Initial release: MacroPert IBD macrophage perturbation model

- scGPT v2 fine-tuned on IBD macrophage scRNA-seq (GSE134809 + 3 datasets)
- Continuous polarization embedding: Pearson r=0.909, IBD AUROC=0.989
- scGPT v1 discrete baseline archived for comparison
- In silico KO predictions for 8 IBD target genes (CellOracle + OmniPath)

.gitattributes

@@ -0,0 +1 @@
+*.pt filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,119 @@
+---
+license: mit
+language:
+- en
+tags:
+- biology
+- single-cell
+- scRNA-seq
+- macrophage
+- IBD
+- inflammatory-bowel-disease
+- perturbation
+- gene-knockout
+- scGPT
+- transformers
+datasets:
+- GEO:GSE134809
+- GEO:GSE116222
+- GEO:GSE182270
+- GEO:GSE148810
+metrics:
+- pearsonr
+- roc_auc
+---
+
+# MacroPert — IBD Macrophage Perturbation Model
+
+A fine-tuned scGPT model for predicting macrophage polarization state and IBD disease status from single-cell RNA-seq, combined with in silico gene knockout predictions for 13 IBD target genes.
+
+## Model Description
+
+Macrophages in inflammatory bowel disease (IBD) exist along a continuous spectrum between pro-inflammatory (M1) and anti-inflammatory (M2) states, rather than in discrete classes. **MacroPert** captures this continuous polarization spectrum and predicts how individual gene knockouts shift macrophage state.
+
+The repository contains:
+
+| File | Description |
+|------|-------------|
+| `scgpt_ibd_v2/best_model.pt` | scGPT v2 fine-tuned weights — continuous polarization model |
+| `scgpt_ibd_v2/test_metrics.json` | Test set performance metrics |
+| `scgpt_ibd_v1/best_model.pt` | scGPT v1 weights — discrete 4-class baseline (archived) |
+| `scgpt_ibd_v1/test_metrics.json` | v1 test metrics |
+| `results/ibd_ko_predictions_combined.json` | In silico KO predictions for 8 IBD target genes |
+
+## Performance
+
+| Metric | scGPT v1 (4-class) | scGPT v2 (continuous) |
+|--------|--------------------|-----------------------|
+| IBD AUROC | 0.915 | **0.989** |
+| Polarization Pearson r | — | **0.909** |
+| Classification Accuracy | 0.754 | — |
+
+## Training Data
+
+Fine-tuned on four IBD macrophage scRNA-seq datasets:
+
+| GEO Accession | Cells | Condition |
+|---------------|-------|-----------|
+| GSE134809 | 13,794 | IBD macrophages (primary) |
+| GSE116222 | ~8,000 | Ulcerative colitis macrophages |
+| GSE182270 | ~6,000 | IBD macrophage substates |
+| GSE148810 | ~5,000 | Crohn's disease macrophages |
+
+## Model Architecture
+
+- **Base:** scGPT pretrained transformer
+- **Fine-tuning objective:**
+  ```
+  L = 0.5 × L_soft_contrastive + 0.3 × L_MSE(pol) + 0.2 × L_BCE(ibd)
+  ```
+- **Polarization score:** `pol_score = z-score(M1_score − M2_score)` using literature-based gene signatures
+  - M1 genes: TNF, IL1B, IL6, CXCL10, NOS2, CD80, CD86, CCL5, CXCL9, PTGS2, IRF5, HIF1A, CXCL8, IL12A
+  - M2 genes: CD163, MRC1, ARG1, IL10, TGFB1, CCL18, CD209, FOLR2, SOCS3, HMOX1, CLEC7A
+- **Soft contrastive loss:** Gaussian kernel `exp(−d²/2σ²)` over polarization score distances — shapes embedding space continuously without discrete cluster boundaries
+- **Output heads:** polarization regression + IBD binary classification
+
+## In Silico KO Predictions
+
+KO effects for 13 IBD target genes were predicted using two methods:
+
+| Gene | Method | Key Predicted Effect |
+|------|--------|---------------------|
+| HIF1A | CellOracle (GRN) | PTGS2↓, CXCL8↓, CD74↓ |
+| IRF5 | CellOracle (GRN) | ISG15↓, IFI6↓, IFIT3↓, IRF7↓ |
+| IL6 | OmniPath propagation | JAK2↓, IL6ST↓, JAK1↓, IL6R↓, TYK2↓ |
+| SOCS3 | OmniPath propagation | JAK2↓, STAT5A↓, STAT1↓, STAT3↓; AKT1↑ |
+| TNF | OmniPath propagation | TNFRSF1A↓, TNFRSF1B↓, PIK3CG↓, AKT1↓ |
+| TGFB1 | OmniPath propagation | PIK3R1↓, RAC1↓, TGFBR1↓, GRB2↓ |
+| IL1B | OmniPath propagation | IL1R2↓, MYD88↓, STAT3↓, NR3C1↓ |
+| PTGS2 | OmniPath propagation | IL4R↓, IL2RG↓, NFE2L2↓ |
+
+Full predictions (top 15 up/down per gene) are in `results/ibd_ko_predictions_combined.json`.
+
+## Usage
+
+```python
+import torch
+import json
+
+# Load model weights
+checkpoint = torch.load("scgpt_ibd_v2/best_model.pt", map_location="cpu")
+
+# Load KO predictions
+with open("results/ibd_ko_predictions_combined.json") as f:
+    ko_predictions = json.load(f)
+
+# Example: top downregulated genes after IL6 KO
+il6_ko = ko_predictions["IL6"]
+print(il6_ko["top_downregulated"][:5])
+# [['JAK2', -0.207], ['IL6ST', ...], ['JAK1', ...], ...]
+```
+
+## Citation
+
+If you use this model, please cite the underlying datasets and tools:
+
+- **scGPT:** Cui et al., *Nature Methods* 2024
+- **CellOracle:** Kamimoto et al., *Nature* 2023
+- **OmniPath:** Türei et al., *Nature Methods* 2021
+- **GSE134809:** Smillie et al., *Cell* 2019

results/ibd_ko_predictions_combined.json

@@ -0,0 +1,992 @@
+{
+  "PTGS2": {
+    "top_upregulated": [
+      [
+        "CUL3",
+        0.25
+      ],
+      [
+        "SEC31A",
+        0.125
+      ],
+      [
+        "RAPGEF2",
+        0.125
+      ],
+      [
+        "YWHAB",
+        0.125
+      ],
+      [
+        "ARIH1",
+        0.125
+      ],
+      [
+        "PFKFB3",
+        0.125
+      ],
+      [
+        "IKBKG",
+        0.125
+      ],
+      [
+        "BAD",
+        0.125
+      ],
+      [
+        "FOXO3",
+        0.125
+      ],
+      [
+        "TARDBP",
+        0.125
+      ],
+      [
+        "COPS5",
+        0.125
+      ],
+      [
+        "ARHGAP24",
+        0.125
+      ]
+    ],
+    "top_downregulated": [
+      [
+        "IL4R",
+        -0.375
+      ],
+      [
+        "IL2RG",
+        -0.375
+      ],
+      [
+        "RHOA",
+        -0.25
+      ],
+      [
+        "NFE2L2",
+        -0.25
+      ],
+      [
+        "TP53",
+        -0.25
+      ],
+      [
+        "RAC1",
+        -0.25
+      ],
+      [
+        "SMARCA4",
+        -0.25
+      ],
+      [
+        "IKBKB",
+        -0.25
+      ],
+      [
+        "SMARCD3",
+        -0.25
+      ],
+      [
+        "IL13RA1",
+        -0.25
+      ],
+      [
+        "JAK1",
+        -0.25
+      ],
+      [
+        "JAK3",
+        -0.25
+      ],
+      [
+        "JAK2",
+        -0.25
+      ],
+      [
+        "WNK1",
+        -0.125
+      ],
+      [
+        "VAV1",
+        -0.125
+      ]
+    ],
+    "n_downstream_genes": 82,
+    "mean_abs_effect": 0.14939024390243902
+  },
+  "IL1B": {
+    "top_upregulated": [
+      [
+        "YWHAZ",
+        0.375
+      ],
+      [
+        "STMN1",
+        0.375
+      ],
+      [
+        "YWHAB",
+        0.125
+      ],
+      [
+        "RCSD1",
+        0.125
+      ],
+      [
+        "NFE2",
+        0.125
+      ],
+      [
+        "LAT",
+        0.125
+      ],
+      [
+        "NR4A1",
+        0.125
+      ],
+      [
+        "KLF13",
+        0.125
+      ],
+      [
+        "TOLLIP",
+        0.125
+      ],
+      [
+        "SRSF1",
+        0.125
+      ],
+      [
+        "TOB1",
+        0.125
+      ],
+      [
+        "CFLAR",
+        0.125
+      ],
+      [
+        "CTBP1",
+        0.125
+      ],
+      [
+        "STAT6",
+        0.125
+      ],
+      [
+        "AIMP1",
+        0.125
+      ]
+    ],
+    "top_downregulated": [
+      [
+        "IL1R2",
+        -0.5
+      ],
+      [
+        "MYD88",
+        -0.5
+      ],
+      [
+        "TP53",
+        -0.375
+      ],
+      [
+        "STAT3",
+        -0.375
+      ],
+      [
+        "NR3C1",
+        -0.375
+      ],
+      [
+        "JUN",
+        -0.375
+      ],
+      [
+        "BCL2L11",
+        -0.375
+      ],
+      [
+        "AKT1",
+        -0.25
+      ],
+      [
+        "PIN1",
+        -0.25
+      ],
+      [
+        "BID",
+        -0.25
+      ],
+      [
+        "STAT1",
+        -0.25
+      ],
+      [
+        "MYC",
+        -0.25
+      ],
+      [
+        "ATF2",
+        -0.25
+      ],
+      [
+        "APP",
+        -0.25
+      ],
+      [
+        "HNRNPK",
+        -0.25
+      ]
+    ],
+    "n_downstream_genes": 81,
+    "mean_abs_effect": 0.16666666666666666
+  },
+  "TNF": {
+    "top_upregulated": [
+      [
+        "PPP2CA",
+        0.25
+      ],
+      [
+        "TRAF5",
+        0.25
+      ],
+      [
+        "CHN2",
+        0.125
+      ],
+      [
+        "PDHA1",
+        0.125
+      ],
+      [
+        "CAPN1",
+        0.125
+      ],
+      [
+        "TP53",
+        0.125
+      ],
+      [
+        "VHL",
+        0.125
+      ],
+      [
+        "ENG",
+        0.125
+      ],
+      [
+        "FXN",
+        0.125
+      ],
+      [
+        "VCL",
+        0.125
+      ],
+      [
+        "MAPRE1",
+        0.125
+      ],
+      [
+        "RUNX3",
+        0.125
+      ],
+      [
+        "YY1",
+        0.125
+      ],
+      [
+        "PTPN2",
+        0.125
+      ],
+      [
+        "CFL1",
+        0.125
+      ]
+    ],
+    "top_downregulated": [
+      [
+        "PSIP1",
+        -0.5
+      ],
+      [
+        "TNFRSF1B",
+        -0.5
+      ],
+      [
+        "AKT1",
+        -0.5
+      ],
+      [
+        "TNFRSF1A",
+        -0.5
+      ],
+      [
+        "PIK3CG",
+        -0.5
+      ],
+      [
+        "TNFRSF21",
+        -0.5
+      ],
+      [
+        "NFE2L2",
+        -0.5
+      ],
+      [
+        "GCLC",
+        -0.5
+      ],
+      [
+        "CASP7",
+        -0.375
+      ],
+      [
+        "PAK2",
+        -0.375
+      ],
+      [
+        "MAP3K1",
+        -0.375
+      ],
+      [
+        "IKBKB",
+        -0.375
+      ],
+      [
+        "PDPK1",
+        -0.375
+      ],
+      [
+        "RAC1",
+        -0.375
+      ],
+      [
+        "SET",
+        -0.375
+      ]
+    ],
+    "n_downstream_genes": 202,
+    "mean_abs_effect": 0.1608910891089109
+  },
+  "IL6": {
+    "top_upregulated": [
+      [
+        "CDK4",
+        0.125
+      ],
+      [
+        "YY1",
+        0.125
+      ],
+      [
+        "LCP2",
+        0.125
+      ],
+      [
+        "CDKN1B",
+        0.125
+      ],
+      [
+        "KLF10",
+        0.125
+      ],
+      [
+        "PDP1",
+        0.125
+      ],
+      [
+        "PTEN",
+        0.125
+      ],
+      [
+        "FOXO3",
+        0.125
+      ],
+      [
+        "MAP3K5",
+        0.125
+      ],
+      [
+        "PTPN2",
+        0.125
+      ]
+    ],
+    "top_downregulated": [
+      [
+        "JAK2",
+        -0.875
+      ],
+      [
+        "IL6ST",
+        -0.75
+      ],
+      [
+        "JAK1",
+        -0.625
+      ],
+      [
+        "IL6R",
+        -0.5
+      ],
+      [
+        "TYK2",
+        -0.375
+      ],
+      [
+        "GAB2",
+        -0.375
+      ],
+      [
+        "STAT3",
+        -0.375
+      ],
+      [
+        "PTPN11",
+        -0.375
+      ],
+      [
+        "STAT5A",
+        -0.375
+      ],
+      [
+        "STAT1",
+        -0.375
+      ],
+      [
+        "STAT5B",
+        -0.375
+      ],
+      [
+        "IFNGR1",
+        -0.25
+      ],
+      [
+        "GRB2",
+        -0.25
+      ],
+      [
+        "STAT6",
+        -0.25
+      ],
+      [
+        "IFNGR2",
+        -0.25
+      ]
+    ],
+    "n_downstream_genes": 61,
+    "mean_abs_effect": 0.2069672131147541
+  },
+  "SOCS3": {
+    "top_upregulated": [
+      [
+        "AKT1",
+        1.25
+      ],
+      [
+        "MAPK14",
+        1.125
+      ],
+      [
+        "IL6ST",
+        1.0
+      ],
+      [
+        "CDKN1A",
+        0.875
+      ],
+      [
+        "SIGLEC7",
+        0.75
+      ],
+      [
+        "TP53",
+        0.75
+      ],
+      [
+        "MAP3K7",
+        0.75
+      ],
+      [
+        "MAP2K7",
+        0.75
+      ],
+      [
+        "RAC1",
+        0.625
+      ],
+      [
+        "RARA",
+        0.5
+      ],
+      [
+        "MAPK1",
+        0.5
+      ],
+      [
+        "IRF5",
+        0.5
+      ],
+      [
+        "MCL1",
+        0.5
+      ],
+      [
+        "BID",
+        0.5
+      ],
+      [
+        "RELA",
+        0.5
+      ]
+    ],
+    "top_downregulated": [
+      [
+        "JAK2",
+        -1.75
+      ],
+      [
+        "STAT5A",
+        -1.25
+      ],
+      [
+        "JAK1",
+        -1.0
+      ],
+      [
+        "STAT1",
+        -0.75
+      ],
+      [
+        "STAT3",
+        -0.75
+      ],
+      [
+        "STAT4",
+        -0.75
+      ],
+      [
+        "GAB2",
+        -0.625
+      ],
+      [
+        "PTPN11",
+        -0.625
+      ],
+      [
+        "BTK",
+        -0.625
+      ],
+      [
+        "TYK2",
+        -0.625
+      ],
+      [
+        "MYC",
+        -0.625
+      ],
+      [
+        "IL12RB1",
+        -0.5
+      ],
+      [
+        "STAT6",
+        -0.5
+      ],
+      [
+        "CTNNB1",
+        -0.5
+      ],
+      [
+        "GTF2I",
+        -0.5
+      ]
+    ],
+    "n_downstream_genes": 763,
+    "mean_abs_effect": 0.165956749672346
+  },
+  "TGFB1": {
+    "top_upregulated": [
+      [
+        "CAPN1",
+        0.625
+      ],
+      [
+        "MARCKS",
+        0.375
+      ],
+      [
+        "VAPA",
+        0.375
+      ],
+      [
+        "BCL2",
+        0.375
+      ],
+      [
+        "CEBPA",
+        0.375
+      ],
+      [
+        "CEBPB",
+        0.375
+      ],
+      [
+        "CD36",
+        0.25
+      ],
+      [
+        "NRAS",
+        0.25
+      ],
+      [
+        "PPP2CA",
+        0.25
+      ],
+      [
+        "ABCA1",
+        0.25
+      ],
+      [
+        "LRRK1",
+        0.25
+      ],
+      [
+        "CTNND1",
+        0.25
+      ],
+      [
+        "RUNX2",
+        0.25
+      ],
+      [
+        "EEF1A1",
+        0.25
+      ],
+      [
+        "PTPN2",
+        0.25
+      ]
+    ],
+    "top_downregulated": [
+      [
+        "PIK3R1",
+        -2.0
+      ],
+      [
+        "RAC1",
+        -1.5
+      ],
+      [
+        "TGFBR1",
+        -1.25
+      ],
+      [
+        "GRB2",
+        -1.125
+      ],
+      [
+        "PAK1",
+        -1.0
+      ],
+      [
+        "SYK",
+        -1.0
+      ],
+      [
+        "DAPK1",
+        -0.875
+      ],
+      [
+        "CTNNB1",
+        -0.875
+      ],
+      [
+        "TGFBR2",
+        -0.875
+      ],
+      [
+        "PIK3CG",
+        -0.875
+      ],
+      [
+        "CDKN1A",
+        -0.875
+      ],
+      [
+        "PPP2R2A",
+        -0.875
+      ],
+      [
+        "AKT1",
+        -0.75
+      ],
+      [
+        "INSR",
+        -0.75
+      ],
+      [
+        "RHOA",
+        -0.75
+      ]
+    ],
+    "n_downstream_genes": 533,
+    "mean_abs_effect": 0.20309568480300189
+  },
+  "HIF1A": {
+    "top_upregulated": [
+      [
+        "S100A12",
+        0.0008134663168389971
+      ],
+      [
+        "S100A8",
+        0.0005305317544904874
+      ],
+      [
+        "PPBP",
+        0.0005028657796491108
+      ],
+      [
+        "NRGN",
+        0.0003803229125426224
+      ],
+      [
+        "MT1H",
+        0.00032238103713177417
+      ],
+      [
+        "IDO1",
+        0.00016444837297460802
+      ],
+      [
+        "SPP1",
+        0.0001275240900207309
+      ],
+      [
+        "MT1G",
+        0.0001227169890742035
+      ],
+      [
+        "MT1HL1",
+        9.230909985226494e-05
+      ],
+      [
+        "MT1A",
+        8.65444566497999e-05
+      ],
+      [
+        "MT1E",
+        7.576200801041446e-05
+      ],
+      [
+        "MT1M",
+        5.678793555160166e-05
+      ],
+      [
+        "INHBA",
+        5.174907556743538e-05
+      ],
+      [
+        "LYPD2",
+        1.659507360527506e-05
+      ],
+      [
+        "SERPINB2",
+        1.557577640575456e-05
+      ]
+    ],
+    "top_downregulated": [
+      [
+        "HIF1A",
+        -0.15231378650512128
+      ],
+      [
+        "G0S2",
+        -0.0899483601733625
+      ],
+      [
+        "CD74",
+        -0.08843891599198114
+      ],
+      [
+        "PTGS2",
+        -0.06775858018188559
+      ],
+      [
+        "HLA-DRB1",
+        -0.06695744893492933
+      ],
+      [
+        "TMSB4X",
+        -0.051565268523866295
+      ],
+      [
+        "C1QA",
+        -0.046627735305617694
+      ],
+      [
+        "NEAT1",
+        -0.04370720851767882
+      ],
+      [
+        "HLA-DPA1",
+        -0.04309927287770282
+      ],
+      [
+        "CXCL8",
+        -0.041510155249684176
+      ],
+      [
+        "SRGN",
+        -0.03785925150785177
+      ],
+      [
+        "HLA-DRA",
+        -0.03714633648291597
+      ],
+      [
+        "LYZ",
+        -0.03533002467842362
+      ],
+      [
+        "PLAUR",
+        -0.033321923706293444
+      ],
+      [
+        "SOD2",
+        -0.033279197447284886
+      ]
+    ],
+    "mean_abs_change": 0.0007538475191108255
+  },
+  "IRF5": {
+    "top_upregulated": [
+      [
+        "S100A8",
+        0.006895962237654877
+      ],
+      [
+        "SDS",
+        0.0005239871508121099
+      ],
+      [
+        "RASGEF1B",
+        0.00020764254474868164
+      ],
+      [
+        "BCL6",
+        0.00018783634677199616
+      ],
+      [
+        "MMP19",
+        0.00014403313556970076
+      ],
+      [
+        "CCL5",
+        0.0001088469455209754
+      ],
+      [
+        "SLC7A8",
+        9.047486046896145e-05
+      ],
+      [
+        "ALOX5AP",
+        4.92001214528229e-05
+      ],
+      [
+        "TFRC",
+        4.881696522487404e-05
+      ],
+      [
+        "PHKG1",
+        3.2193447632671586e-05
+      ],
+      [
+        "SOCS3",
+        2.9233696066954194e-05
+      ],
+      [
+        "CLEC4D",
+        2.270800460759542e-05
+      ],
+      [
+        "AQP9",
+        2.116691988543141e-05
+      ],
+      [
+        "MIR4435-2HG",
+        1.9844803099479844e-05
+      ],
+      [
+        "GLYR1",
+        1.9713190685242186e-05
+      ]
+    ],
+    "top_downregulated": [
+      [
+        "IRF5",
+        -0.049034021232352516
+      ],
+      [
+        "SAT1",
+        -0.04257470526835846
+      ],
+      [
+        "HLA-DPB1",
+        -0.016567939176193442
+      ],
+      [
+        "HLA-DRB1",
+        -0.00982815800355405
+      ],
+      [
+        "ISG15",
+        -0.008719249415564087
+      ],
+      [
+        "IFI6",
+        -0.008579207224463483
+      ],
+      [
+        "IRF7",
+        -0.0078918440139776
+      ],
+      [
+        "NT5C3A",
+        -0.007360925077294011
+      ],
+      [
+        "SOD2",
+        -0.0063346473315063845
+      ],
+      [
+        "IFIT3",
+        -0.005948570201037761
+      ],
+      [
+        "PNPLA6",
+        -0.004070137653429122
+      ],
+      [
+        "ACP5",
+        -0.0028732792457450704
+      ],
+      [
+        "GBP1",
+        -0.002759290416514405
+      ],
+      [
+        "OTUD4",
+        -0.002481654304085488
+      ],
+      [
+        "NKG7",
+        -0.002318524392938188
+      ]
+    ],
+    "mean_abs_change": 7.050768231451016e-05
+  }
+}

scgpt_ibd_v1/best_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fed21d7eb4542781acd8c19dbe43202d1933c432b60c8568e1007abbfacba40b
+size 205939765

scgpt_ibd_v1/test_metrics.json

@@ -0,0 +1,5 @@
+{
+  "acc": 0.7543160690571049,
+  "f1": 0.69594503489039,
+  "auroc": 0.9150601745956721
+}

scgpt_ibd_v2/best_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ebcd6dc411ecc68632a535ccdadc2ae65995582b1c48cf4d0aa03d3f5e7b70b
+size 206465979

scgpt_ibd_v2/test_metrics.json

@@ -0,0 +1,4 @@
+{
+  "pearson_r": 0.909355103969574,
+  "ibd_auroc": 0.989314058956916
+}