computational-drug-discovery

This pipeline provides a way to perform pharmaceutical compounds virtual screening using similarity-based analysis, ligand-based and structure-based techniques. The pipeline contains a collections of modules to perform a variety of analysis.

Computational Drug Discovery Using Machine Learning

A Classification model for identifying stable tautomers

This is a compilation of machine learning projects I have worked on.

Portfolio landing page — computational drug discovery & reproducible omics

Traceable preliminary models for GPCR direct affinity and operational bias-signalling.

End-to-end ML pipeline for SELFIES-based molecular generation, Transformer fine-tuning, docking, and drug-likeness evaluation for HIV-1 protease inhibitors.

Automated CADD pipeline: AlphaFold structure prediction → Vina docking → Lipinski filtering, orchestrated with Nextflow (TREM2 target)

An experiment in agentic AI workflows: driving an autonomous agent through a full computational drug-discovery pipeline (Schrodinger Glide/SiteMap/Jaguar, ChEMBL, AlphaFold) against silicosis targets. Includes the positive-control benchmark that invalidated its own results. Not medical advice.

An experiment in agentic AI workflows: autonomous agents driven through an evidence-gated computational triage pipeline (Schrodinger, AlphaFold, Vina) for DKK-1/LRP6 hair-regrowth hypotheses. Every claim provenance-tagged. Not medical advice.

An experiment in agentic AI workflows: autonomous agents driven through an 8-phase computational screening pipeline (Schrodinger SiteMap/Glide/QikProp/Prime, AlphaFold, pathway ODE modelling) against SOSTDC1. All outputs unvalidated, in arbitrary units. Not medical advice.

Drug Candidate Screener

multiparameter optimisation for filtering drugs by chemical properties

A comparative and ablation study exploring different models for protein-ligand binding affinity prediction. The model categories studied in this project include traditional machine learning, graph neural networks, and structural deep learning models.