_on_shared_ids": true,
+    "cxg_ids_not_in_hg19_table": [],
+    "equal_after_symbol_match": true,
+    "cxg_umis": 1776.0,
+    "our_umis_on_cxg_genes": 1776
+   },
+   {
+    "file": "73024e1c-c5e4-48d2-9ca4-49d2f845c8b9.h5ad",
+    "cell": "CGTTAGAGTGGTCCGT-8_1",
+    "our_umis": 3314,
+    "our_genes_in_cxg_var": 1404,
+    "equal_on_shared_ids": true,
+    "cxg_ids_not_in_hg19_table": [],
+    "equal_after_symbol_match": true,
+    "cxg_umis": 3234.0,
+    "our_umis_on_cxg_genes": 3234
+   },
+   {
+    "file": "e6ef9f09-bf7f-49bf-900c-457c02675411.h5ad",
+    "cell": "CCCATACAGGATCGCA-03_4",
+    "our_umis": 522,
+    "our_genes_in_cxg_var": 280,
+    "equal_on_shared_ids": true,
+    "cxg_ids_not_in_hg19_table": [],
+    "equal_after_symbol_match": true,
+    "cxg_umis": 521.0,
+    "our_umis_on_cxg_genes": 521
+   },
+   {
+    "file": "e6ef9f09-bf7f-49bf-900c-457c02675411.h5ad",
+    "cell": "TGCGGGTCAGCTTAAC-03_2",
+    "our_umis": 4639,
+    "our_genes_in_cxg_var": 1483,
+    "equal_on_shared_ids": true,
+    "cxg_ids_not_in_hg19_table": [],
+    "equal_after_symbol_match": true,
+    "cxg_umis": 4546.0,
+    "our_umis_on_cxg_genes": 4546
+   },
+   {
+    "file": "e6ef9f09-bf7f-49bf-900c-457c02675411.h5ad",
+    "cell": "ATTATCCGTGCAGACA-4_1",
+    "our_umis": 3743,
+    "our_genes_in_cxg_var": 1492,
+    "equal_on_shared_ids": true,
+    "cxg_ids_not_in_hg19_table": [
+     {
+      "cxg_id": "ENSG00000310560",
+      "cxg_feature_name": "PAXX",
+      "cxg_count": 1.0,
+      "our_count_same_symbol": null
+     }
+    ],
+    "equal_after_symbol_match": false,
+    "cxg_umis": 3677.0,
+    "our_umis_on_cxg_genes": 3676
+   },
+   {
+    "file": "e6ef9f09-bf7f-49bf-900c-457c02675411.h5ad",
+    "cell": "TTTGCGCGTGGGTATG-11_4",
+    "our_umis": 2327,
+    "our_genes_in_cxg_var": 932,
+    "equal_on_shared_ids": true,
+    "cxg_ids_not_in_hg19_table": [],
+    "equal_after_symbol_match": true,
+    "cxg_umis": 2292.0,
+    "our_umis_on_cxg_genes": 2292
+   }
+  ],
+  "conclusion": "Not the corpus copy: the CELLxGENE files it derives from hold 372,081 of the object's 411,902 cells on a 30,000-gene Ensembl axis, include sorted-fraction cells and HTO doublets/negatives, but carry no Sorted / HTO-classification columns, so the unsorted HTO-singlet subset needed to match an unsorted PBMC bulk cannot be selected there; the corpus parquet has no barcodes and its own QC. The spot-check (8 random cells) compares the exported counts with CELLxGENE raw/X: equal on every shared Ensembl ID, and the only extra CELLxGENE entries are genes it labels with a newer Ensembl ID (e.g. SOD2 ENSG00000291237 vs hg19 ENSG00000112096), equal when matched by symbol; CELLxGENE drops the remaining hg19 genes (30,000 kept), which is why our cells have more UMIs."
+ },
+ "gene_coverage": {
+  "bulk_canonical_measured": 19060,
+  "sc_canonical_measured": 18976,
+  "pair_valid": 18553,
+  "panel6k_bulk": 5753,
+  "panel6k_sc": 5806,
+  "panel6k_both": 5681,
+  "bulk_mapping_summary": {
+   "schema": "strict_gene_mapping_summary_v1",
+   "source_features": 55765,
+   "canonical_genes": 20097,
+   "mapped_source_features": 19070,
+   "unique_canonical_features_present": 19060,
+   "canonical_coverage_fraction": 0.9484002587450864,
+   "canonical_features_absent": 1037,
+   "ambiguous_features": 59,
+   "unmapped_or_excluded_features": 36695,
+   "version_stripped_features": 0,
+   "source_id_symbol_conflicts": 0,
+   "duplicate_source_id_rows": 0,
+   "duplicate_target_groups": 10,
+   "duplicate_target_rows": 20,
+   "mitochondrial_features_source_axis": 37,
+   "ribosomal_features_source_axis": 83,
+   "source_qc_annotation_unresolved_features": 22641,
+   "status_counts": {
+    "ambiguous_symbol": 59,
+    "excluded_legacy_symbol_annotation_only": 15,
+    "excluded_outside_canonical": 14039,
+    "excluded_withdrawn_symbol": 448,
+    "mapped_alias_symbol": 23,
+    "mapped_approved_symbol": 19,
+    "mapped_canonical_symbol": 18396,
+    "mapped_previous_symbol": 632,
+    "unmapped_symbol": 22134
+   },
+   "presence_mask_semantics": "source_feature_table_membership;not_detection_or_proof_of_assay_coverage",
+   "absent_feature_policy": "masked_missing;never_observed_zero",
+   "duplicate_policy": "flag_only;no_expression_aggregation_performed",
+   "expression_transformation": "none",
+   "canonical_file_sha256": "64f4e30e17763bef37793e7229edb6caa0a7ffe3f3bfa5a08e1d4ca8ea7ea1dc",
+   "canonical_identity_sha256": "13fd2fb93db9578df24e4ac92c430b229f87825f269b404c918e5e673384fe14"
+  },
+  "bulk_duplicate_target_groups_summed": 10,
+  "bulk_has_ribosomal_and_mito_genes": true,
+  "note": "Unlike Giroux, the author bulk table keeps ribosomal-protein and mitochondrial genes."
+ },
+ "bulk_depth": {
+  "assigned_reads_paired_columns": {
+   "min": 11314446.0,
+   "median": 18300389.0,
+   "max": 28220948.0
+  },
+  "by_table": {
+   "B1_bulkRNAseq_countmtx.csv": 13311038.0,
+   "B2_B3_bulkRNAseq_countmtx.csv": 19730193.0
+  }
+ },
+ "pairing": {
+  "bulk_gsms": 73,
+  "bulk_columns": 73,
+  "bulk_columns_by_table": {
+   "B2_B3_bulkRNAseq_countmtx.csv": 50,
+   "B1_bulkRNAseq_countmtx.csv": 23
+  },
+  "bulk_columns_by_kind": {
+   "COVID": 54,
+   "HC": 11,
+   "technical_control": 7,
+   "author_excluded": 1
+  },
+  "bulk_column_corrections": [
+   {
+    "bulk_column": "PHGR0000074_T0",
+    "column_key": "HGR0000074_T0",
+    "bulk_gsm": "GSM4929238",
+    "bulk_title": "Batch 1 bulk RNAseq HGR0000074"
+   }
+  ],
+  "bulk_timepoint_conflicts": [
+   {
+    "bulk_column": "HGR0000089_T1",
+    "bulk_gsm": "GSM4929245",
+    "bulk_title": "Batch 1 bulk RNAseq HGR0000089",
+    "timepoint_conflict": "column name T1 vs GEO characteristic 'Timepoint 0'"
+   }
+  ],
+  "bulk_material": "GEO source 'Peripheral Blood Mononuclear Cells', characteristic 'Unsorted' for all 73 bulk GSMs",
+  "sc_cells_object": 411902,
+  "sc_cells_unsorted": 260456,
+  "sc_cells_selected": 224919,
+  "sc_cells_by_batchsort_hto": {
+   "Doublet": {
+    "B1Sort": 10732,
+    "B1UnSort": 358,
+    "B2Sort": 596,
+    "B2UnSort": 98,
+    "B3Sort": 0,
+    "B3UnSort": 1426
+   },
+   "Negative": {
+    "B1Sort": 0,
+    "B1UnSort": 0,
+    "B2Sort": 0,
+    "B2UnSort": 15723,
+    "B3Sort": 148,
+    "B3UnSort": 17932
+   },
+   "Singlet": {
+    "B1Sort": 23605,
+    "B1UnSort": 6982,
+    "B2Sort": 49582,
+    "B2UnSort": 109828,
+    "B3Sort": 66783,
+    "B3UnSort": 108109
+   }
+  },
+  "sc_droplet_type": {
+   "SNG": 411902
+  },
+  "sc_samples_unsorted": 70,
+  "sc_samples_by_kind": {
+   "COVID": 54,
+   "HC": 11,
+   "technical_control": 5
+  },
+  "hto_metadata_cross_check": {
+   "cells": 411902,
+   "cells_key_found": 411902,
+   "timepoint_agrees": 411902,
+   "timepoint_agrees_selected": 224919,
+   "batch1_sorted_agrees": 41677,
+   "batch1_cells": 41677,
+   "selected_cells": 224919,
+   "selected_key_not_found": {},
+   "selected_timepoint_disagrees": {},
+   "key": "Batch + '|' + Donor + '_' + autoHashcalls (the hashtag that defines the author Sample)"
+  },
+  "pairs_total": 65,
+  "pairs_subjects": 44,
+  "pairs_by_group": {
+   "COVID": 54,
+   "HC": 11
+  },
+  "pairs_by_timepoint": {
+   "T0": 28,
+   "T1": 19,
+   "HC": 11,
+   "T2": 4,
+   "T3": 3
+  },
+  "pairs_by_batch": {
+   "B3": 23,
+   "B1": 22,
+   "B2": 20
+  },
+  "pairs_by_id_match": {
+   "exact_subject_timepoint": 63,
+   "subject_timepoint; bulk column name corrected": 1,
+   "subject_timepoint; GEO bulk timepoint characteristic differs": 1
+  },
+  "subjects_by_group": {
+   "COVID": 33,
+   "HC": 11
+  },
+  "sc_without_bulk": [
+   {
+    "sc_sample": "B1_CHI014_HC",
+    "kind": "technical_control",
+    "reason": "technical control (CHI014 / HD buffy-coat test donor); excluded by task rule"
+   },
+   {
+    "sc_sample": "B2_CHI014_HC",
+    "kind": "technical_control",
+    "reason": "technical control (CHI014 / HD buffy-coat test donor); excluded by task rule"
+   },
+   {
+    "sc_sample": "B2_HDML_PBMC_HC",
+    "kind": "technical_control",
+    "reason": "technical control (CHI014 / HD buffy-coat test donor); excluded by task rule"
+   },
+   {
+    "sc_sample": "B2_HDVO_PBMC_HC",
+    "kind": "technical_control",
+    "reason": "technical control (CHI014 / HD buffy-coat test donor); excluded by task rule"
+   },
+   {
+    "sc_sample": "B3_CHI014_HC",
+    "kind": "technical_control",
+    "reason": "technical control (CHI014 / HD buffy-coat test donor); excluded by task rule"
+   }
+  ],
+  "bulk_without_sc": [
+   {
+    "bulk_column": "CHI-014",
+    "bulk_gsm": "GSM4929236",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "CHI014_B2",
+    "bulk_gsm": "GSM4929256",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "CHI014_B3",
+    "bulk_gsm": "GSM4929282",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "HD_ML_bc",
+    "bulk_gsm": "GSM4929257",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "HD_ML_PBMC",
+    "bulk_gsm": "GSM4929258",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "HD_VO_bc",
+    "bulk_gsm": "GSM4929259",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "HD_VO_PBMC",
+    "bulk_gsm": "GSM4929260",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "SHD8",
+    "bulk_gsm": "GSM4929281",
+    "kind": "author_excluded",
+    "unpaired_reason": "GEO: 'Healthy Control (excluded due to abnormal lymphoma-like B cells found upon analysis)'; the author Seurat object holds 0 cells of this subject"
+   }
+  ],
+  "rule": "A pair is one author bulk count-table column and one author sc sample (unsorted fraction) with the same subject and the same timepoint label (T0-T3, or HC). Bulk column <-> GEO GSM by the GEO title name within the same count table (1:1, all 73). Column subject/timepoint from the column name (one typo corrected: PHGR0000074_T0); the sc subject/timepoint are the author metadata Subject and Timepoint (hashtag-defined; cross-checked against the GEO HTO/timepoint table). Technical controls (CHI014, HD_ML/HD_VO) are not paired. No expression, age-sex or order matching creates a pair."
+ },
+ "identity": {
+  "gene_positions_sha256_ok": true,
+  "panel_gene_ids_match_selected_genes_csv": true,
+  "panel_gene_ids_match_giroux2022_cells_panel6k": true,
+  "bulk_integer_nonnegative_inside_mask": true,
+  "bulk_raw_nan_exactly_outside_bulk_valid": true,
+  "pseudo_raw_sum_equals_sum_of_retained_cells_full_axis": true,
+  "cells_panel_sums_equal_pseudo_raw_sum_positions_exactly": true,
+  "cells_per_row_equal_qc_cell_counts": true,
+  "no_sc_mass_outside_sc_valid": true,
+  "one_sample_id_per_library": true,
+  "sc_cells_are_unsorted_hto_singlet_demuxlet_singlet": true,
+  "sex_concordant_pairs": "63/65",
+  "expression_best_match_is_own_pair": "25/65",
+  "expression_best_match_is_own_pair_centred": "40/65",
+  "expression_reverse_best_match_is_own_pair": "17/65",
+  "expression_residualised_k5_own_subject_top1_bulk_to_sc": "62/65",
+  "expression_residualised_k3_own_subject_top1_bulk_to_sc": "61/65",
+  "expression_residualised_k10_own_subject_top1_bulk_to_sc": "60/65",
+  "expression_residualised_k5_own_subject_top1_sc_to_bulk": "63/65",
+  "expression_residualised_k3_own_subject_top1_sc_to_bulk": "64/65",
+  "expression_residualised_k10_own_subject_top1_sc_to_bulk": "62/65",
+  "expression_residualised_k5_own_library_top1_bulk_to_sc": "61/65",
+  "expression_residualised_k3_own_library_top1_bulk_to_sc": "59/65",
+  "expression_residualised_k10_own_library_top1_bulk_to_sc": "60/65",
+  "expression_residualised_k5_own_timepoint_best_within_subject_bulk_to_sc_multi_timepoint_rows": "37/38",
+  "expression_residualised_k5_own_timepoint_best_within_subject_sc_to_bulk_multi_timepoint_rows": "37/38",
+  "expression_centred_own_timepoint_best_within_subject_bulk_to_sc_multi_timepoint_rows": "30/38",
+  "expression_centred_own_timepoint_best_within_subject_sc_to_bulk_multi_timepoint_rows": "34/38",
+  "isg_peak_timepoint_same_in_both_assays_multi_timepoint_subjects": "16/17",
+  "identity_check_pass": "61/65",
+  "identity_check_fail": [
+   "HGR0000079_T0",
+   "HGR0000079_T1",
+   "HGR0000110_T0"
+  ],
+  "identity_check_borderline": [
+   "HGR0000078_T0"
+  ],
+  "rows_below_200_cells": [
+   "HGR0000078_T0",
+   "HGR0000078_T1",
+   "HGR0000079_T0",
+   "HGR0000079_T1",
+   "HGR0000083_T0",
+   "HGR0000093_T1",
+   "HGR0000110_T0"
+  ],
+  "usable_rows_identity_pass_and_ge_200_cells": 58
+ },
+ "usable_rows": {
+  "rows": 65,
+  "usable": 58,
+  "usable_by_group": {
+   "COVID": 47,
+   "HC": 11
+  },
+  "usable_donors_by_group": {
+   "COVID": 30,
+   "HC": 11
+  },
+  "usable_by_timepoint": {
+   "HC": 11,
+   "T0": 24,
+   "T1": 16,
+   "T2": 4,
+   "T3": 3
+  },
+  "excluded": [
+   "HGR0000078_T0",
+   "HGR0000078_T1",
+   "HGR0000079_T0",
+   "HGR0000079_T1",
+   "HGR0000083_T0",
+   "HGR0000093_T1",
+   "HGR0000110_T0"
+  ]
+ },
+ "sanity": {
+  "definition": "per usable row (folds.csv fold >= 0): Pearson r of log1p(CPM) bulk vs log1p(CPM) sc pseudobulk (pseudo_raw_sum, retained cells) over the row's pair_valid genes; CPM over the same pair_valid genes in both assays; NaN bulk outside bulk_valid is never used",
+  "datasets": {
+   "liu2021": {
+    "rows": 65,
+    "usable_rows": 58,
+    "pearson_log1p_cpm_bulk_vs_pseudobulk": {
+     "n": 58,
+     "median": 0.8978402352314852,
+     "min": 0.8691990435358162,
+     "max": 0.9118583780172661,
+     "q25": 0.8924504463995364,
+     "q75": 0.9009274993985295
+    },
+    "pair_valid_genes_per_row": {
+     "n": 58,
+     "median": 18553.0,
+     "min": 18553.0,
+     "max": 18553.0,
+     "q25": 18553.0,
+     "q75": 18553.0
+    },
+    "retained_cells_per_usable_row": {
+     "n": 58,
+     "median": 2932.0,
+     "min": 240.0,
+     "max": 8731.0,
+     "q25": 565.25,
+     "q75": 5183.25
+    },
+    "retained_cells_total_usable_rows": 190468,
+    "umi_per_cell_source_median_all_cells_usable_rows": 4617.0,
+    "umi_per_cell_canonical_median_all_cells_usable_rows": 4388.0,
+    "umi_per_cell_source_median_of_row_medians": 4540.5,
+    "per_row": [
+     {
+      "sample_id": "AA220014_HC",
+      "batch": "B1",
+      "r": 0.8925,
+      "cells": 419,
+      "umi_median": 4706.0
+     },
+     {
+      "sample_id": "AA220534_HC",
+      "batch": "B1",
+      "r": 0.8692,
+      "cells": 353,
+      "umi_median": 4198.0
+     },
+     {
+      "sample_id": "AA220907_HC",
+      "batch": "B1",
+      "r": 0.8733,
+      "cells": 381,
+      "umi_median": 4267.0
+     },
+     {
+      "sample_id": "HGR0000071_T0",
+      "batch": "B1",
+      "r": 0.8907,
+      "cells": 341,
+      "umi_median": 3291.0
+     },
+     {
+      "sample_id": "HGR0000074_T0",
+      "batch": "B1",
+      "r": 0.8865,
+      "cells": 335,
+      "umi_median": 3248.0
+     },
+     {
+      "sample_id": "HGR0000083_T1",
+      "batch": "B1",
+      "r": 0.8976,
+      "cells": 299,
+      "umi_median": 4265.0
+     },
+     {
+      "sample_id": "HGR0000089_T1",
+      "batch": "B1",
+      "r": 0.8997,
+      "cells": 358,
+      "umi_median": 3784.0
+     },
+     {
+      "sample_id": "HGR0000091_T0",
+      "batch": "B1",
+      "r": 0.8925,
+      "cells": 362,
+      "umi_median": 3991.5
+     },
+     {
+      "sample_id": "HGR0000093_T0",
+      "batch": "B1",
+      "r": 0.873,
+      "cells": 451,
+      "umi_median": 3989.0
+     },
+     {
+      "sample_id": "HGR0000098_T0",
+      "batch": "B1",
+      "r": 0.8849,
+      "cells": 390,
+      "umi_median": 3908.5
+     },
+     {
+      "sample_id": "HGR0000136_T0",
+      "batch": "B1",
+      "r": 0.8824,
+      "cells": 317,
+      "umi_median": 4242.0
+     },
+     {
+      "sample_id": "HGR0000137_T0",
+      "batch": "B1",
+      "r": 0.8853,
+      "cells": 314,
+      "umi_median": 3785.0
+     },
+     {
+      "sample_id": "HGR0000141_T0",
+      "batch": "B1",
+      "r": 0.8937,
+      "cells": 240,
+      "umi_median": 3716.0
+     },
+     {
+      "sample_id": "HGR0000143_T0",
+      "batch": "B1",
+      "r": 0.877,
+      "cells": 280,
+      "umi_median": 3165.0
+     },
+     {
+      "sample_id": "IM000040_HC",
+      "batch": "B1",
+      "r": 0.887,
+      "cells": 373,
+      "umi_median": 4192.0
+     },
+     {
+      "sample_id": "HGR0000042_T0",
+      "batch": "B2",
+      "r": 0.9009,
+      "cells": 2213,
+      "umi_median": 4930.0
+     },
+     {
+      "sample_id": "HGR0000042_T1",
+      "batch": "B2",
+      "r": 0.8993,
+      "cells": 4514,
+      "umi_median": 4559.0
+     },
+     {
+      "sample_id": "HGR0000069_T0",
+      "batch": "B2",
+      "r": 0.8927,
+      "cells": 4419,
+      "umi_median": 4218.0
+     },
+     {
+      "sample_id": "HGR0000069_T1",
+      "batch": "B2",
+      "r": 0.8987,
+      "cells": 4418,
+      "umi_median": 4458.5
+     },
+     {
+      "sample_id": "HGR0000069_T2",
+      "batch": "B2",
+      "r": 0.8992,
+      "cells": 2351,
+      "umi_median": 5119.0
+     },
+     {
+      "sample_id": "HGR0000092_T0",
+      "batch": "B2",
+      "r": 0.8981,
+      "cells": 5316,
+      "umi_median": 4391.0
+     },
+     {
+      "sample_id": "HGR0000092_T1",
+      "batch": "B2",
+      "r": 0.9002,
+      "cells": 2917,
+      "umi_median": 5662.0
+     },
+     {
+      "sample_id": "HGR0000094_T0",
+      "batch": "B2",
+      "r": 0.8967,
+      "cells": 3045,
+      "umi_median": 4099.0
+     },
+     {
+      "sample_id": "HGR0000094_T1",
+      "batch": "B2",
+      "r": 0.8925,
+      "cells": 908,
+      "umi_median": 4851.5
+     },
+     {
+      "sample_id": "HGR0000115_T1",
+      "batch": "B2",
+      "r": 0.8875,
+      "cells": 5186,
+      "umi_median": 4789.0
+     },
+     {
+      "sample_id": "HGR0000124_T0",
+      "batch": "B2",
+      "r": 0.8989,
+      "cells": 8731,
+      "umi_median": 4498.0
+     },
+     {
+      "sample_id": "HGR0000124_T1",
+      "batch": "B2",
+      "r": 0.9052,
+      "cells": 2515,
+      "umi_median": 5509.0
+     },
+     {
+      "sample_id": "HGR0000144_T0",
+      "batch": "B2",
+      "r": 0.9003,
+      "cells": 2243,
+      "umi_median": 5084.0
+     },
+     {
+      "sample_id": "HGR0000144_T2",
+      "batch": "B2",
+      "r": 0.8904,
+      "cells": 6257,
+      "umi_median": 4849.0
+     },
+     {
+      "sample_id": "HGR0000391_T0",
+      "batch": "B2",
+      "r": 0.8976,
+      "cells": 6728,
+      "umi_median": 4004.0
+     },
+     {
+      "sample_id": "HGR0000395_T0",
+      "batch": "B2",
+      "r": 0.9022,
+      "cells": 2786,
+      "umi_median": 5165.5
+     },
+     {
+      "sample_id": "HGR0000418_T1",
+      "batch": "B2",
+      "r": 0.8966,
+      "cells": 6151,
+      "umi_median": 4896.0
+     },
+     {
+      "sample_id": "SHD2_HC",
+      "batch": "B2",
+      "r": 0.8968,
+      "cells": 3484,
+      "umi_median": 5657.5
+     },
+     {
+      "sample_id": "SHD4_HC",
+      "batch": "B2",
+      "r": 0.896,
+      "cells": 7126,
+      "umi_median": 4820.0
+     },
+     {
+      "sample_id": "SHD7_HC",
+      "batch": "B2",
+      "r": 0.9006,
+      "cells": 6375,
+      "umi_median": 4909.0
+     },
+     {
+      "sample_id": "HGR0000051_T0",
+      "batch": "B3",
+      "r": 0.9103,
+      "cells": 2947,
+      "umi_median": 3959.0
+     },
+     {
+      "sample_id": "HGR0000051_T1",
+      "batch": "B3",
+      "r": 0.8995,
+      "cells": 4888,
+      "umi_median": 4153.5
+     },
+     {
+      "sample_id": "HGR0000051_T2",
+      "batch": "B3",
+      "r": 0.8919,
+      "cells": 2109,
+      "umi_median": 5389.0
+     },
+     {
+      "sample_id": "HGR0000051_T3",
+      "batch": "B3",
+      "r": 0.9038,
+      "cells": 5007,
+      "umi_median": 4669.0
+     },
+     {
+      "sample_id": "HGR0000101_T0",
+      "batch": "B3",
+      "r": 0.9009,
+      "cells": 920,
+      "umi_median": 4542.0
+     },
+     {
+      "sample_id": "HGR0000101_T1",
+      "batch": "B3",
+      "r": 0.8981,
+      "cells": 4781,
+      "umi_median": 4493.0
+     },
+     {
+      "sample_id": "HGR0000101_T3",
+      "batch": "B3",
+      "r": 0.9054,
+      "cells": 3855,
+      "umi_median": 4569.0
+     },
+     {
+      "sample_id": "HGR0000102_T0",
+      "batch": "B3",
+      "r": 0.9028,
+      "cells": 4533,
+      "umi_median": 4505.0
+     },
+     {
+      "sample_id": "HGR0000102_T1",
+      "batch": "B3",
+      "r": 0.905,
+      "cells": 5393,
+      "umi_median": 4436.0
+     },
+     {
+      "sample_id": "HGR0000134_T0",
+      "batch": "B3",
+      "r": 0.902,
+      "cells": 5941,
+      "umi_median": 4826.0
+     },
+     {
+      "sample_id": "HGR0000134_T1",
+      "batch": "B3",
+      "r": 0.9048,
+      "cells": 2705,
+      "umi_median": 4892.0
+     },
+     {
+      "sample_id": "HGR0000135_T0",
+      "batch": "B3",
+      "r": 0.8996,
+      "cells": 5972,
+      "umi_median": 3775.5
+     },
+     {
+      "sample_id": "HGR0000135_T1",
+      "batch": "B3",
+      "r": 0.9119,
+      "cells": 4829,
+      "umi_median": 4561.0
+     },
+     {
+      "sample_id": "HGR0000142_T0",
+      "batch": "B3",
+      "r": 0.9048,
+      "cells": 2766,
+      "umi_median": 4355.5
+     },
+     {
+      "sample_id": "HGR0000392_T0",
+      "batch": "B3",
+      "r": 0.9044,
+      "cells": 5261,
+      "umi_median": 4070.0
+     },
+     {
+      "sample_id": "HGR0000392_T3",
+      "batch": "B3",
+      "r": 0.8974,
+      "cells": 1413,
+      "umi_median": 4695.0
+     },
+     {
+      "sample_id": "HGR0000429_T1",
+      "batch": "B3",
+      "r": 0.8951,
+      "cells": 7249,
+      "umi_median": 4539.0
+     },
+     {
+      "sample_id": "HGR0000429_T2",
+      "batch": "B3",
+      "r": 0.9043,
+      "cells": 4895,
+      "umi_median": 4863.0
+     },
+     {
+      "sample_id": "HGR0000430_T1",
+      "batch": "B3",
+      "r": 0.9044,
+      "cells": 1875,
+      "umi_median": 5052.0
+     },
+     {
+      "sample_id": "SHD1_HC",
+      "batch": "B3",
+      "r": 0.8898,
+      "cells": 6174,
+      "umi_median": 4837.0
+     },
+     {
+      "sample_id": "SHD3_HC",
+      "batch": "B3",
+      "r": 0.898,
+      "cells": 5175,
+      "umi_median": 4765.0
+     },
+     {
+      "sample_id": "SHD5_HC",
+      "batch": "B3",
+      "r": 0.8952,
+      "cells": 6397,
+      "umi_median": 5158.0
+     },
+     {
+      "sample_id": "SHD6_HC",
+      "batch": "B3",
+      "r": 0.8966,
+      "cells": 2487,
+      "umi_median": 5200.0
+     }
+    ],
+    "by_batch": {
+     "B1": {
+      "rows": 15,
+      "r_median": 0.8865,
+      "cells_median": 353.0,
+      "umi_median_of_row_medians": 3989.0
+     },
+     "B2": {
+      "rows": 20,
+      "r_median": 0.8984000000000001,
+      "cells_median": 4418.5,
+      "umi_median_of_row_medians": 4850.25
+     },
+     "B3": {
+      "rows": 23,
+      "r_median": 0.902,
+      "cells_median": 4829.0,
+      "umi_median_of_row_medians": 4569.0
+     }
+    }
+   },
+   "giroux2022": {
+    "rows": 26,
+    "usable_rows": 24,
+    "pearson_log1p_cpm_bulk_vs_pseudobulk": {
+     "n": 24,
+     "median": 0.8991432339915457,
+     "min": 0.8579903693274028,
+     "max": 0.912171011362672,
+     "q25": 0.8893585269254364,
+     "q75": 0.90455429225548
+    },
+    "pair_valid_genes_per_row": {
+     "n": 24,
+     "median": 18280.0,
+     "min": 18280.0,
+     "max": 18280.0,
+     "q25": 18280.0,
+     "q75": 18280.0
+    },
+    "retained_cells_per_usable_row": {
+     "n": 24,
+     "median": 10958.0,
+     "min": 5914.0,
+     "max": 15650.0,
+     "q25": 9059.5,
+     "q75": 12024.25
+    },
+    "retained_cells_total_usable_rows": 254250,
+    "umi_per_cell_source_median_all_cells_usable_rows": 3318.0,
+    "umi_per_cell_canonical_median_all_cells_usable_rows": 3179.0,
+    "umi_per_cell_source_median_of_row_medians": 3371.0
+   },
+   "ye2026_sepsis": {
+    "rows": 85,
+    "usable_rows": 82,
+    "pearson_log1p_cpm_bulk_vs_pseudobulk": {
+     "n": 82,
+     "median": 0.9144361928373272,
+     "min": 0.827249093313906,
+     "max": 0.9346381375723359,
+     "q25": 0.8992056268849085,
+     "q75": 0.9238219583968234
+    },
+    "pair_valid_genes_per_row": {
+     "n": 82,
+     "median": 19864.0,
+     "min": 19864.0,
+     "max": 19864.0,
+     "q25": 19864.0,
+     "q75": 19864.0
+    },
+    "retained_cells_per_usable_row": {
+     "n": 82,
+     "median": 7153.5,
+     "min": 2107.0,
+     "max": 15576.0,
+     "q25": 5959.5,
+     "q75": 8269.75
+    },
+    "retained_cells_total_usable_rows": 589478,
+    "umi_per_cell_source_median_all_cells_usable_rows": 4044.0,
+    "umi_per_cell_canonical_median_all_cells_usable_rows": 3889.0,
+    "umi_per_cell_source_median_of_row_medians": 4136.0
+   },
+   "imperial_eskd": {
+    "rows": 172,
+    "usable_rows": 157,
+    "pearson_log1p_cpm_bulk_vs_pseudobulk": {
+     "n": 157,
+     "median": 0.8754111970180192,
+     "min": 0.815300615512258,
+     "max": 0.9035985714099305,
+     "q25": 0.8552299242119961,
+     "q75": 0.8880094216040912
+    },
+    "pair_valid_genes_per_row": {
+     "n": 157,
+     "median": 19656.0,
+     "min": 19656.0,
+     "max": 19656.0,
+     "q25": 19656.0,
+     "q75": 19656.0
+    },
+    "retained_cells_per_usable_row": {
+     "n": 157,
+     "median": 2483.0,
+     "min": 226.0,
+     "max": 11388.0,
+     "q25": 1286.0,
+     "q75": 4765.0
+    },
+    "retained_cells_total_usable_rows": 514260,
+    "umi_per_cell_source_median_all_cells_usable_rows": 2946.0,
+    "umi_per_cell_canonical_median_all_cells_usable_rows": 2813.0,
+    "umi_per_cell_source_median_of_row_medians": 2841.0
+   }
+  }
+ },
+ "hto_negative_sensitivity": {
+  "genes": 32362,
+  "rows_with_>=50_negative_cells": 20,
+  "median_negative_fraction_those_rows": 0.403779248910689,
+  "r_own_singlet_median": 0.9182934727603804,
+  "r_own_negative_median": 0.9005794797648514,
+  "r_own_singlet_plus_negative_median": 0.9178280617774,
+  "rows_where_adding_negatives_raises_r_own": 7,
+  "negatives_own_timepoint_beats_best_other_timepoint": "10/14",
+  "singlets_own_timepoint_beats_best_other_timepoint": "12/14",
+  "note": "r over hg19 symbols shared by the 10x table (unique) and the bulk table; informational only - the dataset keeps the task rule (HTO singlets only)"
+ },
+ "cells_full": {
+  "dataset": "liu2021",
+  "n_cells": 191544,
+  "nnz": 288833735,
+  "sha256": "f262d7f30c70009863fc710f44cc672271e1d845fb4c3e6fa6bf00fd0853af8c",
+  "checks": {
+   "panel_columns_equal_cells_panel6k": true,
+   "row_sums_equal_pseudo_raw_sum": true
+  },
+  "builder": "processing/liu2021_v1_20261002/code/liu2021_full_cells.py (build_full_cells.read_library, same checks, streamed; re-verified on the written file)"
+ }
+}

paper-releases/pbisc-diffusion/updates/20261002T124820Z-crosscohort-liu-default/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/data/liu2021/MANIFEST.json

@@ -0,0 +1,864 @@
+{
+ "schema": "liu2021_paired_ft_v1",
+ "written_utc": "2026-10-02T12:22:27.217313+00:00",
+ "started_utc": "2026-10-02T12:21:53.277358+00:00",
+ "study": "Liu C, Martins AJ, Lau WW, et al. Time-resolved systems immunology reveals a late juncture linked to fatal COVID-19. Cell 2021;184:1836-1857 (doi 10.1016/j.cell.2021.02.018; PMID 33713619); GEO GSE161918 (bulk RNA-seq and CITE-seq in one series)",
+ "use": "extra TRAINING cohort only (its single cells are in the model's pretraining corpus); never a test cohort",
+ "bulk_material": "PBMC (GEO source 'Peripheral Blood Mononuclear Cells', 'Unsorted'; 100,000 cells per sample)",
+ "rows": 65,
+ "subjects": 44,
+ "row_order": "bulk GSM ascending",
+ "pairs_by_group": {
+  "COVID": 54,
+  "HC": 11
+ },
+ "pairs_by_timepoint": {
+  "T0": 28,
+  "T1": 19,
+  "HC": 11,
+  "T2": 4,
+  "T3": 3
+ },
+ "pairs_by_batch": {
+  "B3": 23,
+  "B1": 22,
+  "B2": 20
+ },
+ "pairs_by_subject": {
+  "AA220014": "HC",
+  "AA220534": "HC",
+  "AA220907": "HC",
+  "HGR0000042": "T0,T1",
+  "HGR0000051": "T0,T1,T2,T3",
+  "HGR0000069": "T0,T1,T2",
+  "HGR0000071": "T0",
+  "HGR0000074": "T0",
+  "HGR0000078": "T0,T1",
+  "HGR0000079": "T0,T1",
+  "HGR0000083": "T0,T1",
+  "HGR0000089": "T1",
+  "HGR0000091": "T0",
+  "HGR0000092": "T0,T1",
+  "HGR0000093": "T0,T1",
+  "HGR0000094": "T0,T1",
+  "HGR0000098": "T0",
+  "HGR0000101": "T0,T1,T3",
+  "HGR0000102": "T0,T1",
+  "HGR0000110": "T0",
+  "HGR0000115": "T1",
+  "HGR0000124": "T0,T1",
+  "HGR0000134": "T0,T1",
+  "HGR0000135": "T0,T1",
+  "HGR0000136": "T0",
+  "HGR0000137": "T0",
+  "HGR0000141": "T0",
+  "HGR0000142": "T0",
+  "HGR0000143": "T0",
+  "HGR0000144": "T0,T2",
+  "HGR0000391": "T0",
+  "HGR0000392": "T0,T3",
+  "HGR0000395": "T0",
+  "HGR0000418": "T1",
+  "HGR0000429": "T1,T2",
+  "HGR0000430": "T1",
+  "IM000040": "HC",
+  "SHD1": "HC",
+  "SHD2": "HC",
+  "SHD3": "HC",
+  "SHD4": "HC",
+  "SHD5": "HC",
+  "SHD6": "HC",
+  "SHD7": "HC"
+ },
+ "pairs_by_id_match": {
+  "exact_subject_timepoint": 63,
+  "subject_timepoint; bulk column name corrected": 1,
+  "subject_timepoint; GEO bulk timepoint characteristic differs": 1
+ },
+ "cells": {
+  "n_cells": 191544,
+  "nnz": 82775694,
+  "per_row_min": 114,
+  "per_row_median": 2705.0,
+  "per_row_max": 8731,
+  "median_umi_per_cell_canonical": 4384.0,
+  "median_umi_per_cell_source": 4613.0
+ },
+ "cell_selection": "author Seurat object cells with Sorted == 'N' (unsorted fraction), HTO_classification.global == 'Singlet' and DROPLET.TYPE == 'SNG'; then flag_conservative_review_pass",
+ "retained_cells_rule": "flag_conservative_review_pass = genes>=200 & UMIs>=500 & mito<=20% (qc_core, ImMicro definition); MAD flags are review-only columns",
+ "features_semantics": {
+  "bulk_raw": "author htseq-count raw gene counts (STAR / UCSC hg19) mapped by gene symbol to the canonical axis; NaN where bulk_valid is False (task contract; ImMicro also uses NaN; Giroux/Ye/Imperial store 0); many-to-one symbol groups summed (bundle policy for verified integer counts)",
+  "pseudo_raw_sum": "sum of retained cells' UMI counts on the canonical axis; storage 0 where sc_valid is False",
+  "bulk_valid": "harmonization_valid from qc_bulk.harmonize (symbol in the author table maps uniquely to the canonical gene)",
+  "sc_valid": "harmonization_valid from qc_singlecell.canonical_export (gene in the hg19-1.2.0 10x features table, mapped by Ensembl ID)",
+  "pair_valid": "bulk_valid & sc_valid",
+  "sample_ids": "<subject>_<timepoint>, e.g. HGR0000083_T0 or SHD1_HC (= author sample_name)",
+  "donor_ids": "author Subject; a COVID-19 subject has 1-4 rows (timepoints)",
+  "timepoints": "author timepoint label (T0, T1, T2, T3; HC for healthy controls), string",
+  "groups": "author Class: COVID or HC",
+  "extra_keys": "bulk_gsm, bulk_count_column, bulk_title, sc_sample (author sample_id = batch_subject_timepoint), id_match, batch, geo_bulk_timepoint (GEO characteristic of the bulk GSM), severity (author: Moderate/Severe/Critical; HC), outcome (author 'outcome': alive/deceased; HC), status_meta (author HTO-metadata 'Status': Alive/Deceased/na), days_since_onset, days_since_hospitalized (float, NaN when not given), ever_admitted_to_icu, ward, sex_meta, age_meta, below_200_cells, identity_pass, identity_check, sex_bulk_expr, sex_sc_expr"
+ },
+ "identity": {
+  "sex_rule": "CPM on the source axis (bulk: all 55,765 author-table genes; sc: retained cells, all 32,738 RNA features); score=log2(XIST+1)-log2(Ysum+1), Ysum = RPS4Y1+DDX3Y+UTY+KDM5D in both; female>2, male<-2, else ambiguous (the Ye 2026 rule). sc only, verbatim the Imperial ESKD rule: where the pseudobulk call is ambiguous, the call falls back to per-cell detection (female if <20 % of retained cells have a Y-gene UMI and >=5 % have an XIST UMI; male if >=40 % have a Y-gene UMI and <20 % an XIST UMI), because pooled 10x lanes carry the pool-mates' ambient RNA and 5' capture of XIST is low. Concordant = bulk call = sc call = author metadata Gender. Opposite-sex flag: female call with Y CPM>20 or male call with XIST CPM>50 (bulk)",
+  "expression_rule": "log2(CPM+1) over the 18512 non-chrY canonical genes measured in all 73 bulk columns and all 70 sc samples; top 2000 by the mean of the bulk-mean and sc-mean log CPM; Spearman of each paired bulk vs all sc pseudobulks; centred sensitivity; reverse vs all bulks",
+  "expression_residualised_rule": "within each modality: centre log2(CPM+1) (same genes), remove the top k sample PCs computed on expressed genes (bulk mean >3 and sc mean >1 log2 CPM), keep the 2000 genes with the highest residual variance in both modalities (rank sum), Pearson over all 73 bulk columns x all 70 sc samples (controls and SHD8 included as null candidates; one subject label per person). k=5 primary, k=3 and k=10 sensitivity. Ranks per sample and per subject (each subject scored by its best sample). chrY genes excluded. Uses no pairing information.",
+  "verdict_rule": "pass = sex concordant (bulk, sc, metadata) and residualised k=5 own subject ranked 1st in one direction and within the top 3 subjects in the other and no timepoint failure; fail = sex discordant, or own subject below 3 in both directions, or timepoint failure; else borderline",
+  "timepoint_check": "within the subject's paired rows, rank of the own timepoint on the residualised (k=5) and on the centred (not residualised) expression correlation, both directions; failure = another timepoint of the same subject beats the own one in all four (the Giroux 2022 rule)",
+  "isg_rule": "informational: mean log2(CPM+1) of 12 ISGs (IFI27, IFI44L, ISG15, RSAD2, IFIT1, IFIT3, MX1, OAS1, OAS3, SIGLEC1, IFI6, HERC5); per subject the ISG-peak timepoint in bulk and in sc",
+  "policy": "reported only; no row dropped on identity results; features.npz carries identity_pass / identity_check"
+ },
+ "checks": {
+  "gene_positions_sha256_ok": true,
+  "panel_gene_ids_match_selected_genes_csv": true,
+  "panel_gene_ids_match_giroux2022_cells_panel6k": true,
+  "bulk_integer_nonnegative_inside_mask": true,
+  "bulk_raw_nan_exactly_outside_bulk_valid": true,
+  "pseudo_raw_sum_equals_sum_of_retained_cells_full_axis": true,
+  "cells_panel_sums_equal_pseudo_raw_sum_positions_exactly": true,
+  "cells_per_row_equal_qc_cell_counts": true,
+  "no_sc_mass_outside_sc_valid": true,
+  "one_sample_id_per_library": true,
+  "sc_cells_are_unsorted_hto_singlet_demuxlet_singlet": true,
+  "sex_concordant_pairs": "63/65",
+  "expression_best_match_is_own_pair": "25/65",
+  "expression_best_match_is_own_pair_centred": "40/65",
+  "expression_reverse_best_match_is_own_pair": "17/65",
+  "expression_residualised_k5_own_subject_top1_bulk_to_sc": "62/65",
+  "expression_residualised_k3_own_subject_top1_bulk_to_sc": "61/65",
+  "expression_residualised_k10_own_subject_top1_bulk_to_sc": "60/65",
+  "expression_residualised_k5_own_subject_top1_sc_to_bulk": "63/65",
+  "expression_residualised_k3_own_subject_top1_sc_to_bulk": "64/65",
+  "expression_residualised_k10_own_subject_top1_sc_to_bulk": "62/65",
+  "expression_residualised_k5_own_library_top1_bulk_to_sc": "61/65",
+  "expression_residualised_k3_own_library_top1_bulk_to_sc": "59/65",
+  "expression_residualised_k10_own_library_top1_bulk_to_sc": "60/65",
+  "expression_residualised_k5_own_timepoint_best_within_subject_bulk_to_sc_multi_timepoint_rows": "37/38",
+  "expression_residualised_k5_own_timepoint_best_within_subject_sc_to_bulk_multi_timepoint_rows": "37/38",
+  "expression_centred_own_timepoint_best_within_subject_bulk_to_sc_multi_timepoint_rows": "30/38",
+  "expression_centred_own_timepoint_best_within_subject_sc_to_bulk_multi_timepoint_rows": "34/38",
+  "isg_peak_timepoint_same_in_both_assays_multi_timepoint_subjects": "16/17",
+  "identity_check_pass": "61/65",
+  "identity_check_fail": [
+   "HGR0000079_T0",
+   "HGR0000079_T1",
+   "HGR0000110_T0"
+  ],
+  "identity_check_borderline": [
+   "HGR0000078_T0"
+  ],
+  "rows_below_200_cells": [
+   "HGR0000078_T0",
+   "HGR0000078_T1",
+   "HGR0000079_T0",
+   "HGR0000079_T1",
+   "HGR0000083_T0",
+   "HGR0000093_T1",
+   "HGR0000110_T0"
+  ],
+  "usable_rows_identity_pass_and_ge_200_cells": 58
+ },
+ "sc_sex_ambient": {
+  "rule": "see identity.sex_rule",
+  "samples_called_per_cell": [
+   "B1_AA220907_HC",
+   "B1_HGR0000078_T0",
+   "B1_HGR0000078_T1",
+   "B2_CHI014_HC",
+   "B2_HGR0000144_T0",
+   "B2_HGR0000144_T2",
+   "B3_CHI014_HC"
+  ],
+  "male_meta_frac_cells_Y_min": 0.4140625,
+  "female_meta_frac_cells_Y_max": 0.8723404255319149,
+  "female_meta_samples": [
+   {
+    "sc_sample": "B1_AA220907_HC",
+    "sc_XIST_cpm": 47.693,
+    "sc_Ysum_cpm": 15.8977,
+    "sc_frac_cells_XIST": 0.168,
+    "sc_frac_cells_Y": 0.0682,
+    "sc_sex_call_pseudobulk": "ambiguous",
+    "sc_sex_call": "female"
+   },
+   {
+    "sc_sample": "B1_HGR0000071_T0",
+    "sc_XIST_cpm": 85.6451,
+    "sc_Ysum_cpm": 14.2742,
+    "sc_frac_cells_XIST": 0.2405,
+    "sc_frac_cells_Y": 0.0528,
+    "sc_sex_call_pseudobulk": "female",
+    "sc_sex_call": "female"
+   },
+   {
+    "sc_sample": "B1_HGR0000078_T0",
+    "sc_XIST_cpm": 44.0037,
+    "sc_Ysum_cpm": 11.0009,
+    "sc_frac_cells_XIST": 0.163,
+    "sc_frac_cells_Y": 0.0326,
+    "sc_sex_call_pseudobulk": "ambiguous",
+    "sc_sex_call": "female"
+   },
+   {
+    "sc_sample": "B1_HGR0000078_T1",
+    "sc_XIST_cpm": 53.4007,
+    "sc_Ysum_cpm": 16.1444,
+    "sc_frac_cells_XIST": 0.1658,
+    "sc_frac_cells_Y": 0.0603,
+    "sc_sex_call_pseudobulk": "ambiguous",
+    "sc_sex_call": "female"
+   },
+   {
+    "sc_sample": "B1_HGR0000079_T0",
+    "sc_XIST_cpm": 2.4682,
+    "sc_Ysum_cpm": 972.4769,
+    "sc_frac_cells_XIST": 0.0088,
+    "sc_frac_cells_Y": 0.8158,
+    "sc_sex_call_pseudobulk": "male",
+    "sc_sex_call": "male"
+   },
+   {
+    "sc_sample": "B1_HGR0000079_T1",
+    "sc_XIST_cpm": 0.0,
+    "sc_Ysum_cpm": 612.567,
+    "sc_frac_cells_XIST": 0.0,
+    "sc_frac_cells_Y": 0.8723,
+    "sc_sex_call_pseudobulk": "male",
+    "sc_sex_call": "male"
+   },
+   {
+    "sc_sample": "B1_HGR0000136_T0",
+    "sc_XIST_cpm": 87.0162,
+    "sc_Ysum_cpm": 10.2775,
+    "sc_frac_cells_XIST": 0.2744,
+    "sc_frac_cells_Y": 0.0379,
+    "sc_sex_call_pseudobulk": "female",
+    "sc_sex_call": "female"
+   },
+   {
+    "sc_sample": "B1_HGR0000137_T0",
+    "sc_XIST_cpm": 78.1932,
+    "sc_Ysum_cpm": 8.6012,
+    "sc_frac_cells_XIST": 0.2357,
+    "sc_frac_cells_Y": 0.035,
+    "sc_sex_call_pseudobulk": "female",
+    "sc_sex_call": "female"
+   },
+   {
+    "sc_sample": "B1_HGR0000143_T0",
+    "sc_XIST_cpm": 100.0308,
+    "sc_Ysum_cpm": 15.3894,
+    "sc_frac_cells_XIST": 0.2571,
+    "sc_frac_cells_Y": 0.05,
+    "sc_sex_call_pseudobulk": "female",
+    "sc_sex_call": "female"
+   },
+   {
+    "sc_sample": "B2_HGR0000144_T0",
+    "sc_XIST_cpm": 85.8328,
+    "sc_Ysum_cpm": 30.6027,
+    "sc_frac_cells_XIST": 0.329,
+    "sc_frac_cells_Y": 0.1507,
+    "sc_sex_call_pseudobulk": "ambiguous",
+    "sc_sex_call": "female"
+   },
+   {
+    "sc_sample": "B2_HGR0000144_T2",
+    "sc_XIST_cpm": 103.5851,
+    "sc_Ysum_cpm": 33.3466,
+    "sc_frac_cells_XIST": 0.3443,
+    "sc_frac_cells_Y": 0.1382,
+    "sc_sex_call_pseudobulk": "ambiguous",
+    "sc_sex_call": "female"
+   },
+   {
+    "sc_sample": "B3_HGR0000430_T1",
+    "sc_XIST_cpm": 119.8774,
+    "sc_Ysum_cpm": 22.9717,
+    "sc_frac_cells_XIST": 0.432,
+    "sc_frac_cells_Y": 0.1131,
+    "sc_sex_call_pseudobulk": "female",
+    "sc_sex_call": "female"
+   },
+   {
+    "sc_sample": "B3_SHD6_HC",
+    "sc_XIST_cpm": 117.9891,
+    "sc_Ysum_cpm": 27.7577,
+    "sc_frac_cells_XIST": 0.4057,
+    "sc_frac_cells_Y": 0.1315,
+    "sc_sex_call_pseudobulk": "female",
+    "sc_sex_call": "female"
+   }
+  ]
+ },
+ "excluded": {
+  "quarantined_sc_samples_among_pairs": [],
+  "sc_samples_without_pair": [
+   {
+    "sc_sample": "B1_CHI014_HC",
+    "kind": "technical_control",
+    "reason": "technical control (CHI014 / HD buffy-coat test donor); excluded by task rule"
+   },
+   {
+    "sc_sample": "B2_CHI014_HC",
+    "kind": "technical_control",
+    "reason": "technical control (CHI014 / HD buffy-coat test donor); excluded by task rule"
+   },
+   {
+    "sc_sample": "B2_HDML_PBMC_HC",
+    "kind": "technical_control",
+    "reason": "technical control (CHI014 / HD buffy-coat test donor); excluded by task rule"
+   },
+   {
+    "sc_sample": "B2_HDVO_PBMC_HC",
+    "kind": "technical_control",
+    "reason": "technical control (CHI014 / HD buffy-coat test donor); excluded by task rule"
+   },
+   {
+    "sc_sample": "B3_CHI014_HC",
+    "kind": "technical_control",
+    "reason": "technical control (CHI014 / HD buffy-coat test donor); excluded by task rule"
+   }
+  ],
+  "sc_samples_not_in_pairs_check": [
+   "B1_CHI014_HC",
+   "B2_CHI014_HC",
+   "B2_HDML_PBMC_HC",
+   "B2_HDVO_PBMC_HC",
+   "B3_CHI014_HC"
+  ],
+  "bulk_count_columns_without_pair": [
+   {
+    "bulk_column": "CHI-014",
+    "bulk_gsm": "GSM4929236",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "CHI014_B2",
+    "bulk_gsm": "GSM4929256",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "CHI014_B3",
+    "bulk_gsm": "GSM4929282",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "HD_ML_bc",
+    "bulk_gsm": "GSM4929257",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "HD_ML_PBMC",
+    "bulk_gsm": "GSM4929258",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "HD_VO_bc",
+    "bulk_gsm": "GSM4929259",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "HD_VO_PBMC",
+    "bulk_gsm": "GSM4929260",
+    "kind": "technical_control",
+    "unpaired_reason": "technical control column (CHI014 / HD buffy-coat test); excluded by task rule"
+   },
+   {
+    "bulk_column": "SHD8",
+    "bulk_gsm": "GSM4929281",
+    "kind": "author_excluded",
+    "unpaired_reason": "GEO: 'Healthy Control (excluded due to abnormal lymphoma-like B cells found upon analysis)'; the author Seurat object holds 0 cells of this subject"
+   }
+  ]
+ },
+ "inputs": {
+  "gene_positions": {
+   "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/artifacts/panel_ladder_floor01_1m_20260923/tiers/6k/gene_positions.npy",
+   "sha256": "653ffa45ac519000baa41086c52408009391d6e68cfb218bc6a50d34163fcb24"
+  },
+  "selected_genes": {
+   "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/artifacts/panel_ladder_floor01_1m_20260923/tiers/6k/selected_genes.csv",
+   "sha256": "09248a3930e3fdefe1dbcd5d60cae4be81028526dcf097dd78169f5a415edfb6"
+  },
+  "reference_panel": {
+   "path": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/data/giroux2022/cells_panel6k.npz",
+   "sha256": "7d22c0ff4b3af2b1fd1741421633616b5ab0e679a800ecffb780cc5fd1b507d1"
+  },
+  "canonical_genes": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/strict_qc_v1_20260913/reference/canonical_genes.tsv",
+   "sha256": "484893b6f6ab673566074e0d2f32c4ec6628ced63f7e709b38ec602cb3e776ab"
+  },
+  "bulk_harmonized": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/bulk/GSE161918_htseq/harmonized.h5ad",
+   "sha256": "bf7c69cce88992be7ff6a0ae5702306938bd32d246ca55412157793c7cf06d46"
+  },
+  "sources": [
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/datasets/GSE161918/raw/GSE161918_AllBatches_SeuratObj.rds.gz",
+    "bytes": 5840847302,
+    "sha256": "8d6782c40f7a3a76c56c8e2eb582a531818a40333331eb781f7817fcaa4651e8",
+    "acquisition_receipt": true,
+    "url": "https://ftp.ncbi.nlm.nih.gov/geo/series/GSE161nnn/GSE161918/suppl/GSE161918_AllBatches_SeuratObj.rds.gz",
+    "receipt_sha256": "8d6782c40f7a3a76c56c8e2eb582a531818a40333331eb781f7817fcaa4651e8",
+    "receipt_bytes": 5840847302,
+    "receipt_completed_at": "2026-09-27T05:41:39.527535+00:00",
+    "matches_receipt": true
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/datasets/GSE161918/raw/GSE161918_B1bulkRNAseq_countmtx.csv.gz",
+    "bytes": 1093108,
+    "sha256": "6823ad9f5bc83d565e19f811c405aa48e9c24c5b58376fd1c85204e642f6b287",
+    "acquisition_receipt": true,
+    "url": "https://ftp.ncbi.nlm.nih.gov/geo/series/GSE161nnn/GSE161918/suppl/GSE161918_B1bulkRNAseq_countmtx.csv.gz",
+    "receipt_sha256": "6823ad9f5bc83d565e19f811c405aa48e9c24c5b58376fd1c85204e642f6b287",
+    "receipt_bytes": 1093108,
+    "receipt_completed_at": "2026-09-27T04:45:44.048226+00:00",
+    "matches_receipt": true
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/datasets/GSE161918/raw/GSE161918_B2GSE161918_B3bulkRNAseq_countmtx.csv.gz",
+    "bytes": 2193695,
+    "sha256": "9e8c18dd8ad1ab7b92258b190b87510149ddd7aac7b53746ca5dcaf7d13b6abc",
+    "acquisition_receipt": true,
+    "url": "https://ftp.ncbi.nlm.nih.gov/geo/series/GSE161nnn/GSE161918/suppl/GSE161918_B2GSE161918_B3bulkRNAseq_countmtx.csv.gz",
+    "receipt_sha256": "9e8c18dd8ad1ab7b92258b190b87510149ddd7aac7b53746ca5dcaf7d13b6abc",
+    "receipt_bytes": 2193695,
+    "receipt_completed_at": "2026-09-27T04:45:49.772346+00:00",
+    "matches_receipt": true
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/datasets/GSE161918/metadata/GSE161918_family.soft.gz",
+    "bytes": 27746,
+    "sha256": "d45d2bdf3e48aeacb48a5cf3d5962809bde778a54947841bb73603760a7993c7",
+    "acquisition_receipt": true,
+    "url": "https://ftp.ncbi.nlm.nih.gov/geo/series/GSE161nnn/GSE161918/soft/GSE161918_family.soft.gz",
+    "receipt_sha256": "d45d2bdf3e48aeacb48a5cf3d5962809bde778a54947841bb73603760a7993c7",
+    "receipt_bytes": 27746,
+    "receipt_completed_at": "2026-09-27T04:45:41.191838+00:00",
+    "matches_receipt": true
+   }
+  ],
+  "hto_metadata": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/analyses/paired_inventory_20260927/new_pairmaps/GSE161918_allbatches.HTOandTimepointMetadata.csv.gz",
+   "bytes": 2515,
+   "sha256": "675bc35749b33eee66602e0bce23a8758576a89356bc57979cc02f7d375eb7de",
+   "url": "https://ftp.ncbi.nlm.nih.gov/geo/series/GSE161nnn/GSE161918/suppl/GSE161918_allbatches.HTOandTimepointMetadata.csv.gz",
+   "summary_json_sha256": "675bc35749b33eee66602e0bce23a8758576a89356bc57979cc02f7d375eb7de",
+   "note": "GEO supplementary file cached by the 2026-09-27 pairing audit (read only); used only as an independent cross-check of the Seurat object's (subject, hashtag) -> timepoint / sorted labels",
+   "matches_summary_json": true
+  },
+  "rds_export": [
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/rds_export/STRUCTURE.json",
+    "bytes": 30200,
+    "sha256": "e2835f6cf8ecfc76d3c8fee45ef2586d56b7a67446010da9f7ec8758f240738c"
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/rds_export/cell_metadata.csv.gz",
+    "bytes": 56680999,
+    "sha256": "f1100540e869c70e45bc77a2b701c6b2f43f813e27fff7e1be756014bf79ef4c"
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/rds_export/other_assay_features.json",
+    "bytes": 3149,
+    "sha256": "e672ee195e9e5c124f27ed9176a958b0e23ec871159b327227d777d26ae7edc9"
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/rds_export/rna_cells.txt",
+    "bytes": 9020167,
+    "sha256": "c3f53117e4b5468f7e69132302583f6730962dec0b6d35ce2357ea736d75a043"
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/rds_export/rna_counts_i.int32",
+    "bytes": 2538267272,
+    "sha256": "534aab50e0cffed8a1c7d6abaf70d911a0c29462ca30ccb1c954f9aeefa74249"
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/rds_export/rna_counts_p.int32",
+    "bytes": 1647612,
+    "sha256": "5749155d541868b47d0c0801a2bc816af2c68df8f27218d6b9b8e337e6458bff"
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/rds_export/rna_counts_x.int32",
+    "bytes": 2538267272,
+    "sha256": "83a96baa8699ae613e48aa5db0725a39d31c5870d21cc1fb58ba2a2a336517b6"
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/rds_export/rna_genes.txt",
+    "bytes": 293334,
+    "sha256": "429ed81524920d8ca87e054efb805297db8c384390b28d8aae074d6d12ddea81"
+   },
+   {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/rds_export/rna_meta_features.csv.gz",
+    "bytes": 894613,
+    "sha256": "fa6ce0b156c5ad5368ba957ca096bd292f0a8702860fba8a3f1ad89fb41313d5"
+   }
+  ],
+  "crosswalk": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/reference_crosswalk/cellranger_hg19-1.2.0_features__from_GSE206283_GSM6249265.tsv.gz",
+   "sha256": "d0bdfea9e111c174bb2dd81fd63a8bbaf93829c0d1bef396aaf2f3b6b5774a58",
+   "bytes": 310002
+  }
+ },
+ "code": {
+  "assemble": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/liu2021_assemble.py",
+   "sha256": "6f1ef885f0ea3a1ed57954f4fac3f19c905149b03d1d1a474fdce68ce39fc01f"
+  },
+  "pipeline": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/liu2021_pipeline.py",
+   "sha256": "e32665bd6e4f6e0582f4a0ea9d350a8fd2f652a3e5ee16cf0cab4424bdf63ee7"
+  },
+  "rds_export": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/liu2021_rds_export.R",
+   "sha256": "9f951a328b9756fb1a203e988e11443cb36c6bad671133b1ec52bf878faba576"
+  },
+  "run_step": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/run_step.sh",
+   "sha256": "3f37e1e0dd2c0ab7a733e576ffff370f64fc73436cf219f1b2b056ffd61613b8"
+  },
+  "bundle/qc_bulk.py": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/strict_qc_v1_code_bundle_v5/qc_bulk.py",
+   "sha256": "8c3bd5179a11c25fdeea491d7071bd492faaabc3cff3348a0750de199cddae9c"
+  },
+  "bundle/qc_core.py": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/strict_qc_v1_code_bundle_v5/qc_core.py",
+   "sha256": "b50a175b64044142676c8c55f9630da90d8c0c193c761124de9ef02d517e4851"
+  },
+  "bundle/qc_finalize.py": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/strict_qc_v1_code_bundle_v5/qc_finalize.py",
+   "sha256": "c4acc068779e3823869bafa4cb8a9384b60b3e66693bbacc313758d0bd495f0a"
+  },
+  "bundle/qc_genes.py": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/strict_qc_v1_code_bundle_v5/qc_genes.py",
+   "sha256": "ff93bbc364af6a0822b0cff7c5ef0fb407aa2512bc8afa6f0681754f0fb65c86"
+  },
+  "bundle/qc_singlecell.py": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/strict_qc_v1_code_bundle_v5/qc_singlecell.py",
+   "sha256": "a69ab1e0b1686e466474735b67a398f5d2140936e66a95051d02a6a497a59753"
+  },
+  "bundle/qc_validate_outputs.py": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/strict_qc_v1_code_bundle_v5/qc_validate_outputs.py",
+   "sha256": "8fe39a62b38f40e3e6053e9300abe81818a27f0e39a044e660eb88c094480daa"
+  },
+  "bundle/strict_qc_driver.py": {
+   "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/strict_qc_v1_code_bundle_v5/strict_qc_driver.py",
+   "sha256": "88ec34fb10074334a4e4ba2f4059a9b04ade1ba3b335d2c448a1f67594daa078"
+  }
+ },
+ "processing_dir": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002",
+ "outputs": {
+  "cells_obs.parquet": {
+   "bytes": 6061893,
+   "sha256": "bd2f4d36a89a4e6720ed938a9800a466545e959e88bf5b99b4ccd69c434aad0d"
+  },
+  "cells_panel6k.npz": {
+   "bytes": 156556059,
+   "sha256": "481bddfaaf2980a10cbe4d0f57a1c6f02e6146141dc00736bc898ac915390c8c"
+  },
+  "features.npz": {
+   "bytes": 4764128,
+   "sha256": "124ca7198c529f8fd0856731ab04b2bfefba3691c3ed960056ff10f8ffba29e3"
+  },
+  "identity_expression.csv": {
+   "bytes": 35426,
+   "sha256": "a05013e32d68f1d5f2bbdf690a6c9c7e7d7a0d2d4cb6ae4d4a75a749c738fbb2"
+  },
+  "identity_sex_bulk_all.csv": {
+   "bytes": 12109,
+   "sha256": "702475e9856a9e7abe10a6790c50ae8b0d89fd54f08c54422557f2c0707cc86d"
+  },
+  "identity_sex_sc_all.csv": {
+   "bytes": 17690,
+   "sha256": "9ed743899b555e5016d6655903d45e1f4bca33a071930b889c1c74e8be8c88ed"
+  },
+  "pairs.csv": {
+   "bytes": 59582,
+   "sha256": "6a695e34683ac05e3c195e2a4a3a4e6cc5e7bbbd23b01e13d5b03fd495a114de"
+  },
+  "samples.csv": {
+   "bytes": 2405,
+   "sha256": "ac3808edce13a0a6b1be040d85252bd55fc06a7a4785d040137f846915ee5412"
+  }
+ },
+ "post_assembly": {
+  "written_utc": "2026-10-02T12:40:05.485006+00:00",
+  "cells_full": {
+   "dataset": "liu2021",
+   "n_cells": 191544,
+   "nnz": 288833735,
+   "sha256": "f262d7f30c70009863fc710f44cc672271e1d845fb4c3e6fa6bf00fd0853af8c",
+   "checks": {
+    "panel_columns_equal_cells_panel6k": true,
+    "row_sums_equal_pseudo_raw_sum": true
+   },
+   "builder": "processing/liu2021_v1_20261002/code/liu2021_full_cells.py (build_full_cells.read_library, same checks, streamed; re-verified on the written file)",
+   "bytes": 2312969790
+  },
+  "prepare_dataset": {
+   "command": "CUDA_VISIBLE_DEVICES='' python tools/prepare_dataset.py --dataset liu2021 (cxg_corpus env, bounded systemd unit)",
+   "tool_sha256": "36384262b23cf6260eff645812911c1a44f791ca8e4182765d91f2aabb3ccaef",
+   "PREPARE.json_sha256": "1e5ca31545be24c8b4ed9207de498cbab15d01eb3c8d297749f8557e31f5c0e5",
+   "folds.csv_sha256": "1d8aa9a8af520e73d14386b2573655286e1f5282d0209e0a6c83d4ce33e3dc5f",
+   "fold_sizes": {
+    "0": {
+     "test": 11,
+     "train": 43,
+     "inner_val": 4,
+     "test_groups": {
+      "COVID": 8,
+      "HC": 3
+     },
+     "alpha_genes_estimated": 10928
+    },
+    "1": {
+     "test": 12,
+     "train": 41,
+     "inner_val": 5,
+     "test_groups": {
+      "COVID": 10,
+      "HC": 2
+     },
+     "alpha_genes_estimated": 10883
+    },
+    "2": {
+     "test": 13,
+     "train": 41,
+     "inner_val": 4,
+     "test_groups": {
+      "COVID": 11,
+      "HC": 2
+     },
+     "alpha_genes_estimated": 10888
+    },
+    "3": {
+     "test": 10,
+     "train": 45,
+     "inner_val": 3,
+     "test_groups": {
+      "COVID": 8,
+      "HC": 2
+     },
+     "alpha_genes_estimated": 10925
+    },
+    "4": {
+     "test": 12,
+     "train": 42,
+     "inner_val": 4,
+     "test_groups": {
+      "COVID": 10,
+      "HC": 2
+     },
+     "alpha_genes_estimated": 10915
+    }
+   },
+   "usable_rows": 58,
+   "rows_in_fold_minus1": [
+    "HGR0000078_T0",
+    "HGR0000078_T1",
+    "HGR0000079_T0",
+    "HGR0000079_T1",
+    "HGR0000083_T0",
+    "HGR0000093_T1",
+    "HGR0000110_T0"
+   ],
+   "donors_in_folds": 41,
+   "donors_split_across_folds": 0
+  },
+  "sanity_bulk_vs_pseudobulk": {
+   "definition": "per usable row (folds.csv fold >= 0): Pearson r of log1p(CPM) bulk vs log1p(CPM) sc pseudobulk (pseudo_raw_sum, retained cells) over the row's pair_valid genes; CPM over the same pair_valid genes in both assays; NaN bulk outside bulk_valid is never used",
+   "liu2021": {
+    "rows": 65,
+    "usable_rows": 58,
+    "pearson_log1p_cpm_bulk_vs_pseudobulk": {
+     "n": 58,
+     "median": 0.8978402352314852,
+     "min": 0.8691990435358162,
+     "max": 0.9118583780172661,
+     "q25": 0.8924504463995364,
+     "q75": 0.9009274993985295
+    },
+    "pair_valid_genes_per_row": {
+     "n": 58,
+     "median": 18553.0,
+     "min": 18553.0,
+     "max": 18553.0,
+     "q25": 18553.0,
+     "q75": 18553.0
+    },
+    "retained_cells_per_usable_row": {
+     "n": 58,
+     "median": 2932.0,
+     "min": 240.0,
+     "max": 8731.0,
+     "q25": 565.25,
+     "q75": 5183.25
+    },
+    "retained_cells_total_usable_rows": 190468,
+    "umi_per_cell_source_median_all_cells_usable_rows": 4617.0,
+    "umi_per_cell_canonical_median_all_cells_usable_rows": 4388.0,
+    "umi_per_cell_source_median_of_row_medians": 4540.5,
+    "by_batch": {
+     "B1": {
+      "rows": 15,
+      "r_median": 0.8865,
+      "cells_median": 353.0,
+      "umi_median_of_row_medians": 3989.0
+     },
+     "B2": {
+      "rows": 20,
+      "r_median": 0.8984000000000001,
+      "cells_median": 4418.5,
+      "umi_median_of_row_medians": 4850.25
+     },
+     "B3": {
+      "rows": 23,
+      "r_median": 0.902,
+      "cells_median": 4829.0,
+      "umi_median_of_row_medians": 4569.0
+     }
+    }
+   },
+   "giroux2022": {
+    "rows": 26,
+    "usable_rows": 24,
+    "pearson_log1p_cpm_bulk_vs_pseudobulk": {
+     "n": 24,
+     "median": 0.8991432339915457,
+     "min": 0.8579903693274028,
+     "max": 0.912171011362672,
+     "q25": 0.8893585269254364,
+     "q75": 0.90455429225548
+    },
+    "pair_valid_genes_per_row": {
+     "n": 24,
+     "median": 18280.0,
+     "min": 18280.0,
+     "max": 18280.0,
+     "q25": 18280.0,
+     "q75": 18280.0
+    },
+    "retained_cells_per_usable_row": {
+     "n": 24,
+     "median": 10958.0,
+     "min": 5914.0,
+     "max": 15650.0,
+     "q25": 9059.5,
+     "q75": 12024.25
+    },
+    "retained_cells_total_usable_rows": 254250,
+    "umi_per_cell_source_median_all_cells_usable_rows": 3318.0,
+    "umi_per_cell_canonical_median_all_cells_usable_rows": 3179.0,
+    "umi_per_cell_source_median_of_row_medians": 3371.0
+   },
+   "ye2026_sepsis": {
+    "rows": 85,
+    "usable_rows": 82,
+    "pearson_log1p_cpm_bulk_vs_pseudobulk": {
+     "n": 82,
+     "median": 0.9144361928373272,
+     "min": 0.827249093313906,
+     "max": 0.9346381375723359,
+     "q25": 0.8992056268849085,
+     "q75": 0.9238219583968234
+    },
+    "pair_valid_genes_per_row": {
+     "n": 82,
+     "median": 19864.0,
+     "min": 19864.0,
+     "max": 19864.0,
+     "q25": 19864.0,
+     "q75": 19864.0
+    },
+    "retained_cells_per_usable_row": {
+     "n": 82,
+     "median": 7153.5,
+     "min": 2107.0,
+     "max": 15576.0,
+     "q25": 5959.5,
+     "q75": 8269.75
+    },
+    "retained_cells_total_usable_rows": 589478,
+    "umi_per_cell_source_median_all_cells_usable_rows": 4044.0,
+    "umi_per_cell_canonical_median_all_cells_usable_rows": 3889.0,
+    "umi_per_cell_source_median_of_row_medians": 4136.0
+   },
+   "imperial_eskd": {
+    "rows": 172,
+    "usable_rows": 157,
+    "pearson_log1p_cpm_bulk_vs_pseudobulk": {
+     "n": 157,
+     "median": 0.8754111970180192,
+     "min": 0.815300615512258,
+     "max": 0.9035985714099305,
+     "q25": 0.8552299242119961,
+     "q75": 0.8880094216040912
+    },
+    "pair_valid_genes_per_row": {
+     "n": 157,
+     "median": 19656.0,
+     "min": 19656.0,
+     "max": 19656.0,
+     "q25": 19656.0,
+     "q75": 19656.0
+    },
+    "retained_cells_per_usable_row": {
+     "n": 157,
+     "median": 2483.0,
+     "min": 226.0,
+     "max": 11388.0,
+     "q25": 1286.0,
+     "q75": 4765.0
+    },
+    "retained_cells_total_usable_rows": 514260,
+    "umi_per_cell_source_median_all_cells_usable_rows": 2946.0,
+    "umi_per_cell_canonical_median_all_cells_usable_rows": 2813.0,
+    "umi_per_cell_source_median_of_row_medians": 2841.0
+   }
+  },
+  "hto_negative_sensitivity": {
+   "genes": 32362,
+   "rows_with_>=50_negative_cells": 20,
+   "median_negative_fraction_those_rows": 0.403779248910689,
+   "r_own_singlet_median": 0.9182934727603804,
+   "r_own_negative_median": 0.9005794797648514,
+   "r_own_singlet_plus_negative_median": 0.9178280617774,
+   "rows_where_adding_negatives_raises_r_own": 7,
+   "negatives_own_timepoint_beats_best_other_timepoint": "10/14",
+   "singlets_own_timepoint_beats_best_other_timepoint": "12/14",
+   "note": "r over hg19 symbols shared by the 10x table (unique) and the bulk table; informational only - the dataset keeps the task rule (HTO singlets only)"
+  },
+  "code": {
+   "full_cells": {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/liu2021_full_cells.py",
+    "sha256": "84a29ba381667ee1f93a1e58a9fc9c74b4eabe3684cc9bd9bfa37164e69672aa"
+   },
+   "findings": {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/liu2021_findings.py",
+    "sha256": "56b14abba09366e4361c91d0ceba384bf56cbf57fe0d4dd407091e1f570336c3"
+   },
+   "sanity": {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/liu2021_sanity.py",
+    "sha256": "517e16e0d62a766955496efaacf904dfa46c16154c8b90d3cf017c6069f7705a"
+   },
+   "hto_sensitivity": {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/liu2021_hto_sensitivity.py",
+    "sha256": "409bcffebdb12b277c00ae9ad05af66e32e11ef7a40751afba9c31b0ae29820d"
+   },
+   "run_step": {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/run_step.sh",
+    "sha256": "141e275db26372f93bde504fad285418a0ba379087108d3352236c427b190b67"
+   },
+   "finalize": {
+    "path": "/home/sj_server_1/gblee/datalake/paired-pbmc-bulk-sc/processing/liu2021_v1_20261002/code/liu2021_finalize.py",
+    "sha256": "c841b50d1e3a14333ef3b4450bba4264e6e51d7c8f50117b16360b944590cbbf"
+   }
+  }
+ }
+}

paper-releases/pbisc-diffusion/updates/20261002T124820Z-crosscohort-liu-default/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/data/liu2021/PREPARE.json

@@ -0,0 +1,91 @@
+{
+ "schema": "pbisc_sdd.ft_intradataset.conditions.v1",
+ "dataset": "liu2021",
+ "written_utc": "2026-10-02T12:22:49.933112+00:00",
+ "features_sha256": "124ca7198c529f8fd0856731ab04b2bfefba3691c3ed960056ff10f8ffba29e3",
+ "seed": 20260927,
+ "folds": 5,
+ "inner_frac": 0.1,
+ "cells_per_bulk": 1500,
+ "recipe_k": 1000,
+ "depth_anchor": 3869.603120208939,
+ "builder": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/realbulk_generation_20260922/build_realbulk_conditions.py",
+ "builder_sha256": "e818d69e6801b5699579db7fc450590a266708d1fc2d544743e8624c9ee6a30a",
+ "files": {
+  "conditions_raw_fixed.npz": "ad3131eabcfb11b166ca37c7585c50abb11efb60e2a736fe62de604794535e0c",
+  "conditions_pseudo_fixed.npz": "86c9820cfd2d4ae4b85d097d09bb8bf604280a7af08513f7883b9404c20b3089",
+  "fold0/conditions_cal_fixed.npz": "93421185c4e82f5e58e988dfafa399da5ea99cefb0eba7ff712bc3ad66ad30cd",
+  "fold0/test_raw.npz": "a144d7c9a297f8e8aff2d0e0d518b6e53cfe059bad805bb98da3089876e85c83",
+  "fold0/test_cal.npz": "bdb6283ecc4d1c3646ed4a5fbc7b10de1d531484ce3e3287386d761d351a0da4",
+  "fold0/test_pseudo.npz": "2b81372b28a91ca6aedda71bc133d68e818ef12437176b6cce720cca01d8c9c5",
+  "fold1/conditions_cal_fixed.npz": "fe11ac590051f2c594d9b678cb63dadfb5184bdb29840ada51961537e9dfdffb",
+  "fold1/test_raw.npz": "866c3691666eda2d7ed4750b14e2ffd6065e94c35995a1ad52f5c94f86e155be",
+  "fold1/test_cal.npz": "a04f60ba6e3d51d304a1bfd497f1dd35ea56c5e7bd409bbd2a231b66ff52d212",
+  "fold1/test_pseudo.npz": "7aba8e40a3111550ce6cadcd62f8fbbace1e1364d71d40ed737fcec4a86e5c42",
+  "fold2/conditions_cal_fixed.npz": "9c4978a6724774e827b99ce259e74b3a13fdf1d11361c4a362964e608609162d",
+  "fold2/test_raw.npz": "0eb20a0d837ab6fa4270940af3a7650c04dd4419220971f835a8f800f3177df6",
+  "fold2/test_cal.npz": "0fd5f1adbbd5dca6b2df9c26af0c07f5ac57f98035f9a07fef041d9e9631de48",
+  "fold2/test_pseudo.npz": "a6c2e49020c9f526a65df7097184adcaf1ab754296669178b55d8cdbc77b42cf",
+  "fold3/conditions_cal_fixed.npz": "2a55eee6ead416fab6dcd7817706026844f75f20bb42ab37648e6cad4a92c969",
+  "fold3/test_raw.npz": "ea3440d92aedccb3df460d22dcc0386a13830c6e306c1a5c5861b4ac06bcaec4",
+  "fold3/test_cal.npz": "4273654f1754feab2531443086b4e282fce1adcad42e349b15bc86a396075024",
+  "fold3/test_pseudo.npz": "87d19c959d0ff6538f9de948103e1926d0977fc78bba2d00ba25f4398545f3b9",
+  "fold4/conditions_cal_fixed.npz": "41fa4e01af2f1e5f6c796d18c846e179acc5391d083ebb977d3399fd1adc3fc9",
+  "fold4/test_raw.npz": "441d098b32fee9402b2ecfe0795b1bdd383ab2a19053b2b1185510b27607999e",
+  "fold4/test_cal.npz": "b69d7b5dc487ecedc529ce10b962306ba74ff28fd862c95c3bc8e81879a6067a",
+  "fold4/test_pseudo.npz": "107443816a0f0fa5a40cea43973232a4acf1497651991c9fc80deaf813f836f6"
+ },
+ "fold_sizes": {
+  "0": {
+   "test": 11,
+   "train": 43,
+   "inner_val": 4,
+   "test_groups": {
+    "COVID": 8,
+    "HC": 3
+   },
+   "alpha_genes_estimated": 10928
+  },
+  "1": {
+   "test": 12,
+   "train": 41,
+   "inner_val": 5,
+   "test_groups": {
+    "COVID": 10,
+    "HC": 2
+   },
+   "alpha_genes_estimated": 10883
+  },
+  "2": {
+   "test": 13,
+   "train": 41,
+   "inner_val": 4,
+   "test_groups": {
+    "COVID": 11,
+    "HC": 2
+   },
+   "alpha_genes_estimated": 10888
+  },
+  "3": {
+   "test": 10,
+   "train": 45,
+   "inner_val": 3,
+   "test_groups": {
+    "COVID": 8,
+    "HC": 2
+   },
+   "alpha_genes_estimated": 10925
+  },
+  "4": {
+   "test": 12,
+   "train": 42,
+   "inner_val": 4,
+   "test_groups": {
+    "COVID": 10,
+    "HC": 2
+   },
+   "alpha_genes_estimated": 10915
+  }
+ },
+ "checks": {}
+}

paper-releases/pbisc-diffusion/updates/20261002T124820Z-crosscohort-liu-default/diffusion/dataset/server_1/model8/plans/paired_finetune_intradataset_20260927/data/liu2021/cells_full.BUILD.json

@@ -0,0 +1,11 @@
+{
+ "dataset": "liu2021",
+ "n_cells": 191544,
+ "nnz": 288833735,
+ "sha256": "f262d7f30c70009863fc710f44cc672271e1d845fb4c3e6fa6bf00fd0853af8c",
+ "checks": {
+  "panel_columns_equal_cells_panel6k": true,
+  "row_sums_equal_pseudo_raw_sum": true
+ },
+ "builder": "processing/liu2021_v1_20261002/code/liu2021_full_cells.py (build_full_cells.read_library, same checks, streamed; re-verified on the written file)"
+}