target-identification

A ChEMBL-driven target-discovery pipeline run on the neglected fungal pathogen Madurella mycetomatis — including the audit that caught the pipeline's own artifact (384 duplicate HDAC4 records ranked first).

One-page, reproducible target-confidence summary that resolves the evidence into an explicit TRUST / CAUTION / REFUSE verdict rather than an unrecoverable score.

Example of generated automatic target dossier with an LLM

Falsifiable drug-target validation: a content-addressed audit engine for direction of effect (inhibit vs. activate), built on Open Targets colocalization and benchmarked against approved drug–target pairs. SHA-locked rules; allowed to refuse.

Temporal biomedical knowledge graph to study target disease linkage over time

scFM-Perturb-Bench: scalable scRNA-seq perturbation & gene-synergy benchmark with Geneformer · GSEA · Scanpy. Audit multi-gene interactions & drug-target potential — GPU/Docker ready foundation-model pipeline.

Unofficial community-maintained Python client for the Open Targets Platform GraphQL API

Rank human genes 0-3 (Tdark->Tclin) by druggability on the TCRD Target Development Level scale, from leakage-audited target-independent biological evidence. CV: R2 0.48, macro-F1 0.60, ranking ROC-AUC 0.88.

AnimaFlow-AI: End-to-end AI drug-discovery methodology hub — single-cell & spatial omics → virtual-cell target ID → multimodal screening (small molecule / peptide / nucleic acid). Case studies: COVID-19, kidney, PD-L1.