drug-admet-prediction

What problem does it solve?

It eliminates slow, manual early-stage triage by computing common ADMET-relevant physicochemical descriptors and rule-based drug-likeness heuristics directly from SMILES.

Core Features & Use Cases

• Physicochemical descriptor calculation: Computes molecular weight (average and exact), Wildman-Crippen cLogP, TPSA, HBD/HBA, rotatable bonds, ring counts, aromatic rings, heavy atoms, fractionCSP3, molar refractivity.
• Rule-based triage heuristics: Evaluates Lipinski Rule of Five (pass defined as ≤ 1 violation), Veber oral bioavailability heuristic (TPSA and rotatable-bonds thresholds plus reporting HBD+HBA ≤ 12), and computes QED.
• Workflow for library triage: Suitable for comparing candidate sets early in discovery to flag likely drug-like molecules before any experimental or ML endpoint prediction.

Quick Start

Use the drugdisc MCP tool to compute descriptors and heuristics for a batch of SMILES by asking: run mcp_drugdisc_compute_molecular_descriptors on .agents/skills/drug-admet-prediction/examples/compounds.smi and write results to compounds_admet.json.