Archive 4 checkpoints (1-4)
Browse files- paper-releases/pbisc-diffusion/updates/20261005T175437Z-deck-supplementary-other-tissues-review/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/sec08_cross_cohort_slide_text.json +133 -0
- paper-releases/pbisc-diffusion/updates/20261005T175437Z-deck-supplementary-other-tissues-review/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/sec10_other_tissues.csv +9 -0
- paper-releases/pbisc-diffusion/updates/20261005T175437Z-deck-supplementary-other-tissues-review/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/secS_scratch_rare_states.csv +73 -0
- paper-releases/pbisc-diffusion/updates/20261005T175437Z-deck-supplementary-other-tissues-review/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/secS_scratch_rare_states_bootstrap.csv +169 -0
paper-releases/pbisc-diffusion/updates/20261005T175437Z-deck-supplementary-other-tissues-review/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/sec08_cross_cohort_slide_text.json
ADDED
|
@@ -0,0 +1,133 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"section": "08-cross-cohort",
|
| 3 |
+
"scripts": [
|
| 4 |
+
"/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/tools/sec08_cross_cohort.py"
|
| 5 |
+
],
|
| 6 |
+
"generated_from": "tools/sec08_cross_cohort.py (rerun to refresh; do not edit by hand)",
|
| 7 |
+
"slides": [
|
| 8 |
+
{
|
| 9 |
+
"title": "Pairs from other cohorts do not transfer",
|
| 10 |
+
"layout": "full",
|
| 11 |
+
"figures": [
|
| 12 |
+
"/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/sec08_cross_cohort.png"
|
| 13 |
+
],
|
| 14 |
+
"keywords": [
|
| 15 |
+
"Train on the other cohorts, test the held-out one",
|
| 16 |
+
"Calibration ties or beats them on composition in 4/5",
|
| 17 |
+
"Own-cohort pairs best on coverage 5/5, disease r 4/4"
|
| 18 |
+
],
|
| 19 |
+
"footnote_short": "Leave-one-cohort-out, K = 1000; train on the other cohorts (+ Liu 2021), test the held-out one.",
|
| 20 |
+
"footnote": "Each held-out cohort is scored on all its usable samples against the same frozen reference as its within-cohort benchmark. Zero-shot = base model on real bulk. Calibration from other cohorts = gene-wise alpha (mean pseudobulk CPM / mean bulk CPM) fitted on the training cohorts' pairs, cohort-balanced; 0 target pairs. LoRA r16 / r64 / full fine-tune = trained on the other PBMC cohorts' pairs plus Liu 2021 (training only: its cells are in the pretraining corpus; 44 pairs after excluding 14 hashtag-negative-heavy rows), cohort-balanced batches, inner-validation on 10 % of each training cohort's donors, step-saturation protocol (20k-step blocks, stop when a block lowers the inner-validation loss by < 1 % of the total), value at the last milestone (number in the bar = steps). Ji's bulk is whole blood and is never trained on. Own-cohort LoRA = LoRA r16 on the cohort's own pairs, 5-fold pooled. Giroux's logFC r excludes 196 sex-linked genes. Best on composition among zero-target-pair arms: ImMicro Calibration from other cohorts, Sepsis Calibration from other cohorts, Imperial Full fine-tune, other cohorts, Giroux Calibration from other cohorts, Ji (whole blood) Calibration from other cohorts.",
|
| 21 |
+
"numbers": [
|
| 22 |
+
{
|
| 23 |
+
"value": "ImMicro: L1 zs 0.446 / xcal 0.172 / lora16sat 0.175 / lora64sat 0.199 / fullsat 0.213 / own 0.162; coverage zs 0.19 / xcal 0.57 / lora16sat 0.45 / lora64sat 0.38 / fullsat 0.35 / own 0.78",
|
| 24 |
+
"source": "data/sec08_cross_cohort.csv"
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"value": "Sepsis: L1 zs 0.509 / xcal 0.444 / lora16sat 0.491 / lora64sat 0.570 / fullsat 0.585 / own 0.362; coverage zs 0.07 / xcal 0.46 / lora16sat 0.34 / lora64sat 0.29 / fullsat 0.35 / own 0.83; logFC r zs -0.02 / xcal 0.10 / lora16sat 0.10 / lora64sat 0.23 / fullsat 0.07 / own 0.53",
|
| 28 |
+
"source": "data/sec08_cross_cohort.csv"
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"value": "Imperial: L1 zs 0.803 / xcal 0.651 / lora16sat 0.635 / lora64sat 0.594 / fullsat 0.580 / own 0.376; coverage zs 0.11 / xcal 0.64 / lora16sat 0.68 / lora64sat 0.72 / fullsat 0.61 / own 0.85; logFC r zs 0.64 / xcal 0.69 / lora16sat 0.49 / lora64sat 0.58 / fullsat 0.46 / own 0.82",
|
| 32 |
+
"source": "data/sec08_cross_cohort.csv"
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"value": "Giroux: L1 zs 0.337 / xcal 0.202 / lora16sat 0.431 / lora64sat 0.410 / fullsat 0.465 / own 0.258; coverage zs 0.29 / xcal 0.89 / lora16sat 0.82 / lora64sat 0.81 / fullsat 0.78 / own 0.93; logFC r zs 0.34 / xcal 0.24 / lora16sat 0.30 / lora64sat 0.28 / fullsat 0.20 / own 0.46",
|
| 36 |
+
"source": "data/sec08_cross_cohort.csv"
|
| 37 |
+
},
|
| 38 |
+
{
|
| 39 |
+
"value": "Ji (whole blood): L1 zs 0.725 / xcal 0.565 / lora16sat 0.740 / lora64sat 0.620 / fullsat 1.030 / own 0.504; coverage zs 0.06 / xcal 0.07 / lora16sat 0.34 / lora64sat 0.40 / fullsat 0.08 / own 0.86; logFC r zs 0.07 / xcal 0.13 / lora16sat 0.05 / lora64sat 0.16 / fullsat 0.14 / own 0.48",
|
| 40 |
+
"source": "data/sec08_cross_cohort.csv"
|
| 41 |
+
}
|
| 42 |
+
]
|
| 43 |
+
},
|
| 44 |
+
{
|
| 45 |
+
"title": "Longer training on other cohorts: gap stays",
|
| 46 |
+
"layout": "full",
|
| 47 |
+
"figures": [
|
| 48 |
+
"/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/presentations/2026-09-30--paper-summary--r00/figures/sec08_cross_cohort_steps.png"
|
| 49 |
+
],
|
| 50 |
+
"keywords": [
|
| 51 |
+
"To saturation: r16 20–31 min, full 59–107 min",
|
| 52 |
+
"Full fine-tune drifts in Sepsis and Ji (Ji L1 0.84 → 1.03)",
|
| 53 |
+
"So PBISC fits LoRA on a few pairs per cohort"
|
| 54 |
+
],
|
| 55 |
+
"footnote_short": "Test metrics at every 20k-step milestone (best inner-validation state so far). K = 1000.",
|
| 56 |
+
"footnote": "Same runs as the previous slide. Each point = the milestone checkpoint at the end of a 20k-step block (or at the early stop), sampled and scored on the held-out cohort. Dashed = calibration from other cohorts; dotted = LoRA on the cohort's own pairs. Training stopped (steps): LoRA r16 ImMicro 28k, Sepsis 26k, Imperial 40k, Giroux 35k, Ji 40k; LoRA r64 ImMicro 39k, Sepsis 59k, Imperial 69k, Giroux 69k, Ji 51k; Full fine-tune ImMicro 65k, Sepsis 108k, Imperial 93k, Giroux 120k, Ji 86k. Saturation steps across cohorts: r16 up to 40k, r64 up to 69k, full up to 120k. Training time to saturation (RUN.json wall seconds, one GB10): LoRA r16 ImMicro 21 min, Sepsis 20 min, Imperial 30 min, Giroux 28 min, Ji 31 min; LoRA r64 ImMicro 27 min, Sepsis 44 min, Imperial 54 min, Giroux 53 min, Ji 41 min; Full fine-tune ImMicro 59 min, Sepsis 93 min, Imperial 82 min, Giroux 107 min, Ji 77 min. Gap = no milestone of any mode reaches the own-cohort LoRA's coverage (true in all five cohorts); full fine-tune composition L1 from the first to the last milestone: ImMicro +0.005, Sepsis +0.087, Imperial -0.052, Giroux -0.078, Ji +0.192 (drift >= 0.05 in Sepsis and Ji; improved by >= 0.05 in Imperial and Giroux).",
|
| 57 |
+
"numbers": [
|
| 58 |
+
{
|
| 59 |
+
"value": "ImMicro LoRA r16: 20k L1 0.163 cov 0.45, 28k L1 0.175 cov 0.45",
|
| 60 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"value": "ImMicro LoRA r64: 20k L1 0.196 cov 0.44, 39k L1 0.199 cov 0.38",
|
| 64 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"value": "ImMicro Full fine-tune: 20k L1 0.208 cov 0.39, 40k L1 0.211 cov 0.37, 60k L1 0.223 cov 0.35, 65k L1 0.213 cov 0.35",
|
| 68 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"value": "Sepsis LoRA r16: 20k L1 0.500 cov 0.35, 26k L1 0.491 cov 0.34",
|
| 72 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"value": "Sepsis LoRA r64: 20k L1 0.550 cov 0.29, 40k L1 0.564 cov 0.29, 59k L1 0.570 cov 0.29",
|
| 76 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"value": "Sepsis Full fine-tune: 20k L1 0.498 cov 0.33, 40k L1 0.530 cov 0.33, 60k L1 0.546 cov 0.34, 80k L1 0.576 cov 0.34, 100k L1 0.577 cov 0.35, 108k L1 0.585 cov 0.35",
|
| 80 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"value": "Imperial LoRA r16: 20k L1 0.657 cov 0.68, 40k L1 0.635 cov 0.68",
|
| 84 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"value": "Imperial LoRA r64: 20k L1 0.622 cov 0.68, 40k L1 0.603 cov 0.71, 60k L1 0.577 cov 0.72, 69k L1 0.594 cov 0.72",
|
| 88 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"value": "Imperial Full fine-tune: 20k L1 0.632 cov 0.60, 40k L1 0.609 cov 0.60, 60k L1 0.565 cov 0.61, 80k L1 0.576 cov 0.60, 93k L1 0.580 cov 0.61",
|
| 92 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"value": "Giroux LoRA r16: 20k L1 0.416 cov 0.79, 35k L1 0.431 cov 0.82",
|
| 96 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 97 |
+
},
|
| 98 |
+
{
|
| 99 |
+
"value": "Giroux LoRA r64: 20k L1 0.491 cov 0.80, 40k L1 0.469 cov 0.82, 60k L1 0.430 cov 0.82, 69k L1 0.410 cov 0.81",
|
| 100 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"value": "Giroux Full fine-tune: 20k L1 0.543 cov 0.75, 40k L1 0.532 cov 0.75, 60k L1 0.499 cov 0.76, 80k L1 0.466 cov 0.76, 100k L1 0.490 cov 0.76, 120k L1 0.465 cov 0.78",
|
| 104 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"value": "Ji (whole blood) LoRA r16: 20k L1 0.757 cov 0.27, 40k L1 0.740 cov 0.34",
|
| 108 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"value": "Ji (whole blood) LoRA r64: 20k L1 0.639 cov 0.38, 40k L1 0.668 cov 0.39, 51k L1 0.620 cov 0.40",
|
| 112 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 113 |
+
},
|
| 114 |
+
{
|
| 115 |
+
"value": "Ji (whole blood) Full fine-tune: 20k L1 0.838 cov 0.17, 40k L1 1.006 cov 0.11, 60k L1 1.032 cov 0.08, 80k L1 1.052 cov 0.07, 86k L1 1.030 cov 0.08",
|
| 116 |
+
"source": "data/sec08_cross_cohort_steps.csv"
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"value": "Training time to saturation, LoRA r16: ImMicro 21 min, Sepsis 20 min, Imperial 30 min, Giroux 28 min, Ji 31 min",
|
| 120 |
+
"source": "data/sec08_cross_cohort.csv (minutes = RUN.json wall_seconds / 60)"
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"value": "Training time to saturation, LoRA r64: ImMicro 27 min, Sepsis 44 min, Imperial 54 min, Giroux 53 min, Ji 41 min",
|
| 124 |
+
"source": "data/sec08_cross_cohort.csv (minutes = RUN.json wall_seconds / 60)"
|
| 125 |
+
},
|
| 126 |
+
{
|
| 127 |
+
"value": "Training time to saturation, Full fine-tune: ImMicro 59 min, Sepsis 93 min, Imperial 82 min, Giroux 107 min, Ji 77 min",
|
| 128 |
+
"source": "data/sec08_cross_cohort.csv (minutes = RUN.json wall_seconds / 60)"
|
| 129 |
+
}
|
| 130 |
+
]
|
| 131 |
+
}
|
| 132 |
+
]
|
| 133 |
+
}
|
paper-releases/pbisc-diffusion/updates/20261005T175437Z-deck-supplementary-other-tissues-review/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/sec10_other_tissues.csv
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
row,pretrained_lora,from_scratch,better,source
|
| 2 |
+
Corpus pretraining,"33.1 M train cells, 15.9 h, once",none,,sec01_model_numbers.csv train_split_cells; sec06_ft_method.csv pretrain_wall_hours
|
| 3 |
+
"Training, sepsis fold 0 (one GPU)",3–5 min,61–86 min,pre,sec07_no_pretraining.csv minutes
|
| 4 |
+
5 pairs: donor composition r,0.70,0.17,pre,sec07_no_pretraining.csv comp_r
|
| 5 |
+
58 pairs + recipes: donor composition r,0.80,0.77,,sec07_no_pretraining.csv comp_r
|
| 6 |
+
58 pairs + recipes: coverage,0.87,0.88,,sec07_no_pretraining.csv coverage
|
| 7 |
+
58 pairs + recipes: disease logFC r,0.45,0.54,scr,sec07_no_pretraining.csv logfc_r
|
| 8 |
+
"58 pairs + recipes: 13 rare states, n.s. on average",closer on ISG-high T,closer on DN2 B,,secS_scratch_rare_states_bootstrap.csv
|
| 9 |
+
Other cohorts' pairs only (sepsis held out): coverage,0.34 (LoRA r16),0.33,,sec08_cross_scratch.csv coverage (masked)
|
paper-releases/pbisc-diffusion/updates/20261005T175437Z-deck-supplementary-other-tissues-review/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/secS_scratch_rare_states.csv
ADDED
|
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
panel,setting,init,arm,state,gen_over_obs,source
|
| 2 |
+
same cohort,5 pairs,pretrained + LoRA,lora_n5__raw,C1Q-high monocyte (reference),1.0748895160897185,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 3 |
+
same cohort,5 pairs,from scratch,scratch_n5__raw,C1Q-high monocyte (reference),1.0051456341688143,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 4 |
+
same cohort,5 pairs,pretrained + LoRA,lora_n5__raw,MS1 monocyte (IL1R2+),0.1519091168143733,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 5 |
+
same cohort,5 pairs,from scratch,scratch_n5__raw,MS1 monocyte (IL1R2+),0.0956144201318488,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 6 |
+
same cohort,5 pairs,pretrained + LoRA,lora_n5__raw,Cycling monocyte (Ye 2026),0.1118744839276425,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 7 |
+
same cohort,5 pairs,from scratch,scratch_n5__raw,Cycling monocyte (Ye 2026),0.0613771980274272,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 8 |
+
same cohort,5 pairs,pretrained + LoRA,lora_n5__raw,Cycling T/NK,1.1540403567864272,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 9 |
+
same cohort,5 pairs,from scratch,scratch_n5__raw,Cycling T/NK,0.5082497711353814,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 10 |
+
same cohort,5 pairs,pretrained + LoRA,lora_n5__raw,ISG-high T,0.4568528677112091,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 11 |
+
same cohort,5 pairs,from scratch,scratch_n5__raw,ISG-high T,0.1456249982993058,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 12 |
+
same cohort,5 pairs,pretrained + LoRA,lora_n5__raw,Plasmablast (label),1.874440793077007,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 13 |
+
same cohort,5 pairs,from scratch,scratch_n5__raw,Plasmablast (label),1.1246644758462043,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 14 |
+
same cohort,5 pairs,pretrained + LoRA,lora_n5__raw,DN2 / ABC B,1.106914735930293,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 15 |
+
same cohort,5 pairs,from scratch,scratch_n5__raw,DN2 / ABC B,1.6346401785938245,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 16 |
+
same cohort,5 pairs,pretrained + LoRA,lora_n5__raw,Granulocyte-like (LDG),0.0545427046558268,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 17 |
+
same cohort,5 pairs,from scratch,scratch_n5__raw,Granulocyte-like (LDG),0.1287207829877513,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 18 |
+
same cohort,58 pairs + recipes,pretrained + LoRA,lora_recipe__raw,C1Q-high monocyte (reference),0.6257734890436245,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 19 |
+
same cohort,58 pairs + recipes,from scratch,scratch_recipe__raw,C1Q-high monocyte (reference),0.5324998918978538,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 20 |
+
same cohort,58 pairs + recipes,pretrained + LoRA,lora_recipe__raw,MS1 monocyte (IL1R2+),0.3010185021290256,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 21 |
+
same cohort,58 pairs + recipes,from scratch,scratch_recipe__raw,MS1 monocyte (IL1R2+),0.4620999194245931,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 22 |
+
same cohort,58 pairs + recipes,pretrained + LoRA,lora_recipe__raw,Cycling monocyte (Ye 2026),0.096242802121991,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 23 |
+
same cohort,58 pairs + recipes,from scratch,scratch_recipe__raw,Cycling monocyte (Ye 2026),0.0869463596409315,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 24 |
+
same cohort,58 pairs + recipes,pretrained + LoRA,lora_recipe__raw,Cycling T/NK,0.9873293455925888,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 25 |
+
same cohort,58 pairs + recipes,from scratch,scratch_recipe__raw,Cycling T/NK,0.4790315538632603,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 26 |
+
same cohort,58 pairs + recipes,pretrained + LoRA,lora_recipe__raw,ISG-high T,0.5329305241249853,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 27 |
+
same cohort,58 pairs + recipes,from scratch,scratch_recipe__raw,ISG-high T,0.183511576926034,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 28 |
+
same cohort,58 pairs + recipes,pretrained + LoRA,lora_recipe__raw,Plasmablast (label),2.299794665352174,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 29 |
+
same cohort,58 pairs + recipes,from scratch,scratch_recipe__raw,Plasmablast (label),1.1751301895059698,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 30 |
+
same cohort,58 pairs + recipes,pretrained + LoRA,lora_recipe__raw,DN2 / ABC B,0.2826701429139615,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 31 |
+
same cohort,58 pairs + recipes,from scratch,scratch_recipe__raw,DN2 / ABC B,0.4592885537544296,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 32 |
+
same cohort,58 pairs + recipes,pretrained + LoRA,lora_recipe__raw,Granulocyte-like (LDG),0.0252858299372933,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 33 |
+
same cohort,58 pairs + recipes,from scratch,scratch_recipe__raw,Granulocyte-like (LDG),0.0585732288086649,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_fold0_summary.csv
|
| 34 |
+
other cohorts,"held-out sepsis, masked",Own-cohort LoRA (reference),within_real_lora__raw,C1Q-high monocyte (reference),0.5097477488339228,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 35 |
+
other cohorts,"held-out sepsis, masked",Own-cohort LoRA (reference),within_real_lora__raw,MS1 monocyte (IL1R2+),0.7152499696281058,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 36 |
+
other cohorts,"held-out sepsis, masked",Own-cohort LoRA (reference),within_real_lora__raw,Cycling monocyte (Ye 2026),0.2364276312802053,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 37 |
+
other cohorts,"held-out sepsis, masked",Own-cohort LoRA (reference),within_real_lora__raw,Cycling T/NK,0.5240570256538553,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 38 |
+
other cohorts,"held-out sepsis, masked",Own-cohort LoRA (reference),within_real_lora__raw,ISG-high T,0.579747348736621,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 39 |
+
other cohorts,"held-out sepsis, masked",Own-cohort LoRA (reference),within_real_lora__raw,Plasmablast (label),0.7550341795451672,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 40 |
+
other cohorts,"held-out sepsis, masked",Own-cohort LoRA (reference),within_real_lora__raw,DN2 / ABC B,0.2861262573588353,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 41 |
+
other cohorts,"held-out sepsis, masked",Own-cohort LoRA (reference),within_real_lora__raw,Granulocyte-like (LDG),0.0612618062971696,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 42 |
+
other cohorts,"held-out sepsis, masked",Calibration,xcalL_K1000,C1Q-high monocyte (reference),0.0756199128382113,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 43 |
+
other cohorts,"held-out sepsis, masked",Calibration,xcalL_K1000,MS1 monocyte (IL1R2+),0.7253936473575063,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 44 |
+
other cohorts,"held-out sepsis, masked",Calibration,xcalL_K1000,Cycling monocyte (Ye 2026),0.2563415841564834,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 45 |
+
other cohorts,"held-out sepsis, masked",Calibration,xcalL_K1000,Cycling T/NK,0.4789398200183921,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 46 |
+
other cohorts,"held-out sepsis, masked",Calibration,xcalL_K1000,ISG-high T,1.540999282816366,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 47 |
+
other cohorts,"held-out sepsis, masked",Calibration,xcalL_K1000,Plasmablast (label),0.7829141153384702,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 48 |
+
other cohorts,"held-out sepsis, masked",Calibration,xcalL_K1000,DN2 / ABC B,0.2844684111816357,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 49 |
+
other cohorts,"held-out sepsis, masked",Calibration,xcalL_K1000,Granulocyte-like (LDG),0.0670251882295056,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 50 |
+
other cohorts,"held-out sepsis, masked",LoRA r16,locoL_lora16sat_s26750_K1000,C1Q-high monocyte (reference),0.0379077438589531,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 51 |
+
other cohorts,"held-out sepsis, masked",LoRA r16,locoL_lora16sat_s26750_K1000,MS1 monocyte (IL1R2+),0.1234847151564598,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 52 |
+
other cohorts,"held-out sepsis, masked",LoRA r16,locoL_lora16sat_s26750_K1000,Cycling monocyte (Ye 2026),0.0732107677203869,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 53 |
+
other cohorts,"held-out sepsis, masked",LoRA r16,locoL_lora16sat_s26750_K1000,Cycling T/NK,0.1324783671914411,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 54 |
+
other cohorts,"held-out sepsis, masked",LoRA r16,locoL_lora16sat_s26750_K1000,ISG-high T,1.7106889714081193,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 55 |
+
other cohorts,"held-out sepsis, masked",LoRA r16,locoL_lora16sat_s26750_K1000,Plasmablast (label),1.3763584629387786,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 56 |
+
other cohorts,"held-out sepsis, masked",LoRA r16,locoL_lora16sat_s26750_K1000,DN2 / ABC B,0.144552472274974,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 57 |
+
other cohorts,"held-out sepsis, masked",LoRA r16,locoL_lora16sat_s26750_K1000,Granulocyte-like (LDG),0.0168835070232936,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 58 |
+
other cohorts,"held-out sepsis, masked",Full FT,locoL_fullsat_s108500_K1000,C1Q-high monocyte (reference),0.0836144742186387,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 59 |
+
other cohorts,"held-out sepsis, masked",Full FT,locoL_fullsat_s108500_K1000,MS1 monocyte (IL1R2+),0.2720343761735265,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 60 |
+
other cohorts,"held-out sepsis, masked",Full FT,locoL_fullsat_s108500_K1000,Cycling monocyte (Ye 2026),0.1215710447520997,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 61 |
+
other cohorts,"held-out sepsis, masked",Full FT,locoL_fullsat_s108500_K1000,Cycling T/NK,0.3601223567794363,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 62 |
+
other cohorts,"held-out sepsis, masked",Full FT,locoL_fullsat_s108500_K1000,ISG-high T,1.979966388121804,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 63 |
+
other cohorts,"held-out sepsis, masked",Full FT,locoL_fullsat_s108500_K1000,Plasmablast (label),2.205359819078424,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 64 |
+
other cohorts,"held-out sepsis, masked",Full FT,locoL_fullsat_s108500_K1000,DN2 / ABC B,0.2858536073071018,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 65 |
+
other cohorts,"held-out sepsis, masked",Full FT,locoL_fullsat_s108500_K1000,Granulocyte-like (LDG),0.0112905444991257,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 66 |
+
other cohorts,"held-out sepsis, masked",From scratch,locoL_scratchsat_s162000_K1000,C1Q-high monocyte (reference),0.0730704382830807,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 67 |
+
other cohorts,"held-out sepsis, masked",From scratch,locoL_scratchsat_s162000_K1000,MS1 monocyte (IL1R2+),0.1610585398379573,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 68 |
+
other cohorts,"held-out sepsis, masked",From scratch,locoL_scratchsat_s162000_K1000,Cycling monocyte (Ye 2026),0.036144289390185,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 69 |
+
other cohorts,"held-out sepsis, masked",From scratch,locoL_scratchsat_s162000_K1000,Cycling T/NK,0.0753149959777236,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 70 |
+
other cohorts,"held-out sepsis, masked",From scratch,locoL_scratchsat_s162000_K1000,ISG-high T,1.101270082650952,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 71 |
+
other cohorts,"held-out sepsis, masked",From scratch,locoL_scratchsat_s162000_K1000,Plasmablast (label),0.8728126838148352,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 72 |
+
other cohorts,"held-out sepsis, masked",From scratch,locoL_scratchsat_s162000_K1000,DN2 / ABC B,0.2658286721720235,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
| 73 |
+
other cohorts,"held-out sepsis, masked",From scratch,locoL_scratchsat_s162000_K1000,Granulocyte-like (LDG),0.0207409780840793,/home/sj_server_1/PBISC/immun-pbmc-fm/model_8_sparse_data_diffusion/plans/rare_states_screen_20261004/out/sepsis_loco_masked_summary.csv
|
paper-releases/pbisc-diffusion/updates/20261005T175437Z-deck-supplementary-other-tissues-review/diffusion/dataset/server_1/model8/presentations/2026-09-30--paper-summary--r00/data/secS_scratch_rare_states_bootstrap.csv
ADDED
|
@@ -0,0 +1,169 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
scope,comparison,state,D,lo,hi,verdict,gen_obs_scratch,gen_obs_pretrained,both_fail
|
| 2 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,C1Q-high monocyte (reference),-0.09638436565278168,-0.40334494118252984,0.503738654466638,n.s.,1.0051456341688143,1.0748895160897185,False
|
| 3 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,Cycling T/NK,0.7407459856596705,-0.4335843856163559,1.7010584397129025,n.s.,0.5082497711353814,1.1540403567864272,False
|
| 4 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,Cycling activated T (CD38/HLA-DR),0.1348984054856852,-0.37133135502498055,0.5678164057375715,n.s.,0.4036934520495799,0.4432616631047784,False
|
| 5 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,Proliferating T (label),-0.8633209988729602,-1.273426124454765,-0.45923905189985353,scratch closer,1.3662885940822904,2.489230760285939,False
|
| 6 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,Plasmablast (label),-0.7351207153458899,-1.1128940564059335,0.04245892693745224,n.s.,1.1246644758462043,1.874440793077007,False
|
| 7 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,Plasmablast (markers),0.5223421471283156,-0.08200037352811988,0.9911329631590571,n.s.,0.3499864638546936,0.5030108484838021,False
|
| 8 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,MS1 monocyte (IL1R2+),0.6654756139651865,-0.45121555659765816,1.2260593437047718,n.s.,0.0956144201318488,0.1519091168143733,True
|
| 9 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,Cycling monocyte (Ye 2026),0.8453441140557731,-0.4831358413116413,2.226372851215401,n.s.,0.0613771980274272,0.1118744839276425,True
|
| 10 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,ISG-high T,1.63292526943556,-0.1603939651742089,3.0408663361428254,n.s.,0.1456249982993058,0.4568528677112091,False
|
| 11 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,DN2 / ABC B,0.5621383873017773,-0.5026131994607549,0.8075359933114971,n.s.,1.6346401785938245,1.106914735930293,False
|
| 12 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,HSPC,0.9730329523997308,-0.5799955807683735,1.9356754782070302,n.s.,0.4793212885154062,0.9586425770308123,False
|
| 13 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,Cytotoxic CD4 T,-0.2424073878912305,-0.7487359238657371,0.60884390845494,n.s.,1.5159448029096037,1.793379478458955,False
|
| 14 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,Granulocyte-like (LDG),-1.230531766146465,-2.073124387057781,-0.3268809504107658,scratch closer,0.1287207829877513,0.0545427046558268,True
|
| 15 |
+
same cohort (fold 0),5 pairs: scratch vs pretrained,MEAN of 13 rare states,0.22377981857864399,-0.07807423645532428,0.4055025835760699,n.s.,,,False
|
| 16 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,C1Q-high monocyte (reference),1.0144443440139763,-0.2644976312582007,1.3425387936606266,n.s.,0.4669732296338602,1.0533623276746888,False
|
| 17 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,Cycling T/NK,1.1474496357267074,0.20029372692540598,1.7826469808355645,pretrained closer,0.4159012861180642,0.9315431693091268,False
|
| 18 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,Cycling activated T (CD38/HLA-DR),0.37221080998747036,-0.3699066360852732,0.9577408383194574,n.s.,0.3149812026848403,0.4088720702830687,False
|
| 19 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,Proliferating T (label),-0.40055126941507,-0.8151013272719655,0.20407495783315543,n.s.,1.127184984582887,1.4897925068632638,False
|
| 20 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,Plasmablast (label),-1.4839325804899113,-2.2683758540881427,0.3443673101846472,n.s.,1.2400146784970971,3.474924854858144,False
|
| 21 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,Plasmablast (markers),0.5802468944674085,-1.004564184545843,0.8828558649585746,n.s.,0.6420129193391261,1.0398975306960363,False
|
| 22 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,MS1 monocyte (IL1R2+),-0.9720542970967039,-1.441552576635392,0.03495522040514784,n.s.,0.4046166451994816,0.2059643894111682,False
|
| 23 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,Cycling monocyte (Ye 2026),0.048444239956909385,-0.946967464912957,1.0703514066151243,n.s.,0.0867530899911269,0.0895573940039075,True
|
| 24 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,ISG-high T,1.1244275627500213,-0.2311471053349451,2.2289746571282327,n.s.,0.1695550987537442,0.3730783265918156,False
|
| 25 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,DN2 / ABC B,-0.8490216919990585,-1.7771690347400295,0.023006929294094997,n.s.,0.6336882030608005,0.3506570213062786,False
|
| 26 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,HSPC,-0.3245382108713872,-2.1210792925881425,0.986916513186777,n.s.,1.3382759622367466,0.5735468409586056,False
|
| 27 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,Cytotoxic CD4 T,0.16131217900752481,-0.23230354979720874,0.5375235938367285,n.s.,0.8002029614840337,0.8952312048786761,False
|
| 28 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,Granulocyte-like (LDG),-0.16946257869589676,-1.0183954189978437,0.5074786552816878,n.s.,0.0430958201462403,0.038254166326095,True
|
| 29 |
+
same cohort (fold 0),58 pairs: scratch vs pretrained,MEAN of 13 rare states,0.019151925949383884,-0.3497161225066392,0.2546273435529553,n.s.,,,False
|
| 30 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",C1Q-high monocyte (reference),0.9699447974426848,-0.09868473199385874,1.2492042857727352,n.s.,0.4669732296338602,0.9168073770721404,False
|
| 31 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",Cycling T/NK,0.8149659272233328,-0.14237804961215828,1.868799958794884,n.s.,0.4159012861180642,1.336003190869205,False
|
| 32 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",Cycling activated T (CD38/HLA-DR),0.8698209660111975,0.2387230862423245,1.3948694641077908,pretrained closer,0.3149812026848403,0.5783226998371942,False
|
| 33 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",Proliferating T (label),-0.7088112784688725,-1.0306301016947128,-0.09694143142512351,scratch closer,1.127184984582887,1.84558283627878,False
|
| 34 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",Plasmablast (label),-1.4008546921365193,-2.017565762894665,-0.13381039033541087,scratch closer,1.2400146784970971,3.280271387884763,False
|
| 35 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",Plasmablast (markers),0.4295262256046736,-0.8657475279772928,1.009683927683646,n.s.,0.6420129193391261,1.1544749803921568,False
|
| 36 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",MS1 monocyte (IL1R2+),-0.16389316906287776,-0.8080204962792746,0.8570150050284154,n.s.,0.4046166451994816,0.3610970873786407,False
|
| 37 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",Cycling monocyte (Ye 2026),0.38522056330494703,-0.4292317666723376,1.2852406747680627,n.s.,0.0867530899911269,0.1139278671116504,True
|
| 38 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",ISG-high T,1.3927547304550578,-0.22266490235817124,2.6639404609201933,n.s.,0.1695550987537442,0.449464931449181,False
|
| 39 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",DN2 / ABC B,-0.7639308560690516,-1.3960959826234434,0.008289749140269096,n.s.,0.6336882030608005,0.3723257190531011,False
|
| 40 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",HSPC,0.11111797526313144,-1.0175212392815238,0.9971729320571884,n.s.,1.3382759622367466,0.7821093285799168,False
|
| 41 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",Cytotoxic CD4 T,-0.09379404096049504,-0.6299532177064382,0.4771681280433422,n.s.,0.8002029614840337,1.322165614831606,False
|
| 42 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",Granulocyte-like (LDG),-0.058730405306818056,-0.6302063444543992,0.503024744268593,n.s.,0.0430958201462403,0.0413254274053143,True
|
| 43 |
+
same cohort (fold 0),"58 pairs, full FT: scratch vs pretrained",MEAN of 13 rare states,0.13717974948464545,-0.138205083434171,0.406414119140758,n.s.,,,False
|
| 44 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,C1Q-high monocyte (reference),0.23147832976094374,-0.6444828041325054,0.7831020209592113,n.s.,0.5324998918978538,0.6257734890436245,False
|
| 45 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,Cycling T/NK,1.0313584688830588,-0.007737800903293942,1.5261565007422315,n.s.,0.4790315538632603,0.9873293455925888,False
|
| 46 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,Cycling activated T (CD38/HLA-DR),0.7050792150097724,-0.07540090978841459,1.4965074490393966,n.s.,0.2789361146046564,0.4570400536740661,False
|
| 47 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,Proliferating T (label),-0.7906932831358676,-1.0778472033559245,-0.08462973996059603,scratch closer,0.9771784814130092,1.7603168470614865,False
|
| 48 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,Plasmablast (label),-0.966525295448912,-1.5399884127006007,0.3582787412709382,n.s.,1.1751301895059698,2.299794665352174,False
|
| 49 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,Plasmablast (markers),-0.04670154684237399,-0.8876767403962907,0.3428566825586785,n.s.,0.7341320942359123,0.7107369395999606,False
|
| 50 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,MS1 monocyte (IL1R2+),-0.6175073847140813,-1.3786106739282946,0.3991309511710483,n.s.,0.4620999194245931,0.3010185021290256,False
|
| 51 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,Cycling monocyte (Ye 2026),0.14103101323570266,-0.7745918555515998,0.9761299716889105,n.s.,0.0869463596409315,0.096242802121991,True
|
| 52 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,ISG-high T,1.5235914814499734,0.08093111693918656,2.490361386456709,pretrained closer,0.183511576926034,0.5329305241249853,False
|
| 53 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,DN2 / ABC B,-0.6964542951210411,-1.2062211206138205,-0.07741922116319098,scratch closer,0.4592885537544296,0.2826701429139615,False
|
| 54 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,HSPC,-0.41861624780456425,-1.7336393333007316,0.5314081445933891,n.s.,0.955911401597676,0.7106993464052287,False
|
| 55 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,Cytotoxic CD4 T,-0.05526303146927253,-0.43084623539700634,0.5229947903941299,n.s.,1.1875538020086085,1.2337156640207918,False
|
| 56 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,Granulocyte-like (LDG),-1.1932258274675727,-2.1808584719933295,-0.35555346733430965,scratch closer,0.0585732288086649,0.0252858299372933,True
|
| 57 |
+
same cohort (fold 0),58 pairs + recipes: scratch vs pretrained,MEAN of 13 rare states,-0.0886498772049411,-0.37881608569253233,0.17775621997033475,n.s.,,,False
|
| 58 |
+
other cohorts (masked),scratch final vs LoRA r16,C1Q-high monocyte (reference),-0.9398422531946626,-1.4812569582522366,-0.3842868828583468,scratch closer,0.0730704382830807,0.0379077438589531,True
|
| 59 |
+
other cohorts (masked),scratch final vs LoRA r16,Cycling T/NK,0.8072698127302047,0.248934432894902,1.4514191504143368,pretrained closer,0.0753149959777236,0.1324783671914411,True
|
| 60 |
+
other cohorts (masked),scratch final vs LoRA r16,Cycling activated T (CD38/HLA-DR),0.2570155934302445,-0.29089230152227374,0.830875595062909,n.s.,0.0773662098147176,0.0926063375201911,True
|
| 61 |
+
other cohorts (masked),scratch final vs LoRA r16,Proliferating T (label),0.5907143986802106,0.04985700715874261,0.7958496116779148,pretrained closer,0.5370940666848425,0.8091027108625667,False
|
| 62 |
+
other cohorts (masked),scratch final vs LoRA r16,Plasmablast (label),-0.26519728866868686,-0.9457371816009248,0.46238913679596844,n.s.,0.8728126838148352,1.3763584629387786,False
|
| 63 |
+
other cohorts (masked),scratch final vs LoRA r16,Plasmablast (markers),0.4522411102413281,0.16834515186401586,0.7004957574273725,pretrained closer,0.575387732850919,0.7872880321522852,False
|
| 64 |
+
other cohorts (masked),scratch final vs LoRA r16,MS1 monocyte (IL1R2+),-0.38283623708405656,-1.3085054706058583,0.4609970044361858,n.s.,0.1610585398379573,0.1234847151564598,True
|
| 65 |
+
other cohorts (masked),scratch final vs LoRA r16,Cycling monocyte (Ye 2026),0.9953325723416415,0.21641031257678986,1.988385892690813,pretrained closer,0.036144289390185,0.0732107677203869,True
|
| 66 |
+
other cohorts (masked),scratch final vs LoRA r16,ISG-high T,-0.635190820381804,-1.148651397024437,0.9311349769312188,n.s.,1.101270082650952,1.7106889714081193,False
|
| 67 |
+
other cohorts (masked),scratch final vs LoRA r16,DN2 / ABC B,-0.8770550923027058,-1.3139905303996187,-0.45847103735444433,scratch closer,0.2658286721720235,0.144552472274974,False
|
| 68 |
+
other cohorts (masked),scratch final vs LoRA r16,HSPC,0.02066277251324722,-0.2947880655331509,0.3749902528486137,n.s.,1.6623998475609756,1.6387741863740826,False
|
| 69 |
+
other cohorts (masked),scratch final vs LoRA r16,Cytotoxic CD4 T,-0.18988686501240049,-0.48393289329063893,0.1334332861891434,n.s.,1.280693050065492,1.4609809697922689,False
|
| 70 |
+
other cohorts (masked),scratch final vs LoRA r16,Granulocyte-like (LDG),-0.29440789423630065,-1.0058786575786873,0.37521067507068123,n.s.,0.0207409780840793,0.0168835070232936,True
|
| 71 |
+
other cohorts (masked),scratch final vs LoRA r16,MEAN of 13 rare states,-0.03547539930336464,-0.20030350260969534,0.1825932174060262,n.s.,,,False
|
| 72 |
+
other cohorts (masked),scratch 50k vs LoRA r16,C1Q-high monocyte (reference),-1.3547531571789344,-1.8929420298054431,-0.8496898409719716,scratch closer,0.097617131023328,0.0379077438589531,True
|
| 73 |
+
other cohorts (masked),scratch 50k vs LoRA r16,Cycling T/NK,0.4735216502326387,-0.07239663280785381,1.1085811590007866,n.s.,0.0952358961428523,0.1324783671914411,True
|
| 74 |
+
other cohorts (masked),scratch 50k vs LoRA r16,Cycling activated T (CD38/HLA-DR),-0.5864880021293,-1.0701056575962133,-0.11495175332315999,scratch closer,0.1394398688005852,0.0926063375201911,True
|
| 75 |
+
other cohorts (masked),scratch 50k vs LoRA r16,Proliferating T (label),0.8863581784899157,0.3206185106926778,1.1041293316246865,pretrained closer,0.4374677199553131,0.8091027108625667,False
|
| 76 |
+
other cohorts (masked),scratch 50k vs LoRA r16,Plasmablast (label),0.5777315973083845,-0.34547826846836394,1.3607712960693341,n.s.,0.486476430679064,1.3763584629387786,False
|
| 77 |
+
other cohorts (masked),scratch 50k vs LoRA r16,Plasmablast (markers),0.9176426580346617,0.6288715764582294,1.1689196004281415,pretrained closer,0.4167028335163746,0.7872880321522852,False
|
| 78 |
+
other cohorts (masked),scratch 50k vs LoRA r16,MS1 monocyte (IL1R2+),-0.8521433242970446,-1.6530990790812128,-0.17982239233503802,scratch closer,0.223095102272663,0.1234847151564598,True
|
| 79 |
+
other cohorts (masked),scratch 50k vs LoRA r16,Cycling monocyte (Ye 2026),0.1423543368497917,-0.5029410745745533,0.8585021111224792,n.s.,0.066330557196999,0.0732107677203869,True
|
| 80 |
+
other cohorts (masked),scratch 50k vs LoRA r16,ISG-high T,-0.5704273360224283,-1.2620338081871827,0.8218439064218466,n.s.,1.1518875975903249,1.7106889714081193,False
|
| 81 |
+
other cohorts (masked),scratch 50k vs LoRA r16,DN2 / ABC B,-0.8018508493379279,-1.208167792482542,-0.3876737806268563,scratch closer,0.2522649798804725,0.144552472274974,False
|
| 82 |
+
other cohorts (masked),scratch 50k vs LoRA r16,HSPC,-0.5400149431225341,-0.9125363862791656,-0.002222557625658367,scratch closer,1.1259051281084682,1.6387741863740826,False
|
| 83 |
+
other cohorts (masked),scratch 50k vs LoRA r16,Cytotoxic CD4 T,-0.5197481186139914,-0.7143417206199745,0.17928173338527303,n.s.,1.018659639026527,1.4609809697922689,False
|
| 84 |
+
other cohorts (masked),scratch 50k vs LoRA r16,Granulocyte-like (LDG),-1.7957741365395297,-2.5070092819608836,-1.1565196694359625,scratch closer,0.0593575526053286,0.0168835070232936,True
|
| 85 |
+
other cohorts (masked),scratch 50k vs LoRA r16,MEAN of 13 rare states,-0.3095070343327922,-0.4464178224698228,-0.10882495167192426,scratch closer,,,False
|
| 86 |
+
other cohorts (masked),scratch final vs LoRA r64,C1Q-high monocyte (reference),-0.30257429386118995,-0.7918896328510745,0.21973507407442058,n.s.,0.0730704382830807,0.0591392264864203,True
|
| 87 |
+
other cohorts (masked),scratch final vs LoRA r64,Cycling T/NK,1.345204159242897,0.9228415416224076,1.888336185393764,pretrained closer,0.0753149959777236,0.192695697524953,True
|
| 88 |
+
other cohorts (masked),scratch final vs LoRA r64,Cycling activated T (CD38/HLA-DR),0.8258242987212041,0.37748883121010357,1.2962481545059665,pretrained closer,0.0773662098147176,0.1375560902973552,True
|
| 89 |
+
other cohorts (masked),scratch final vs LoRA r64,Proliferating T (label),0.8364772684993874,-0.3289177699176432,1.119649931090606,n.s.,0.5370940666848425,1.0411324350234112,False
|
| 90 |
+
other cohorts (masked),scratch final vs LoRA r64,Plasmablast (label),0.17722921484504883,-0.4767138954371683,0.3279236027417153,n.s.,0.8728126838148352,1.0127813410628468,False
|
| 91 |
+
other cohorts (masked),scratch final vs LoRA r64,Plasmablast (markers),0.2359697109805161,-0.04748860808344772,0.5123541313031872,n.s.,0.575387732850919,0.6776568347595588,False
|
| 92 |
+
other cohorts (masked),scratch final vs LoRA r64,MS1 monocyte (IL1R2+),-0.33859066662184567,-0.7945376640128864,0.22479087186379937,n.s.,0.1610585398379573,0.1273018562854049,True
|
| 93 |
+
other cohorts (masked),scratch final vs LoRA r64,Cycling monocyte (Ye 2026),0.8130796822582997,0.13522723218480706,1.6293918961252702,pretrained closer,0.036144289390185,0.0642254180793674,True
|
| 94 |
+
other cohorts (masked),scratch final vs LoRA r64,ISG-high T,-0.8073973936889911,-1.3218790883186395,0.9144157932205702,n.s.,1.101270082650952,1.9275342903606627,False
|
| 95 |
+
other cohorts (masked),scratch final vs LoRA r64,DN2 / ABC B,-0.42917290797173324,-0.7494286534468191,-0.12272598808629208,scratch closer,0.2658286721720235,0.1973846287934167,False
|
| 96 |
+
other cohorts (masked),scratch final vs LoRA r64,HSPC,-0.7907470862246386,-1.0586332454803513,-0.49347209874927306,scratch closer,1.6623998475609756,2.8785893981645128,False
|
| 97 |
+
other cohorts (masked),scratch final vs LoRA r64,Cytotoxic CD4 T,-0.39737359336524014,-0.6821216357583157,-0.04942697151218194,scratch closer,1.280693050065492,1.687132112528452,False
|
| 98 |
+
other cohorts (masked),scratch final vs LoRA r64,Granulocyte-like (LDG),-0.36139122128073087,-1.029290878369477,0.25278265906087827,n.s.,0.0207409780840793,0.0160564593584755,True
|
| 99 |
+
other cohorts (masked),scratch final vs LoRA r64,MEAN of 13 rare states,0.062041320887152605,-0.08041388284481239,0.22392736414091247,n.s.,,,False
|
| 100 |
+
other cohorts (masked),scratch 50k vs LoRA r64,C1Q-high monocyte (reference),-0.7174851978454617,-1.2425380279584697,-0.21856730397006693,scratch closer,0.097617131023328,0.0591392264864203,True
|
| 101 |
+
other cohorts (masked),scratch 50k vs LoRA r64,Cycling T/NK,1.011455996745331,0.5128580815725398,1.6010896277273572,pretrained closer,0.0952358961428523,0.192695697524953,True
|
| 102 |
+
other cohorts (masked),scratch 50k vs LoRA r64,Cycling activated T (CD38/HLA-DR),-0.017679296838340353,-0.5110747195694431,0.4511392951026028,n.s.,0.1394398688005852,0.1375560902973552,True
|
| 103 |
+
other cohorts (masked),scratch 50k vs LoRA r64,Proliferating T (label),1.1321210483090927,-0.028228451345985776,1.425678776361177,n.s.,0.4374677199553131,1.0411324350234112,False
|
| 104 |
+
other cohorts (masked),scratch 50k vs LoRA r64,Plasmablast (label),1.0201581008221203,0.14643266240205596,1.2870570074568883,pretrained closer,0.486476430679064,1.0127813410628468,False
|
| 105 |
+
other cohorts (masked),scratch 50k vs LoRA r64,Plasmablast (markers),0.7013712587738496,0.38320491672195256,1.0089367624855619,pretrained closer,0.4167028335163746,0.6776568347595588,False
|
| 106 |
+
other cohorts (masked),scratch 50k vs LoRA r64,MS1 monocyte (IL1R2+),-0.8078977538348338,-1.2497497645895825,-0.3624171214903144,scratch closer,0.223095102272663,0.1273018562854049,True
|
| 107 |
+
other cohorts (masked),scratch 50k vs LoRA r64,Cycling monocyte (Ye 2026),-0.03989855323355007,-0.7815647610540041,0.6879033887654453,n.s.,0.066330557196999,0.0642254180793674,True
|
| 108 |
+
other cohorts (masked),scratch 50k vs LoRA r64,ISG-high T,-0.7426339093296155,-1.4173858428296646,0.8287112535206411,n.s.,1.1518875975903249,1.9275342903606627,False
|
| 109 |
+
other cohorts (masked),scratch 50k vs LoRA r64,DN2 / ABC B,-0.35396866500695534,-0.7061965024804587,-0.04654763188446811,scratch closer,0.2522649798804725,0.1973846287934167,False
|
| 110 |
+
other cohorts (masked),scratch 50k vs LoRA r64,HSPC,-1.3514248018604198,-1.6865037547859747,-0.8189735046579152,scratch closer,1.1259051281084682,2.8785893981645128,False
|
| 111 |
+
other cohorts (masked),scratch 50k vs LoRA r64,Cytotoxic CD4 T,-0.727234846966831,-0.9260981443008456,0.06563722206425136,n.s.,1.018659639026527,1.687132112528452,False
|
| 112 |
+
other cohorts (masked),scratch 50k vs LoRA r64,Granulocyte-like (LDG),-1.8627574635839599,-2.605581180944553,-1.2414507929821672,scratch closer,0.0593575526053286,0.0160564593584755,True
|
| 113 |
+
other cohorts (masked),scratch 50k vs LoRA r64,MEAN of 13 rare states,-0.21199031414227493,-0.3740271400127492,-0.024982602943402044,scratch closer,,,False
|
| 114 |
+
other cohorts (masked),scratch final vs Full FT,C1Q-high monocyte (reference),0.19403551507647965,-0.351884192263881,0.754624122948642,n.s.,0.0730704382830807,0.0836144742186387,True
|
| 115 |
+
other cohorts (masked),scratch final vs Full FT,Cycling T/NK,2.243342627096989,1.7148101924288308,2.9094222132682077,pretrained closer,0.0753149959777236,0.3601223567794363,False
|
| 116 |
+
other cohorts (masked),scratch final vs Full FT,Cycling activated T (CD38/HLA-DR),1.7328617585054076,1.1949141038363795,2.3796574813704408,pretrained closer,0.0773662098147176,0.2587405682137551,False
|
| 117 |
+
other cohorts (masked),scratch final vs Full FT,Proliferating T (label),0.7460975359058798,-0.3892684290502386,1.1782281495075542,n.s.,0.5370940666848425,1.108491769417387,False
|
| 118 |
+
other cohorts (masked),scratch final vs Full FT,Plasmablast (label),-0.9452430143756747,-1.6504596302387624,-0.017955494492291692,scratch closer,0.8728126838148352,2.205359819078424,False
|
| 119 |
+
other cohorts (masked),scratch final vs Full FT,Plasmablast (markers),0.2284056347260518,-0.7108681335840631,1.0834996592318542,n.s.,0.575387732850919,1.483040798427553,False
|
| 120 |
+
other cohorts (masked),scratch final vs Full FT,MS1 monocyte (IL1R2+),0.7555470583629571,-0.05400671879647958,1.4987092439334269,n.s.,0.1610585398379573,0.2720343761735265,False
|
| 121 |
+
other cohorts (masked),scratch final vs Full FT,Cycling monocyte (Ye 2026),1.7222006529460039,0.9717618404898799,2.617077214436303,pretrained closer,0.036144289390185,0.1215710447520997,True
|
| 122 |
+
other cohorts (masked),scratch final vs Full FT,ISG-high T,-0.8460723633239322,-1.108060451522082,0.9054150335186909,n.s.,1.101270082650952,1.979966388121804,False
|
| 123 |
+
other cohorts (masked),scratch final vs Full FT,DN2 / ABC B,0.10365472054293035,-0.18033952886897991,0.4132510554460348,n.s.,0.2658286721720235,0.2858536073071018,False
|
| 124 |
+
other cohorts (masked),scratch final vs Full FT,HSPC,-0.8562177298636451,-1.1953363810081004,-0.49380180397446,scratch closer,1.6623998475609756,3.0124311265969803,False
|
| 125 |
+
other cohorts (masked),scratch final vs Full FT,Cytotoxic CD4 T,-0.3864165648740653,-0.7037211238281685,-0.019453545586445676,scratch closer,1.280693050065492,1.674139458539571,False
|
| 126 |
+
other cohorts (masked),scratch final vs Full FT,Granulocyte-like (LDG),-0.8586315634949617,-1.890937921032326,0.035860473252906575,n.s.,0.0207409780840793,0.0112905444991257,True
|
| 127 |
+
other cohorts (masked),scratch final vs Full FT,MEAN of 13 rare states,0.2948895590177246,0.06771755050657054,0.5662305786807542,pretrained closer,,,False
|
| 128 |
+
other cohorts (masked),scratch 50k vs Full FT,C1Q-high monocyte (reference),-0.22087538890779213,-0.784564478420458,0.2491936171327289,n.s.,0.097617131023328,0.0836144742186387,True
|
| 129 |
+
other cohorts (masked),scratch 50k vs Full FT,Cycling T/NK,1.9095944645994232,1.355165132912577,2.608860288302769,pretrained closer,0.0952358961428523,0.3601223567794363,False
|
| 130 |
+
other cohorts (masked),scratch 50k vs Full FT,Cycling activated T (CD38/HLA-DR),0.8893581629458631,0.4143225488070337,1.3572959068606238,pretrained closer,0.1394398688005852,0.2587405682137551,False
|
| 131 |
+
other cohorts (masked),scratch 50k vs Full FT,Proliferating T (label),1.0417413157155848,-0.12906956870827427,1.510409337071922,n.s.,0.4374677199553131,1.108491769417387,False
|
| 132 |
+
other cohorts (masked),scratch 50k vs Full FT,Plasmablast (label),-0.10231412839860332,-1.0710516270350647,0.844091871177576,n.s.,0.486476430679064,2.205359819078424,False
|
| 133 |
+
other cohorts (masked),scratch 50k vs Full FT,Plasmablast (markers),0.6938071825193853,-0.21631389306094356,1.489549357080578,n.s.,0.4167028335163746,1.483040798427553,False
|
| 134 |
+
other cohorts (masked),scratch 50k vs Full FT,MS1 monocyte (IL1R2+),0.286239971149969,-0.3939720291160393,0.9296628580435381,n.s.,0.223095102272663,0.2720343761735265,False
|
| 135 |
+
other cohorts (masked),scratch 50k vs Full FT,Cycling monocyte (Ye 2026),0.8692224174541541,0.18344871454015327,1.6365969274509506,pretrained closer,0.066330557196999,0.1215710447520997,True
|
| 136 |
+
other cohorts (masked),scratch 50k vs Full FT,ISG-high T,-0.7813088789645566,-1.1814321978769222,0.7962039140767088,n.s.,1.1518875975903249,1.979966388121804,False
|
| 137 |
+
other cohorts (masked),scratch 50k vs Full FT,DN2 / ABC B,0.17885896350770825,-0.09408336005872026,0.4967592511270396,n.s.,0.2522649798804725,0.2858536073071018,False
|
| 138 |
+
other cohorts (masked),scratch 50k vs Full FT,HSPC,-1.4168954454994263,-1.789331678772034,-0.8422323921667151,scratch closer,1.1259051281084682,3.0124311265969803,False
|
| 139 |
+
other cohorts (masked),scratch 50k vs Full FT,Cytotoxic CD4 T,-0.7162778184756562,-0.9359696470170399,0.011634165459463264,n.s.,1.018659639026527,1.674139458539571,False
|
| 140 |
+
other cohorts (masked),scratch 50k vs Full FT,Granulocyte-like (LDG),-2.3599978057981907,-3.21555245457478,-1.7196780324607068,scratch closer,0.0593575526053286,0.0112905444991257,True
|
| 141 |
+
other cohorts (masked),scratch 50k vs Full FT,MEAN of 13 rare states,0.020857923988297163,-0.19166006371074568,0.29692253428186594,n.s.,,,False
|
| 142 |
+
other cohorts (masked),scratch final vs Calibration,C1Q-high monocyte (reference),0.04546914811396574,-0.3951113874597825,0.52274308187319,n.s.,0.0730704382830807,0.0756199128382113,True
|
| 143 |
+
other cohorts (masked),scratch final vs Calibration,Cycling T/NK,2.6529624337361684,2.1517335383223846,3.3073185664360616,pretrained closer,0.0753149959777236,0.4789398200183921,False
|
| 144 |
+
other cohorts (masked),scratch final vs Calibration,Cycling activated T (CD38/HLA-DR),1.6883850556112758,1.2158646929836037,2.2376033764567627,pretrained closer,0.0773662098147176,0.2511255852558967,False
|
| 145 |
+
other cohorts (masked),scratch final vs Calibration,Proliferating T (label),0.25879576461217246,-0.03545927650478682,0.5210061675639664,n.s.,0.5370940666848425,0.6427550534927856,False
|
| 146 |
+
other cohorts (masked),scratch final vs Calibration,Plasmablast (label),-0.15676403587092375,-0.5272827875454912,0.17991900189093293,n.s.,0.8728126838148352,0.7829141153384702,False
|
| 147 |
+
other cohorts (masked),scratch final vs Calibration,Plasmablast (markers),-0.6282827108318532,-0.9793522864366764,-0.3092871206094582,scratch closer,0.575387732850919,0.3722333254400505,False
|
| 148 |
+
other cohorts (masked),scratch final vs Calibration,MS1 monocyte (IL1R2+),2.169409775072883,1.631537988685901,2.821702051153473,pretrained closer,0.1610585398379573,0.7253936473575063,False
|
| 149 |
+
other cohorts (masked),scratch final vs Calibration,Cycling monocyte (Ye 2026),2.786909671954928,1.914715849342894,3.820315742898634,pretrained closer,0.036144289390185,0.2563415841564834,False
|
| 150 |
+
other cohorts (masked),scratch final vs Calibration,ISG-high T,-0.4844143163516037,-0.7392731669069954,0.7872816843396875,n.s.,1.101270082650952,1.540999282816366,False
|
| 151 |
+
other cohorts (masked),scratch final vs Calibration,DN2 / ABC B,0.09724138504993296,-0.33393658371559376,0.5420016220935218,n.s.,0.2658286721720235,0.2844684111816357,False
|
| 152 |
+
other cohorts (masked),scratch final vs Calibration,HSPC,-0.11742530853383282,-0.4349890273381199,0.23995625718635763,n.s.,1.6623998475609756,1.8039200694254132,False
|
| 153 |
+
other cohorts (masked),scratch final vs Calibration,Cytotoxic CD4 T,-0.6599714919809385,-0.9826707342187848,-0.30118796681813464,scratch closer,1.280693050065492,2.0243331855625573,False
|
| 154 |
+
other cohorts (masked),scratch final vs Calibration,Granulocyte-like (LDG),1.6735405771288532,1.1698934191495043,2.2272795862307078,pretrained closer,0.0207409780840793,0.0670251882295056,True
|
| 155 |
+
other cohorts (masked),scratch final vs Calibration,MEAN of 13 rare states,0.7173735344393097,0.5663771634658857,0.947893361742983,pretrained closer,,,False
|
| 156 |
+
other cohorts (masked),scratch 50k vs Calibration,C1Q-high monocyte (reference),-0.36944175587030603,-0.7819933792238812,0.007149966641778575,n.s.,0.097617131023328,0.0756199128382113,True
|
| 157 |
+
other cohorts (masked),scratch 50k vs Calibration,Cycling T/NK,2.319214271238602,1.7822726534071889,2.965265535211989,pretrained closer,0.0952358961428523,0.4789398200183921,False
|
| 158 |
+
other cohorts (masked),scratch 50k vs Calibration,Cycling activated T (CD38/HLA-DR),0.8448814600517314,0.43338135261428923,1.2807351193720404,pretrained closer,0.1394398688005852,0.2511255852558967,False
|
| 159 |
+
other cohorts (masked),scratch 50k vs Calibration,Proliferating T (label),0.5544395444218776,0.2727184777866661,0.8107420205685348,pretrained closer,0.4374677199553131,0.6427550534927856,False
|
| 160 |
+
other cohorts (masked),scratch 50k vs Calibration,Plasmablast (label),0.6861648501061477,0.20356222868621457,1.1362125207560048,pretrained closer,0.486476430679064,0.7829141153384702,False
|
| 161 |
+
other cohorts (masked),scratch 50k vs Calibration,Plasmablast (markers),-0.16288116303851963,-0.5346806860641824,0.18104582924724735,n.s.,0.4167028335163746,0.3722333254400505,False
|
| 162 |
+
other cohorts (masked),scratch 50k vs Calibration,MS1 monocyte (IL1R2+),1.700102687859895,1.2684313447389266,2.2211876988918404,pretrained closer,0.223095102272663,0.7253936473575063,False
|
| 163 |
+
other cohorts (masked),scratch 50k vs Calibration,Cycling monocyte (Ye 2026),1.9339314364630782,1.2358712847188935,2.54673483626598,pretrained closer,0.066330557196999,0.2563415841564834,False
|
| 164 |
+
other cohorts (masked),scratch 50k vs Calibration,ISG-high T,-0.4196508319922281,-0.8007422155181345,0.687560102927006,n.s.,1.1518875975903249,1.540999282816366,False
|
| 165 |
+
other cohorts (masked),scratch 50k vs Calibration,DN2 / ABC B,0.17244562801471086,-0.24203170042588462,0.5767229759673134,n.s.,0.2522649798804725,0.2844684111816357,False
|
| 166 |
+
other cohorts (masked),scratch 50k vs Calibration,HSPC,-0.6781030241696141,-1.0414484477583763,-0.15146410237438945,scratch closer,1.1259051281084682,1.8039200694254132,False
|
| 167 |
+
other cohorts (masked),scratch 50k vs Calibration,Cytotoxic CD4 T,-0.9898327455825294,-1.2010359815364504,-0.32355640436681626,scratch closer,1.018659639026527,2.0243331855625573,False
|
| 168 |
+
other cohorts (masked),scratch 50k vs Calibration,Granulocyte-like (LDG),0.17217433482562416,-0.24990642186411455,0.5894812601459527,n.s.,0.0593575526053286,0.0670251882295056,True
|
| 169 |
+
other cohorts (masked),scratch 50k vs Calibration,MEAN of 13 rare states,0.4433418994098823,0.29882799917880265,0.6609261768384087,pretrained closer,,,False
|