Archive 4 checkpoints (1-4)

.gitattributes

@@ -222,3 +222,5 @@ paper-releases/pbisc-diffusion/updates/20261012T130738Z-strictpairs-implementati
 paper-releases/pbisc-diffusion/updates/20261012T130738Z-strictpairs-implementation/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/scsemi/ye2026_sepsis/disease/draw4/adopted_legacy/scsemi_project/models/fastreconst2_Res-PS22 filter=lfs diff=lfs merge=lfs -text
 paper-releases/pbisc-diffusion/updates/20261012T130738Z-strictpairs-implementation/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/scsemi/ye2026_sepsis/disease/draw4/adopted_legacy/scsemi_project/models/fastreconst2_Skin-S142 filter=lfs diff=lfs merge=lfs -text
 paper-releases/pbisc-diffusion/updates/20261012T130738Z-strictpairs-implementation/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/scsemi/ye2026_sepsis/disease/draw4/adopted_legacy/scsemi_project/models/fastreconst2_Skin-S46 filter=lfs diff=lfs merge=lfs -text
+paper-releases/pbisc-diffusion/updates/20261012T233237Z-figure-layout-todo/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/layout_preview_20261013/PBISC_Figure_layout_TODO_20261013.pdf filter=lfs diff=lfs merge=lfs -text
+paper-releases/pbisc-diffusion/updates/20261012T233237Z-figure-layout-todo/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/layout_preview_20261013/PBISC_Figure_layout_TODO_20261013.pptx filter=lfs diff=lfs merge=lfs -text

paper-releases/pbisc-diffusion/updates/20261012T233237Z-figure-layout-todo/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/layout_preview_20261013/PANEL_MAP.csv

@@ -0,0 +1,40 @@
+figure,panel,title,state,dependency
+1,a-e,Complete training-to-inference concept,AVAILABLE,fig1_overview.pdf
+2,a-c,Architecture and gene-panel comparisons,AVAILABLE,"fig2_ablation_and_support.pdf, top row"
+2,d-f,Full fine-tuning and LoRA rank selection,HISTORICAL 5-FOLD,"fig2_ablation_and_support.pdf, middle row"
+2,g,Precision curve,TODO,"Exact n = 5 / 10 / 20 / 40
+Five outer folds
+2/20 units scored
+LoRA, same-n calibration and scSemiProfiler; raw/ref-cal control lines."
+2,h,Coverage curve,TODO,"Exact n = 5 / 10 / 20 / 40
+Five outer folds
+2/20 units scored
+LoRA, same-n calibration and scSemiProfiler; raw/ref-cal control lines."
+2,i,Composition curve,TODO,"Exact n = 5 / 10 / 20 / 40
+Five outer folds
+2/20 units scored
+LoRA, same-n calibration and scSemiProfiler; raw/ref-cal control lines."
+3,a,Common-reference UMAP,TODO,"Real | PBISC-LoRA | scSemiProfiler | external ref-cal | same-pair calibration
+Real-reference coordinates only; shared cell-type colors; no forced integration."
+3,b,Canonical marker profiles,TODO,"Heatmap + dot plot
+Fixed marker list, gene order and expression scale; dot size = detection fraction."
+3,c,Cell-type composition,TODO,"Observed and generated fractions
+Per-donor composition error, with the same cell-type ordering across methods."
+3,d,Primary quality metrics,TODO,"Precision | coverage | composition L1
+Matched reference, donors and output cell budget."
+3,e,Additional realism,TODO,"C2ST | MMD2 | E_N
+Real-versus-real control; no interpretation of UMAP overlap as a score."
+3,f,Operating time and compute,TODO,"Small supporting panel: adaptation time + target-generation time.
+Exclude the supplied pretraining cost, GPU queue waits and evaluation time."
+4,a,Exact-five-pair comparison,TODO,"Five draws per cohort; one pair per donor.
+Pearson / Spearman logFC, partial correlation and top-DEG overlap; patient-level uncertainty."
+4,b,Existing disease five-fold contrasts,HISTORICAL ESTIMATOR,"fig4_historical_5fold_separate.pdf, logFC + partial-r rows"
+4,c,Gene-level validation,TODO,"Observed-generated scatter / heatmap, CAT curves and patient-aware DEG sensitivity."
+4,d,Matched deconvolution,TODO,BayesPrism / CIBERSORTx on the same support pairs; supported expression/composition endpoints.
+5,a,State-frequency and detection map,TODO,"All predefined state x cohort combinations; observed frequency, donor support and explicit zero detections."
+5,b,State identity and location,TODO,"Common UMAP with rare-state zoom; marker heatmap, dot plot and same-cell co-detection."
+5,c,Patient-level reproduction,TODO,"Observed-generated abundance scatter, missed-state frequency and disease direction."
+5,d,All-state robustness,TODO,Retain zero draws and real donor IDs; common-depth / equal-cell-budget sensitivity; required markers.
+5,e,Any C1Q versus multi-gene C1Q rule,DESCRIPTIVE,fig5_behcet_rules_diagnostic.pdf
+6,a,Dataset and reproduction design,TODO,"Place the cohort, published target finding, paired-adaptation subset and held-out evaluation here after data preparation."
+6,b,Reproduction and validation,TODO,Place the reproduction result and independent supporting evidence here. No Figure 6 analysis is started by this layout task.

paper-releases/pbisc-diffusion/updates/20261012T233237Z-figure-layout-todo/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/layout_preview_20261013/PANEL_MAP.json

@@ -0,0 +1,163 @@
+[
+  {
+    "figure": 1,
+    "panel": "a-e",
+    "title": "Complete training-to-inference concept",
+    "state": "AVAILABLE",
+    "dependency": "fig1_overview.pdf"
+  },
+  {
+    "figure": 2,
+    "panel": "a-c",
+    "title": "Architecture and gene-panel comparisons",
+    "state": "AVAILABLE",
+    "dependency": "fig2_ablation_and_support.pdf, top row"
+  },
+  {
+    "figure": 2,
+    "panel": "d-f",
+    "title": "Full fine-tuning and LoRA rank selection",
+    "state": "HISTORICAL 5-FOLD",
+    "dependency": "fig2_ablation_and_support.pdf, middle row"
+  },
+  {
+    "figure": 2,
+    "panel": "g",
+    "title": "Precision curve",
+    "state": "TODO",
+    "dependency": "Exact n = 5 / 10 / 20 / 40\nFive outer folds\n2/20 units scored\nLoRA, same-n calibration and scSemiProfiler; raw/ref-cal control lines."
+  },
+  {
+    "figure": 2,
+    "panel": "h",
+    "title": "Coverage curve",
+    "state": "TODO",
+    "dependency": "Exact n = 5 / 10 / 20 / 40\nFive outer folds\n2/20 units scored\nLoRA, same-n calibration and scSemiProfiler; raw/ref-cal control lines."
+  },
+  {
+    "figure": 2,
+    "panel": "i",
+    "title": "Composition curve",
+    "state": "TODO",
+    "dependency": "Exact n = 5 / 10 / 20 / 40\nFive outer folds\n2/20 units scored\nLoRA, same-n calibration and scSemiProfiler; raw/ref-cal control lines."
+  },
+  {
+    "figure": 3,
+    "panel": "a",
+    "title": "Common-reference UMAP",
+    "state": "TODO",
+    "dependency": "Real | PBISC-LoRA | scSemiProfiler | external ref-cal | same-pair calibration\nReal-reference coordinates only; shared cell-type colors; no forced integration."
+  },
+  {
+    "figure": 3,
+    "panel": "b",
+    "title": "Canonical marker profiles",
+    "state": "TODO",
+    "dependency": "Heatmap + dot plot\nFixed marker list, gene order and expression scale; dot size = detection fraction."
+  },
+  {
+    "figure": 3,
+    "panel": "c",
+    "title": "Cell-type composition",
+    "state": "TODO",
+    "dependency": "Observed and generated fractions\nPer-donor composition error, with the same cell-type ordering across methods."
+  },
+  {
+    "figure": 3,
+    "panel": "d",
+    "title": "Primary quality metrics",
+    "state": "TODO",
+    "dependency": "Precision | coverage | composition L1\nMatched reference, donors and output cell budget."
+  },
+  {
+    "figure": 3,
+    "panel": "e",
+    "title": "Additional realism",
+    "state": "TODO",
+    "dependency": "C2ST | MMD2 | E_N\nReal-versus-real control; no interpretation of UMAP overlap as a score."
+  },
+  {
+    "figure": 3,
+    "panel": "f",
+    "title": "Operating time and compute",
+    "state": "TODO",
+    "dependency": "Small supporting panel: adaptation time + target-generation time.\nExclude the supplied pretraining cost, GPU queue waits and evaluation time."
+  },
+  {
+    "figure": 4,
+    "panel": "a",
+    "title": "Exact-five-pair comparison",
+    "state": "TODO",
+    "dependency": "Five draws per cohort; one pair per donor.\nPearson / Spearman logFC, partial correlation and top-DEG overlap; patient-level uncertainty."
+  },
+  {
+    "figure": 4,
+    "panel": "b",
+    "title": "Existing disease five-fold contrasts",
+    "state": "HISTORICAL ESTIMATOR",
+    "dependency": "fig4_historical_5fold_separate.pdf, logFC + partial-r rows"
+  },
+  {
+    "figure": 4,
+    "panel": "c",
+    "title": "Gene-level validation",
+    "state": "TODO",
+    "dependency": "Observed-generated scatter / heatmap, CAT curves and patient-aware DEG sensitivity."
+  },
+  {
+    "figure": 4,
+    "panel": "d",
+    "title": "Matched deconvolution",
+    "state": "TODO",
+    "dependency": "BayesPrism / CIBERSORTx on the same support pairs; supported expression/composition endpoints."
+  },
+  {
+    "figure": 5,
+    "panel": "a",
+    "title": "State-frequency and detection map",
+    "state": "TODO",
+    "dependency": "All predefined state x cohort combinations; observed frequency, donor support and explicit zero detections."
+  },
+  {
+    "figure": 5,
+    "panel": "b",
+    "title": "State identity and location",
+    "state": "TODO",
+    "dependency": "Common UMAP with rare-state zoom; marker heatmap, dot plot and same-cell co-detection."
+  },
+  {
+    "figure": 5,
+    "panel": "c",
+    "title": "Patient-level reproduction",
+    "state": "TODO",
+    "dependency": "Observed-generated abundance scatter, missed-state frequency and disease direction."
+  },
+  {
+    "figure": 5,
+    "panel": "d",
+    "title": "All-state robustness",
+    "state": "TODO",
+    "dependency": "Retain zero draws and real donor IDs; common-depth / equal-cell-budget sensitivity; required markers."
+  },
+  {
+    "figure": 5,
+    "panel": "e",
+    "title": "Any C1Q versus multi-gene C1Q rule",
+    "state": "DESCRIPTIVE",
+    "dependency": "fig5_behcet_rules_diagnostic.pdf"
+  },
+  {
+    "figure": 6,
+    "panel": "a",
+    "title": "Dataset and reproduction design",
+    "state": "TODO",
+    "dependency": "Place the cohort, published target finding, paired-adaptation subset and held-out evaluation here after data preparation."
+  },
+  {
+    "figure": 6,
+    "panel": "b",
+    "title": "Reproduction and validation",
+    "state": "TODO",
+    "dependency": "Place the reproduction result and independent supporting evidence here. No Figure 6 analysis is started by this layout task."
+  }
+]

paper-releases/pbisc-diffusion/updates/20261012T233237Z-figure-layout-todo/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/layout_preview_20261013/PBISC_Figure_layout_TODO_20261013.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da38d26a2acc938f50f34b10a1d2718ef366accfbd0db4a2e7b4b1aabedc2661
+size 937630

paper-releases/pbisc-diffusion/updates/20261012T233237Z-figure-layout-todo/diffusion/dataset/server_1/model8/plans/paper_figures_strictpairs_20261012/layout_preview_20261013/PBISC_Figure_layout_TODO_20261013.pptx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1202b2af47cfd9b9451cb6ed102a619f68681ff7685d59a346f6a415ab75a55
+size 648050