Archive 4 checkpoints (1-4)
Browse files- paper-releases/pbisc-diffusion/updates/20261008T050202Z-scsemi-imperial-deck/diffusion/dataset/server_1/model8/plans/bayesprism_paired_20261006/results/RESULTS_cibersortx.md +27 -0
- paper-releases/pbisc-diffusion/updates/20261008T050202Z-scsemi-imperial-deck/diffusion/dataset/server_1/model8/plans/bayesprism_paired_20261006/results/RESULTS_full.md +27 -0
- paper-releases/pbisc-diffusion/updates/20261008T050202Z-scsemi-imperial-deck/diffusion/dataset/server_1/model8/plans/bayesprism_paired_20261006/results/imperial_eskd/cv_pooled_cibersortx/summary.json +198 -0
- paper-releases/pbisc-diffusion/updates/20261008T050202Z-scsemi-imperial-deck/diffusion/dataset/server_1/model8/plans/bayesprism_paired_20261006/results/imperial_eskd/cv_pooled_full/summary.json +166 -0
paper-releases/pbisc-diffusion/updates/20261008T050202Z-scsemi-imperial-deck/diffusion/dataset/server_1/model8/plans/bayesprism_paired_20261006/results/RESULTS_cibersortx.md
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# CIBERSORTx with the same paired donors vs LoRA (paired benchmark; gene mode cibersortx)
|
| 2 |
+
|
| 3 |
+
CIBERSORTx in place of BayesPrism ('BP' columns): official Docker Fractions, single-cell reference (<=400 cells per type from the same paired donors) with S-mode batch correction, then HiRes per-sample cell-type expression on the scoring genes (tools/run_cibersortx.sh). One fraction estimate, so 'final' = 'first'. HiRes needs more samples than cell types: splits with <= 8 test samples (Ji and Giroux CV folds) have composition only.
|
| 4 |
+
|
| 5 |
+
Same donors, observed truth, contrast and scorer as the benchmark (score_ft.py functions; the recomputed bulk broadcast equals the stored one in every split). BayesPrism reference = the split's paired donors' observed cells (states = gate | AIFI L2); mixture = the test samples' real bulk. Composition L1 in cell units (lower is better): 'final' and 'first' BayesPrism theta. logFC r = case-vs-control per-type logFC (mean CP10K) correlation with observed, mean of T/B/Mono/NK; BayesPrism Z zero-filled on genes it lacks ('covered' = only genes it models, broadcast recomputed there). DEG = top-50 overlap summed over types.
|
| 6 |
+
|
| 7 |
+
## Five-donor draws (mean ± SD over 5 draws)
|
| 8 |
+
|
| 9 |
+
| cohort | L1 BP final | L1 BP first | L1 LoRA-5 | r BP | r BP (covered) | r LoRA-5 | r broadcast | partial r BP | partial r LoRA-5 | DEG BP | DEG LoRA-5 | DEG broadcast |
|
| 10 |
+
|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
|
| 11 |
+
| ImMicro (5) | 0.491 ± 0.095 | 0.491 ± 0.095 | 0.176 ± 0.012 | – | – | – | – | – | – | – | – | – |
|
| 12 |
+
| Sepsis (5) | 0.582 ± 0.024 | 0.582 ± 0.024 | 0.466 ± 0.044 | 0.069 ± 0.023 | 0.106 ± 0.043 | 0.350 ± 0.162 | 0.320 ± 0.024 | -0.010 ± 0.026 | 0.305 ± 0.189 | 25.4 ± 3.4 | 16.4 ± 6.7 | 20.0 ± 1.9 |
|
| 13 |
+
| Ji ChILI (5) | 0.827 ± 0.145 | 0.827 ± 0.145 | 0.513 ± 0.115 | 0.158 ± 0.054 | 0.263 ± 0.071 | 0.333 ± 0.077 | 0.234 ± 0.026 | 0.005 ± 0.046 | 0.300 ± 0.076 | 0.0 ± 0.0 | 0.2 ± 0.4 | 1.0 ± 2.2 |
|
| 14 |
+
| Imperial (5) | 0.646 ± 0.026 | 0.646 ± 0.026 | 0.503 ± 0.022 | 0.283 ± 0.011 | 0.363 ± 0.016 | 0.749 ± 0.024 | 0.656 ± 0.009 | 0.061 ± 0.014 | 0.603 ± 0.033 | 30.8 ± 4.6 | 113.6 ± 6.3 | 74.8 ± 7.7 |
|
| 15 |
+
| Giroux (5) | 0.669 ± 0.124 | 0.669 ± 0.124 | 0.256 ± 0.030 | 0.095 ± 0.050 | 0.201 ± 0.088 | 0.297 ± 0.082 | 0.492 ± 0.085 | 0.029 ± 0.011 | 0.189 ± 0.054 | 0.0 ± 0.0 | 0.2 ± 0.4 | 0.0 ± 0.0 |
|
| 16 |
+
|
| 17 |
+
Draws where LoRA-5 beats CIBERSORTx: composition 24/25 (vs CIBERSORTx's better theta), logFC r 17/17, partial r 17/17, top-50 DEG 7/20 (ties 9).
|
| 18 |
+
|
| 19 |
+
## 5-fold CV, pooled (CIBERSORTx reference = each fold's LoRA training + inner-validation pairs)
|
| 20 |
+
|
| 21 |
+
| cohort | L1 BP final | L1 BP first | L1 LoRA | L1 scSemi | r BP | r BP (covered) | r LoRA | r scSemi | r broadcast | partial r BP | partial r LoRA | DEG BP | DEG LoRA | DEG scSemi | DEG broadcast |
|
| 22 |
+
|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
|
| 23 |
+
| ImMicro | 0.507 | 0.507 | 0.162 | 0.217 | – | – | – | – | – | – | – | – | – | – | – |
|
| 24 |
+
| Sepsis | 0.607 | 0.607 | 0.362 | 0.502 | -0.036 | -0.125 | 0.527 | 0.435 | 0.316 | -0.074 | 0.469 | 0 | 23 | 6 | 21 |
|
| 25 |
+
| Ji ChILI | 0.914 | 0.914 | 0.504 | 0.575 | – | – | – | 0.273 | – | – | – | – | – | 0 | – |
|
| 26 |
+
| Imperial | 0.592 | 0.592 | 0.376 | 0.520 | 0.211 | 0.369 | 0.819 | 0.705 | 0.661 | 0.016 | 0.690 | 1 | 143 | 87 | 84 |
|
| 27 |
+
| Giroux | 0.591 | 0.591 | 0.258 | 0.243 | – | – | – | 0.537 | – | – | – | – | – | 0 | – |
|
paper-releases/pbisc-diffusion/updates/20261008T050202Z-scsemi-imperial-deck/diffusion/dataset/server_1/model8/plans/bayesprism_paired_20261006/results/RESULTS_full.md
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# BayesPrism with the same paired donors vs LoRA (paired benchmark; gene mode full)
|
| 2 |
+
|
| 3 |
+
BayesPrism genes: its recommended set (canonical protein-coding minus chrX/Y/M and RPS/RPL/MRPS/MRPL, measured by the sc assay; 15-19k genes).
|
| 4 |
+
|
| 5 |
+
Same donors, observed truth, contrast and scorer as the benchmark (score_ft.py functions; the recomputed bulk broadcast equals the stored one in every split). BayesPrism reference = the split's paired donors' observed cells (states = gate | AIFI L2); mixture = the test samples' real bulk. Composition L1 in cell units (lower is better): 'final' and 'first' BayesPrism theta. logFC r = case-vs-control per-type logFC (mean CP10K) correlation with observed, mean of T/B/Mono/NK; BayesPrism Z zero-filled on genes it lacks ('covered' = only genes it models, broadcast recomputed there). DEG = top-50 overlap summed over types.
|
| 6 |
+
|
| 7 |
+
## Five-donor draws (mean ± SD over 5 draws)
|
| 8 |
+
|
| 9 |
+
| cohort | L1 BP final | L1 BP first | L1 LoRA-5 | r BP | r BP (covered) | r LoRA-5 | r broadcast | partial r BP | partial r LoRA-5 | DEG BP | DEG LoRA-5 | DEG broadcast |
|
| 10 |
+
|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
|
| 11 |
+
| ImMicro (5) | 0.594 ± 0.088 | 0.550 ± 0.164 | 0.176 ± 0.012 | – | – | – | – | – | – | – | – | – |
|
| 12 |
+
| Sepsis (5) | 0.643 ± 0.118 | 0.543 ± 0.218 | 0.466 ± 0.044 | 0.241 ± 0.035 | 0.258 ± 0.036 | 0.350 ± 0.162 | 0.320 ± 0.024 | 0.095 ± 0.035 | 0.305 ± 0.189 | 27.0 ± 2.0 | 16.4 ± 6.7 | 20.0 ± 1.9 |
|
| 13 |
+
| Ji ChILI (5) | 0.943 ± 0.117 | 0.952 ± 0.161 | 0.513 ± 0.115 | 0.193 ± 0.013 | 0.212 ± 0.013 | 0.333 ± 0.077 | 0.234 ± 0.026 | 0.032 ± 0.020 | 0.300 ± 0.076 | 2.0 ± 2.4 | 0.2 ± 0.4 | 1.0 ± 2.2 |
|
| 14 |
+
| Imperial (5) | 0.962 ± 0.086 | 0.845 ± 0.063 | 0.503 ± 0.022 | 0.598 ± 0.033 | 0.614 ± 0.034 | 0.749 ± 0.024 | 0.656 ± 0.009 | 0.187 ± 0.039 | 0.603 ± 0.033 | 72.4 ± 10.1 | 113.6 ± 6.3 | 74.8 ± 7.7 |
|
| 15 |
+
| Giroux (5) | 1.807 ± 0.047 | 1.455 ± 0.361 | 0.256 ± 0.030 | 0.371 ± 0.060 | 0.374 ± 0.060 | 0.297 ± 0.082 | 0.492 ± 0.085 | 0.117 ± 0.060 | 0.189 ± 0.054 | 0.2 ± 0.4 | 0.2 ± 0.4 | 0.0 ± 0.0 |
|
| 16 |
+
|
| 17 |
+
Draws where LoRA-5 beats BayesPrism: composition 22/25 (vs BayesPrism's better theta), logFC r 15/20, partial r 18/20, top-50 DEG 6/20 (ties 6).
|
| 18 |
+
|
| 19 |
+
## 5-fold CV, pooled (BayesPrism reference = each fold's LoRA training + inner-validation pairs)
|
| 20 |
+
|
| 21 |
+
| cohort | L1 BP final | L1 BP first | L1 LoRA | L1 scSemi | r BP | r BP (covered) | r LoRA | r scSemi | r broadcast | partial r BP | partial r LoRA | DEG BP | DEG LoRA | DEG scSemi | DEG broadcast |
|
| 22 |
+
|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
|
| 23 |
+
| ImMicro | 0.575 | 0.909 | 0.162 | 0.217 | – | – | – | – | – | – | – | – | – | – | – |
|
| 24 |
+
| Sepsis | 0.603 | 0.475 | 0.362 | 0.502 | 0.255 | 0.253 | 0.527 | 0.435 | 0.316 | 0.095 | 0.469 | 37 | 23 | 6 | 21 |
|
| 25 |
+
| Ji ChILI | 0.883 | 1.055 | 0.504 | 0.575 | 0.186 | 0.204 | 0.479 | 0.273 | 0.240 | 0.020 | 0.452 | 4 | 15 | 0 | 4 |
|
| 26 |
+
| Imperial | 0.887 | 1.022 | 0.376 | 0.520 | 0.568 | 0.584 | 0.819 | 0.705 | 0.661 | 0.112 | 0.690 | 91 | 143 | 87 | 84 |
|
| 27 |
+
| Giroux | 1.803 | 0.878 | 0.258 | 0.243 | 0.406 | 0.408 | 0.461 | 0.537 | 0.575 | 0.022 | 0.323 | 3 | 5 | 0 | 5 |
|
paper-releases/pbisc-diffusion/updates/20261008T050202Z-scsemi-imperial-deck/diffusion/dataset/server_1/model8/plans/bayesprism_paired_20261006/results/imperial_eskd/cv_pooled_cibersortx/summary.json
ADDED
|
@@ -0,0 +1,198 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"dataset": "imperial_eskd",
|
| 3 |
+
"comparison": "cv",
|
| 4 |
+
"draw": null,
|
| 5 |
+
"gene_mode": "cibersortx",
|
| 6 |
+
"stored_scores": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/imperial_eskd/pooled",
|
| 7 |
+
"lora_arm": "real_lora__raw",
|
| 8 |
+
"n_donors": 157,
|
| 9 |
+
"contrast": true,
|
| 10 |
+
"genes_scoring": 5927,
|
| 11 |
+
"genes_covered_by_Z_all_donors": 4120,
|
| 12 |
+
"bayesprism": {
|
| 13 |
+
"comp_l1": 0.5918271288164071,
|
| 14 |
+
"comp_l1_by_variant": {
|
| 15 |
+
"final_cells": 0.5918271288164071,
|
| 16 |
+
"first_cells": 0.5918271288164071,
|
| 17 |
+
"final_mrna": 0.6633005066634348,
|
| 18 |
+
"first_mrna": 0.6633005066634348
|
| 19 |
+
},
|
| 20 |
+
"mean_fraction": {
|
| 21 |
+
"final_cells": {
|
| 22 |
+
"T cell": 0.3685,
|
| 23 |
+
"B cell": 0.07,
|
| 24 |
+
"Monocyte": 0.324,
|
| 25 |
+
"NK cell": 0.1331,
|
| 26 |
+
"DC": 0.0376,
|
| 27 |
+
"Other": 0.0668
|
| 28 |
+
},
|
| 29 |
+
"first_cells": {
|
| 30 |
+
"T cell": 0.3685,
|
| 31 |
+
"B cell": 0.07,
|
| 32 |
+
"Monocyte": 0.324,
|
| 33 |
+
"NK cell": 0.1331,
|
| 34 |
+
"DC": 0.0376,
|
| 35 |
+
"Other": 0.0668
|
| 36 |
+
},
|
| 37 |
+
"final_mrna": {
|
| 38 |
+
"T cell": 0.3397,
|
| 39 |
+
"B cell": 0.0751,
|
| 40 |
+
"Monocyte": 0.3483,
|
| 41 |
+
"NK cell": 0.0983,
|
| 42 |
+
"DC": 0.0674,
|
| 43 |
+
"Other": 0.0712
|
| 44 |
+
},
|
| 45 |
+
"first_mrna": {
|
| 46 |
+
"T cell": 0.3397,
|
| 47 |
+
"B cell": 0.0751,
|
| 48 |
+
"Monocyte": 0.3483,
|
| 49 |
+
"NK cell": 0.0983,
|
| 50 |
+
"DC": 0.0674,
|
| 51 |
+
"Other": 0.0712
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"logfc_r_mean_cp10k": 0.21113497684457377,
|
| 55 |
+
"partial_r_mean_cp10k": 0.015667623661280307,
|
| 56 |
+
"logfc_r_mean_cp10k_covered_genes": 0.3693097049105145,
|
| 57 |
+
"deg_top50_sum": 1,
|
| 58 |
+
"deg_top50_sum_covered_genes": 3,
|
| 59 |
+
"per_type_logfc_r": {
|
| 60 |
+
"T cell": 0.1634,
|
| 61 |
+
"B cell": -0.0253,
|
| 62 |
+
"Monocyte": 0.4332,
|
| 63 |
+
"NK cell": 0.2733
|
| 64 |
+
}
|
| 65 |
+
},
|
| 66 |
+
"observed_mean_fraction": {
|
| 67 |
+
"T cell": 0.4973,
|
| 68 |
+
"B cell": 0.1192,
|
| 69 |
+
"Monocyte": 0.1692,
|
| 70 |
+
"NK cell": 0.1696,
|
| 71 |
+
"DC": 0.0303,
|
| 72 |
+
"Other": 0.0144
|
| 73 |
+
},
|
| 74 |
+
"lora": {
|
| 75 |
+
"comp_l1": 0.37638560875745664,
|
| 76 |
+
"coverage": 0.8500127757431224,
|
| 77 |
+
"logfc_r_mean_cp10k": 0.8187448084273437,
|
| 78 |
+
"deg_top50_sum": 143,
|
| 79 |
+
"partial_r_mean_cp10k": 0.6899314750539278
|
| 80 |
+
},
|
| 81 |
+
"broadcast": {
|
| 82 |
+
"logfc_r_mean_cp10k": 0.6606053132085035,
|
| 83 |
+
"logfc_r_mean_cp10k_covered_genes": 0.5831498842343769,
|
| 84 |
+
"deg_top50_sum": 84
|
| 85 |
+
},
|
| 86 |
+
"other_arms": {
|
| 87 |
+
"pseudo_full__cal": {
|
| 88 |
+
"comp_l1": 0.4613345498351681,
|
| 89 |
+
"logfc_r_mean_cp10k": 0.8081921526275624,
|
| 90 |
+
"deg_top50_sum": 129
|
| 91 |
+
},
|
| 92 |
+
"pseudo_full__raw": {
|
| 93 |
+
"comp_l1": 0.8128075891074484,
|
| 94 |
+
"logfc_r_mean_cp10k": 0.46880822702400493,
|
| 95 |
+
"deg_top50_sum": 54
|
| 96 |
+
},
|
| 97 |
+
"zs__cal": {
|
| 98 |
+
"comp_l1": 0.4943429956255003,
|
| 99 |
+
"logfc_r_mean_cp10k": 0.6787028713891314,
|
| 100 |
+
"deg_top50_sum": 111
|
| 101 |
+
},
|
| 102 |
+
"zs__pseudo": {
|
| 103 |
+
"comp_l1": 0.29976390097878974,
|
| 104 |
+
"logfc_r_mean_cp10k": 0.6815503704232228,
|
| 105 |
+
"deg_top50_sum": 101
|
| 106 |
+
},
|
| 107 |
+
"zs__raw": {
|
| 108 |
+
"comp_l1": 0.8077907232980434,
|
| 109 |
+
"logfc_r_mean_cp10k": 0.6439837621001347,
|
| 110 |
+
"deg_top50_sum": 106
|
| 111 |
+
}
|
| 112 |
+
},
|
| 113 |
+
"gates": {
|
| 114 |
+
"observed_composition_equal": true,
|
| 115 |
+
"broadcast_r_equal": true,
|
| 116 |
+
"broadcast_deg_equal": true
|
| 117 |
+
},
|
| 118 |
+
"method": "cibersortx",
|
| 119 |
+
"has_cell_type_expression": true,
|
| 120 |
+
"n_case": 113,
|
| 121 |
+
"n_control": 34,
|
| 122 |
+
"deg_per_type": {
|
| 123 |
+
"zero_filled": {
|
| 124 |
+
"bp": {
|
| 125 |
+
"T cell": 1,
|
| 126 |
+
"B cell": 0,
|
| 127 |
+
"Monocyte": 0,
|
| 128 |
+
"NK cell": 0,
|
| 129 |
+
"DC": 0
|
| 130 |
+
},
|
| 131 |
+
"broadcast": {
|
| 132 |
+
"T cell": 17,
|
| 133 |
+
"B cell": 14,
|
| 134 |
+
"Monocyte": 22,
|
| 135 |
+
"NK cell": 16,
|
| 136 |
+
"DC": 15
|
| 137 |
+
}
|
| 138 |
+
},
|
| 139 |
+
"covered": {
|
| 140 |
+
"bp": {
|
| 141 |
+
"T cell": 1,
|
| 142 |
+
"B cell": null,
|
| 143 |
+
"Monocyte": 2,
|
| 144 |
+
"NK cell": 0,
|
| 145 |
+
"DC": null
|
| 146 |
+
},
|
| 147 |
+
"broadcast": {
|
| 148 |
+
"T cell": 2,
|
| 149 |
+
"B cell": null,
|
| 150 |
+
"Monocyte": 3,
|
| 151 |
+
"NK cell": 0,
|
| 152 |
+
"DC": null
|
| 153 |
+
}
|
| 154 |
+
}
|
| 155 |
+
},
|
| 156 |
+
"hires_imputed_genes": {
|
| 157 |
+
"cv0|T cell": 4426,
|
| 158 |
+
"cv0|B cell": 2689,
|
| 159 |
+
"cv0|Monocyte": 1638,
|
| 160 |
+
"cv0|NK cell": 821,
|
| 161 |
+
"cv0|DC": 894,
|
| 162 |
+
"cv0|Other": 356,
|
| 163 |
+
"cv1|T cell": 4036,
|
| 164 |
+
"cv1|B cell": 0,
|
| 165 |
+
"cv1|Monocyte": 1454,
|
| 166 |
+
"cv1|NK cell": 1407,
|
| 167 |
+
"cv1|DC": 1807,
|
| 168 |
+
"cv1|Other": 740,
|
| 169 |
+
"cv2|T cell": 4094,
|
| 170 |
+
"cv2|B cell": 421,
|
| 171 |
+
"cv2|Monocyte": 2242,
|
| 172 |
+
"cv2|NK cell": 89,
|
| 173 |
+
"cv2|DC": 0,
|
| 174 |
+
"cv2|Other": 796,
|
| 175 |
+
"cv3|T cell": 3714,
|
| 176 |
+
"cv3|B cell": 1356,
|
| 177 |
+
"cv3|Monocyte": 1041,
|
| 178 |
+
"cv3|NK cell": 1271,
|
| 179 |
+
"cv3|DC": 1977,
|
| 180 |
+
"cv3|Other": 21,
|
| 181 |
+
"cv4|T cell": 3799,
|
| 182 |
+
"cv4|B cell": 2620,
|
| 183 |
+
"cv4|Monocyte": 2621,
|
| 184 |
+
"cv4|NK cell": 297,
|
| 185 |
+
"cv4|DC": 394,
|
| 186 |
+
"cv4|Other": 114
|
| 187 |
+
},
|
| 188 |
+
"scsemi_cv": {
|
| 189 |
+
"pooled": {
|
| 190 |
+
"comp_l1": 0.5203155086290455,
|
| 191 |
+
"logfc_r_mean_cp10k": 0.7053598964433999,
|
| 192 |
+
"deg_top50_sum": 87
|
| 193 |
+
}
|
| 194 |
+
},
|
| 195 |
+
"tool_sha256": "b541f68dc286497e8347d487af58ff3550a383d126eeabe347c0726e4fb28799",
|
| 196 |
+
"score_ft_sha256": "ec5f40b4cea6d5062d0138f3b934abb6465e34cbfeac869c8a0906924c033aec",
|
| 197 |
+
"seconds": 6.7
|
| 198 |
+
}
|
paper-releases/pbisc-diffusion/updates/20261008T050202Z-scsemi-imperial-deck/diffusion/dataset/server_1/model8/plans/bayesprism_paired_20261006/results/imperial_eskd/cv_pooled_full/summary.json
ADDED
|
@@ -0,0 +1,166 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"dataset": "imperial_eskd",
|
| 3 |
+
"comparison": "cv",
|
| 4 |
+
"draw": null,
|
| 5 |
+
"gene_mode": "full",
|
| 6 |
+
"stored_scores": "/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/paired_finetune_intradataset_20260927/scores/imperial_eskd/pooled",
|
| 7 |
+
"lora_arm": "real_lora__raw",
|
| 8 |
+
"n_donors": 157,
|
| 9 |
+
"contrast": true,
|
| 10 |
+
"genes_scoring": 5927,
|
| 11 |
+
"genes_covered_by_Z_all_donors": 5502,
|
| 12 |
+
"bayesprism": {
|
| 13 |
+
"comp_l1": 0.8870870830413495,
|
| 14 |
+
"comp_l1_by_variant": {
|
| 15 |
+
"final_cells": 0.8870870830413495,
|
| 16 |
+
"first_cells": 1.0215635566013643,
|
| 17 |
+
"final_mrna": 0.9300862395332192,
|
| 18 |
+
"first_mrna": 1.0602851494967092
|
| 19 |
+
},
|
| 20 |
+
"mean_fraction": {
|
| 21 |
+
"final_cells": {
|
| 22 |
+
"T cell": 0.1721,
|
| 23 |
+
"B cell": 0.0612,
|
| 24 |
+
"Monocyte": 0.4002,
|
| 25 |
+
"NK cell": 0.3518,
|
| 26 |
+
"DC": 0.0006,
|
| 27 |
+
"Other": 0.014
|
| 28 |
+
},
|
| 29 |
+
"first_cells": {
|
| 30 |
+
"T cell": 0.0801,
|
| 31 |
+
"B cell": 0.0973,
|
| 32 |
+
"Monocyte": 0.3567,
|
| 33 |
+
"NK cell": 0.2763,
|
| 34 |
+
"DC": 0.0018,
|
| 35 |
+
"Other": 0.1878
|
| 36 |
+
},
|
| 37 |
+
"final_mrna": {
|
| 38 |
+
"T cell": 0.149,
|
| 39 |
+
"B cell": 0.0657,
|
| 40 |
+
"Monocyte": 0.4861,
|
| 41 |
+
"NK cell": 0.2819,
|
| 42 |
+
"DC": 0.0012,
|
| 43 |
+
"Other": 0.0161
|
| 44 |
+
},
|
| 45 |
+
"first_mrna": {
|
| 46 |
+
"T cell": 0.066,
|
| 47 |
+
"B cell": 0.0997,
|
| 48 |
+
"Monocyte": 0.4149,
|
| 49 |
+
"NK cell": 0.212,
|
| 50 |
+
"DC": 0.0032,
|
| 51 |
+
"Other": 0.2041
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"logfc_r_mean_cp10k": 0.5684615839866303,
|
| 55 |
+
"partial_r_mean_cp10k": 0.11228628082302743,
|
| 56 |
+
"logfc_r_mean_cp10k_covered_genes": 0.5843506673991777,
|
| 57 |
+
"deg_top50_sum": 91,
|
| 58 |
+
"deg_top50_sum_covered_genes": 91,
|
| 59 |
+
"per_type_logfc_r": {
|
| 60 |
+
"T cell": 0.3768,
|
| 61 |
+
"B cell": 0.5227,
|
| 62 |
+
"Monocyte": 0.7265,
|
| 63 |
+
"NK cell": 0.6478
|
| 64 |
+
}
|
| 65 |
+
},
|
| 66 |
+
"observed_mean_fraction": {
|
| 67 |
+
"T cell": 0.4973,
|
| 68 |
+
"B cell": 0.1192,
|
| 69 |
+
"Monocyte": 0.1692,
|
| 70 |
+
"NK cell": 0.1696,
|
| 71 |
+
"DC": 0.0303,
|
| 72 |
+
"Other": 0.0144
|
| 73 |
+
},
|
| 74 |
+
"lora": {
|
| 75 |
+
"comp_l1": 0.37638560875745664,
|
| 76 |
+
"coverage": 0.8500127757431224,
|
| 77 |
+
"logfc_r_mean_cp10k": 0.8187448084273437,
|
| 78 |
+
"deg_top50_sum": 143,
|
| 79 |
+
"partial_r_mean_cp10k": 0.6899314750539278
|
| 80 |
+
},
|
| 81 |
+
"broadcast": {
|
| 82 |
+
"logfc_r_mean_cp10k": 0.6606053132085035,
|
| 83 |
+
"logfc_r_mean_cp10k_covered_genes": 0.6388431994811914,
|
| 84 |
+
"deg_top50_sum": 84
|
| 85 |
+
},
|
| 86 |
+
"other_arms": {
|
| 87 |
+
"pseudo_full__cal": {
|
| 88 |
+
"comp_l1": 0.4613345498351681,
|
| 89 |
+
"logfc_r_mean_cp10k": 0.8081921526275624,
|
| 90 |
+
"deg_top50_sum": 129
|
| 91 |
+
},
|
| 92 |
+
"pseudo_full__raw": {
|
| 93 |
+
"comp_l1": 0.8128075891074484,
|
| 94 |
+
"logfc_r_mean_cp10k": 0.46880822702400493,
|
| 95 |
+
"deg_top50_sum": 54
|
| 96 |
+
},
|
| 97 |
+
"zs__cal": {
|
| 98 |
+
"comp_l1": 0.4943429956255003,
|
| 99 |
+
"logfc_r_mean_cp10k": 0.6787028713891314,
|
| 100 |
+
"deg_top50_sum": 111
|
| 101 |
+
},
|
| 102 |
+
"zs__pseudo": {
|
| 103 |
+
"comp_l1": 0.29976390097878974,
|
| 104 |
+
"logfc_r_mean_cp10k": 0.6815503704232228,
|
| 105 |
+
"deg_top50_sum": 101
|
| 106 |
+
},
|
| 107 |
+
"zs__raw": {
|
| 108 |
+
"comp_l1": 0.8077907232980434,
|
| 109 |
+
"logfc_r_mean_cp10k": 0.6439837621001347,
|
| 110 |
+
"deg_top50_sum": 106
|
| 111 |
+
}
|
| 112 |
+
},
|
| 113 |
+
"gates": {
|
| 114 |
+
"observed_composition_equal": true,
|
| 115 |
+
"broadcast_r_equal": true,
|
| 116 |
+
"broadcast_deg_equal": true
|
| 117 |
+
},
|
| 118 |
+
"method": "bayesprism",
|
| 119 |
+
"has_cell_type_expression": true,
|
| 120 |
+
"n_case": 113,
|
| 121 |
+
"n_control": 34,
|
| 122 |
+
"deg_per_type": {
|
| 123 |
+
"zero_filled": {
|
| 124 |
+
"bp": {
|
| 125 |
+
"T cell": 15,
|
| 126 |
+
"B cell": 7,
|
| 127 |
+
"Monocyte": 28,
|
| 128 |
+
"NK cell": 22,
|
| 129 |
+
"DC": 19
|
| 130 |
+
},
|
| 131 |
+
"broadcast": {
|
| 132 |
+
"T cell": 17,
|
| 133 |
+
"B cell": 14,
|
| 134 |
+
"Monocyte": 22,
|
| 135 |
+
"NK cell": 16,
|
| 136 |
+
"DC": 15
|
| 137 |
+
}
|
| 138 |
+
},
|
| 139 |
+
"covered": {
|
| 140 |
+
"bp": {
|
| 141 |
+
"T cell": 15,
|
| 142 |
+
"B cell": 7,
|
| 143 |
+
"Monocyte": 28,
|
| 144 |
+
"NK cell": 22,
|
| 145 |
+
"DC": 19
|
| 146 |
+
},
|
| 147 |
+
"broadcast": {
|
| 148 |
+
"T cell": 15,
|
| 149 |
+
"B cell": 12,
|
| 150 |
+
"Monocyte": 21,
|
| 151 |
+
"NK cell": 14,
|
| 152 |
+
"DC": 16
|
| 153 |
+
}
|
| 154 |
+
}
|
| 155 |
+
},
|
| 156 |
+
"scsemi_cv": {
|
| 157 |
+
"pooled": {
|
| 158 |
+
"comp_l1": 0.5203155086290455,
|
| 159 |
+
"logfc_r_mean_cp10k": 0.7053598964433999,
|
| 160 |
+
"deg_top50_sum": 87
|
| 161 |
+
}
|
| 162 |
+
},
|
| 163 |
+
"tool_sha256": "b541f68dc286497e8347d487af58ff3550a383d126eeabe347c0726e4fb28799",
|
| 164 |
+
"score_ft_sha256": "ec5f40b4cea6d5062d0138f3b934abb6465e34cbfeac869c8a0906924c033aec",
|
| 165 |
+
"seconds": 8.4
|
| 166 |
+
}
|