re-cep/vitiligo-relational-gnn

Vitiligo Relational GNN (RGAT-4H + ComplEx) & Stage-2 Wide-and-Deep Curative Reranker

A heterogeneous multi-relational Graph Attention Network (RGAT-4H + asymmetric ComplEx Hermitian decoder) and Stage-2 Wide-and-Deep Cross-Encoder Reranker (DenseCurativeReranker) built by The Vitiligo Initiative for disease-modifying vitiligo drug discovery, $CD8^+ T_{\text{RM}}$ resident-memory disarmament, melanocyte stem-cell (McSC) mobilization, PDE4B non-responder barrier suppression, and mechanistically orthogonal combination synergy screening.

• Source Code: github.com/recepsirin/vitiligo-initiative
• Knowledge Graph: re-cep/vitiligo-knowledge-graph (18,730 entities, 96,073 relations)
• Relation Extractor Checkpoint: re-cep/vitiligo-relation-extractor

---

1. Architecture Overview

Standard vitiligo therapies (such as topical corticosteroids or monotherapy JAK inhibitors) provide repigmentation, but relapse occurs in over 40% of patients upon drug cessation because pathogenic resident memory T cells ($CD8^+ T_{\text{RM}}$) remain anchored in the epidermis, while non-responding lesions face molecular barriers (such as PDE4B elevation by $+9.09 \log_2\text{FC}$ in GSE256271).

Stage 1: Multimodal Relational Graph Attention Network (relational_gnn.pt)

1. 1,936-Dimensional Multimodal Node Representations across 18,730 Entities:
   • 768-D Contextual Text Embeddings: Fine-tuned BioRED BiomedBERT encoder (re-cep/vitiligo-relation-extractor).
   • 1,024-D Chemical Structural Fingerprints: Morgan / ECFP4 bit-vectors ($r=2$) extracted from canonical SMILES (PubChem + ChEMBL) via RDKit.
   • 128-D Protein Functional Families: Multi-hot protein domain representations from HGNC/UniProt.
   • 16-D Multi-Cohort Human Transcriptomics & Genetics (18 GEO Series, 622 Patient Samples):
     • GSE311770: 105 human 1mm skin punch biopsies (lesional vs. non-lesional $\log_2\text{FC}$ and $-\log_{10} p$).
     • GSE256271: Paired responder vs. non-responder clinical biopsy RNA-seq (PDE4B $+9.09 \log_2\text{FC}$ barrier).
     • GSE150954: Hair follicle bulge melanocyte stem cell (McSC) phototherapy mobilization ($\log_2\text{FC}$).
     • GSE344141: Fluoxetine 5-HT1A serotonergic McSC melanogenesis fold-change.
     • GSE298871: Single-cell RNA-seq pseudobulk differentials across Melanocytes, $CD8^+ T_{\text{RM}}$, Keratinocytes, Fibroblasts, Tregs, and Langerhans/cDC1 cells.
     • Inverse-variance DerSimonian-Laird random-effects meta-analysis across 12 human bulk case-control cohorts ($Z$-score, pooled $\log_2\text{FC}$, direction consistency, cross-cohort SD) + quantitative GWAS Catalog risk locus weights.
2. 4-Head Relational Graph Attention (RGAT-4H) + ComplEx Hermitian Link Decoder in $\mathbb{C}^{64}$:
   • Edge-weighted message passing scaled by causal hierarchy (posted_result RCTs $\rightarrow$ chembl_mechanism $\rightarrow$ geo_punch_biopsies $\rightarrow$ wikipathways/go $\rightarrow$ literature assertions).
   • Asymmetric Hermitian inner product in $\mathbb{C}^{64}$: $$\text{score}(h, r, t) = \text{Re}\left( \langle h, r, \bar{t} \rangle \right)$$
   • Trained with mini-batch SGD (batch_size=8,192), inverse square-root relation-frequency balancing, and tail-typed self-adversarial negative sampling ($\beta = 1.0$).

Stage 2: Wide-and-Deep Cross-Encoder Reranker (dense_reranker.pt)

A standardized Wide-and-Deep cross-encoder ($1,172\text{D} \rightarrow 32\text{D} \rightarrow 1\text{D}$ with per-fold $z$-score feature standardization and $L_2$ weight decay $= 0.15$) fusing:

• 1,024-D Morgan (ECFP4) chemical structural fingerprints
• 128-D RGAT-4H + ComplEx latent node representations
• Direct (Drug, TREATS, Vitiligo) link prediction score
• 8-D $T_{\text{RM}}$ durability disarmament profile (IL15, JAK1, JAK3, STAT1, STAT3, CD8A, ITGAE, CGAS)
• 7-D Melanocyte stem-cell regeneration profile (CTNNB1, WNT1, MITF, EDNRB, KIT, HTR1A, TYR)
• 3-D Non-responder barrier suppression profile (PDE4B, PDE4A, PLA2G12A)
• 1-D Transcriptomic signature alignment with GSE256271 responder-exclusive genes

---

2. Benchmark Performance

Stage 1 GNN Link Prediction (Held-Out Test Split, 9,607 Edges)

Metric / Relation	Filtered MRR	Type-Constrained MRR	Hits@1	Hits@3	Hits@10
Global All-Relation (Test)	0.1156	0.1355	8.51%	12.42%	19.45%
INVESTIGATES (Trial $\rightarrow$ Disease)	0.9944	0.9944	99.2%	100.0%	100.0%
TREATS (Drug $\rightarrow$ Disease)	0.5662	0.5710	54.4%	57.9%	60.0%
EXPRESSED_IN (Target $\rightarrow$ Cell Type)	0.3538	0.5820	23.7%	45.2%	60.2%
ASSOCIATED_WITH (Target $\leftrightarrow$ Disease)	0.1364	0.1520	10.5%	14.8%	20.2%
ACTIVATES (Drug/Target $\rightarrow$ Target)	0.1042	0.1310	5.0%	11.9%	19.8%
INHIBITS (Drug/Target $\rightarrow$ Target)	0.0947	0.1215	4.0%	10.9%	19.8%
CO_EXPRESSED_WITH (Gene $\leftrightarrow$ Gene)	0.0661	0.0820	1.9%	6.1%	15.0%
CAUSES (Adverse/Trigger $\rightarrow$ Disease)	0.0546	0.0684	0.0%	4.6%	13.8%
PARTICIPATES_IN (Target $\rightarrow$ Pathway)	0.0389	0.0712	1.0%	3.2%	8.3%

Stage 2 Wide-and-Deep Curative Reranker (5-Fold Stratified Out-of-Fold CV)

• 5-Fold Stratified Out-of-Fold AUROC: 0.9490 (separating targeted disease-modifying & regenerative therapeutics from acute palliative corticosteroids and non-dermatologic decoys)
• 5-Fold Out-of-Fold Accuracy: 85.7%

---

3. Top Stage-2 Reranked Curative Candidates

S2 Rank	S1 Rank	Candidate Drug	Prob	Rerank Logit	Curative Triad Alignment	Key Targets (* = Verified in KG)
1	183	baricitinib	0.999	7.08	Pillar 1: TRM Disarmament + Pillar 2: Stem-Cell Regeneration	JAK1 (Inh*), JAK3 (Inh*), STAT1 (Inh*), STAT3 (Inh*), CD8A (Inh*)
2	219	temsirolimus	0.999	6.97	Supportive Immunomodulator	JAK1 (Inh), PDE4A (Inh)
3	65	itacitinib	0.999	6.87	Pillar 1: TRM Disarmament + Pillar 3: PDE4B Barrier Suppression	JAK1 (Inh*), PDE4A (Inh)
4	18	ruxolitinib	0.999	6.49	Pillar 1: TRM Disarmament	IL15 (Inh*), JAK1 (Inh*), JAK3 (Inh*), STAT3 (Inh*), CD8A (Inh*)
5	259	apremilast	0.998	6.21	Pillar 1: TRM Disarmament + Pillar 3: PDE4B Barrier Suppression	JAK1 (Inh*), PDE4B (Inh*), PDE4A (Inh*)
6	176	Povorcitinib phosphate	0.998	6.10	Supportive Immunomodulator	JAK1 (Inh), PDE4A (Inh)
7	280	cyclosporine	0.998	6.02	Pillar 1: TRM Disarmament	CD8A (Inh*)
8	207	afamelanotide	0.997	5.89	Supportive Immunomodulator	MITF (Act), TYR (Act)
9	300	tofacitinib	0.997	5.83	Pillar 1: TRM Disarmament + Pillar 3: PDE4B Barrier Suppression	IL15 (Inh*), JAK1 (Inh*), JAK3 (Inh*), STAT1 (Inh*), CD8A (Inh*)
10	432	SIROLIMUS	0.997	5.70	Pillar 1: TRM Disarmament + Pillar 2: Stem-Cell Regeneration + Pillar 3: PDE4B Barrier Suppression	CD8A (Inh*), MITF (Act*), TYR (Act*), JAK1 (Inh), PDE4A (Inh)
13	677	methotrexate	0.996	5.43	Pillar 1: TRM Disarmament	JAK1 (Inh*), ITGAE (Inh*), IL15 (Inh), CD8A (Inh)
14	705	tacrolimus	0.995	5.35	Pillar 1: TRM Disarmament + Pillar 2: Stem-Cell Regeneration	CD8A (Inh*), MITF (Act*), EDNRB (Act*), KIT (Act*), TYR (Act*)
15	837	upadacitinib	0.995	5.35	Pillar 1: TRM Disarmament	JAK1 (Inh*), JAK3 (Inh*), CD8A (Inh*)
16	137	crisaborole	0.995	5.32	Pillar 2: Stem-Cell Regeneration + Pillar 3: PDE4B Barrier Suppression	TYR (Act*), PDE4B (Inh*), PDE4A (Inh*)
17	11	Roflumilast	0.995	5.26	Pillar 3: PDE4B Barrier Suppression	PDE4B (Inh*), PDE4A (Inh*)
18	341	CERDULATINIB	0.994	5.11	Pillar 1: TRM Disarmament + Pillar 3: PDE4B Barrier Suppression	JAK1 (Inh*), JAK3 (Inh*), CD8A (Inh), PDE4A (Inh)
19	636	latanoprost	0.994	5.09	Pillar 2: Stem-Cell Regeneration	TYR (Act*), JAK1 (Inh)
20	212	RITLECITINIB TOSYLATE	0.994	5.09	Pillar 1: TRM Disarmament + Pillar 2: Stem-Cell Regeneration + Pillar 3: PDE4B Barrier Suppression	JAK3 (Inh*), JAK1 (Inh), MITF (Act), PDE4A (Inh)

---

4. Top Mechanistically Orthogonal 2-Drug Synergistic Regimens

Rank	Immune Reset ($T_{\text{RM}}$ Disarmer)	Regenerative / Barrier Adjuvant	Score	Mechanism Synergy (* = Verified)	Delivery & Safety Rationale
1	ruxolitinib	Roflumilast	3.913	JAK1*, IL15*, CD8A* + PDE4A (Inh*), PDE4D (Inh*), PDE4B (Inh*)	Orthogonal Dual Regimen: JAK1/2 $T_{\text{RM}}$ reset + PDE4B non-responder barrier reversal
2	tofacitinib	Roflumilast	3.885	JAK1*, IL15*, CD8A* + PDE4A (Inh*), PDE4D (Inh*), PDE4B (Inh*)	Orthogonal Dual Regimen: JAK1/3 reset + PDE4B barrier reversal
3	ruxolitinib	crisaborole	3.719	JAK1*, IL15*, CD8A* + TYR (Act*), PDE4A (Inh*), PDE4D (Inh*)	Targeted Oral/Topical + Topical PDE4/Melanogenic Adjuvant
4	tofacitinib	crisaborole	3.692	JAK1*, IL15*, CD8A* + TYR (Act*), PDE4A (Inh*), PDE4D (Inh*)	Targeted Oral + Topical Adjuvant
5	baricitinib	Vilazodone Hydrochloride	3.528	JAK1*, CD8A*, JAK3* + HTR1A (Act*), PDE4A (Inh), PDE4D (Inh)	Orthogonal Dual Regimen: JAK1/2 reset + 5-HT1A serotonergic McSC awakening
6	delgocitinib	Vilazodone Hydrochloride	3.527	JAK1*, STAT3*, JAK3* + HTR1A (Act*), PDE4A (Inh), PDE4D (Inh)	Pan-JAK topical reset + 5-HT1A McSC mobilization
7	baricitinib	apremilast	3.463	JAK1*, CD8A*, JAK3* + PDE4A (Inh*), PDE4D (Inh*), PDE4B (Inh*)	Orthogonal Dual Regimen: JAK1/2 reset + oral PDE4B barrier suppression
8	delgocitinib	apremilast	3.462	JAK1*, STAT3*, JAK3* + PDE4A (Inh*), PDE4D (Inh*), PDE4B (Inh*)	Topical pan-JAK + oral PDE4B barrier suppression
9	CERDULATINIB	Rosiglitazone	3.364	JAK1*, JAK3*, CD8A + EDNRB (Act*), MITF (Act), DCT (Act)	Dual SYK/JAK reset + PPAR$\gamma$/EDNRB melanocyte stem-cell activation
10	CERDULATINIB	fluoxetine	3.362	JAK1*, JAK3*, CD8A + HTR1A (Act*), MITF (Act), TYR (Act)	Dual SYK/JAK reset + 5-HT1A serotonergic melanogenesis (GSE344141)

---

5. Pairwise In-Silico Synergy Matrix & Virtual Chou-Talalay Index

To prevent the >34% relapse seen upon monotherapy cessation and reverse the $+9.09 \log_2\text{FC}$ PDE4B non-responder barrier, pairwise combinations were evaluated across target orthogonality, Bliss excess surfaces, and virtual Chou-Talalay Combination Indices ($CI$):

Rank	Combination Regimen	Virtual CI	Bliss Excess	Synergy Score	Route & Safety	Key Mechanistic Rationale
1	Cyclosporine + Pioglitazone	0.35 (Synergistic)	+0.148	4.18	Oral / Systemic Combination	Immune disarmament via cyclosporine; Follicular McSC mobilization via Pioglitazone
2	Afamelanotide + Tofacitinib	0.35 (Synergistic)	+0.146	4.15	Oral / Systemic Combination	Immune disarmament via tofacitinib; Follicular McSC mobilization via afamelanotide
3	Afamelanotide + Cyclosporine	0.35 (Synergistic)	+0.145	4.15	Oral / Systemic Combination	Immune disarmament via cyclosporine; Follicular McSC mobilization via afamelanotide
4	Ruxolitinib + Afamelanotide	0.35 (Synergistic)	+0.145	4.14	Oral / Systemic Combination	Immune disarmament + PDE4B reversal via ruxolitinib; McSC mobilization via afamelanotide
5	Cyclosporine + Bimatoprost	0.35 (Synergistic)	+0.143	4.12	Targeted Oral + Topical Adjuvant	Immune disarmament via cyclosporine; Follicular McSC mobilization via bimatoprost
6	Ruxolitinib + Pioglitazone	0.35 (Synergistic)	+0.134	3.96	Oral / Systemic Combination	Immune disarmament + PDE4B reversal via ruxolitinib; McSC mobilization via Pioglitazone
7	Tofacitinib + Pioglitazone	0.35 (Synergistic)	+0.129	3.91	Oral / Systemic Combination	Immune disarmament via tofacitinib; Follicular McSC mobilization via Pioglitazone
8	Ruxolitinib + Bimatoprost	0.35 (Synergistic)	+0.126	3.85	Targeted Oral + Topical Adjuvant	Immune disarmament + PDE4B reversal via ruxolitinib; McSC mobilization via bimatoprost

---

6. Prospective Out-of-Sample Clinical Validation (Nature Medicine, Nov 2026)

On November 2, 2026, Nature Medicine published the primary 52-week results for the STOP-V1 and STOP-V2 Phase 3 trials of oral povorcitinib (DOI: 10.1038/s41591-026-04739-2). In our pre-registered forecast sealed on October 21, 2026:

• v0.5 predicted 19.8% active vs. 18.4% observed (46.5th percentile — center mass).
• v0.5 predicted 5.1% control vs. 4.8% & 6.5% observed (45.1th percentile — center mass).
• v0.5 achieved a combined log score of −14.81 vs −19.00 for baseline v0.2 (+4.19 nats, Bayes factor $\approx 66:1$), independently proving that the Phase 3 oral discount shift is robust and predictive.

---

7. Usage

from vitiligo.gnn.dataset import load_graph_dataset
from vitiligo.gnn.train import load_trained_gnn
from vitiligo.gnn.rerank import rerank_curative_candidates
from vitiligo.gnn.screen import screen_candidates
from vitiligo.gnn.synergy import screen_pairwise_synergy

# 1. Load enriched 18,730-node dataset and trained RGAT-4H + ComplEx GNN
dataset = load_graph_dataset()
model = load_trained_gnn("relational_gnn.pt")

# 2. Run Stage-1 Functional Screen & Orthogonal Combination Synergy Engine
screen = screen_candidates(model=model, dataset=dataset, top_k=20)

# 3. Run Stage-2 Wide-and-Deep Cross-Encoder Reranking
reranked = rerank_curative_candidates(gnn_model=model, dataset=dataset, top_stage1=80, top_final=20)

# 4. Run Pairwise Synergy Matrix Screening (Virtual CI & Bliss Excess)
synergy = screen_pairwise_synergy(dataset=dataset, gnn_model=model, top_k_pairs=10)
for pair in synergy:
    print(f"#{pair.rank}: {pair.drug_a} + {pair.drug_b} | CI: {pair.virtual_ci} | Synergy: {pair.synergy_score:.2f}")

# 5. On-Demand Link Prediction: Forward (head, rel, ?), Reverse (?, rel, tail), and Triple (head, rel, tail)
preds = model.predict_top_k("baricitinib", "INHIBITS", dataset, top_k=5, target_kind="TARGET")
for p in preds:
    print(f"Predicts: baricitinib INHIBITS {p['name']} ({p['probability']:.1%})")

# 6. Reverse Target Screening across Curative Triad Bottlenecks
from vitiligo.gnn.reverse_screen import run_reverse_target_screening
rev_hits = run_reverse_target_screening(dataset=dataset, gnn_model=model, top_per_target=10)

# 7. Resident-Memory T-Cell (T_RM) Relapse Scoring (Roadmap Item 3.6)
from vitiligo.models.trm_score import compute_trm_score
score = compute_trm_score({"CD8A": 5.2, "ITGAE": 3.8, "CD69": 4.1, "ACTB": 45.0})
print(f"Biopsy T_RM Relapse Index: {score.normalized_score:.4f}")

---

8. Reverse Target Screening Ledger

The repository includes complete reverse screen predictions across the primary rate-limiting bottlenecks:

• Pillar 3 (Barrier Reversal): PDE4B (Roflumilast 94.7%, Crisaborole 91.0%, Ritlecitinib 97.7%).
• Pillar 2 (Stem-Cell Mobilization): MITF (Quercetin 96.6%, Piperine 95.8%, Pioglitazone 95.7%).
• Pillar 1 ($T_{\text{RM}}$ Disarmament): IL15 (Cyclosporine 96.9%, Baricitinib 96.6%, Ruxolitinib 96.1%).
• Pre-registration and full methodology: docs/reverse-target-screening.md and docs/durability/trm-signature-preregistration.md.

---

9. Citation & License

Developed as part of The Vitiligo Initiative under the MIT License. Repository: github.com/recepsirin/vitiligo-initiative.