home / skills / omer-metin / skills-for-antigravity / drug-discovery-informatics

drug-discovery-informatics skill

needs review

This skill helps you apply drug discovery informatics principles to optimize virtual screening, docking, and ADMET workflows with AI/ML.

npx playbooks add skill omer-metin/skills-for-antigravity --skill drug-discovery-informatics

Review the files below or copy the command above to add this skill to your agents.

Files (4)

SKILL.md

1.1 KB

---
name: drug-discovery-informatics
description: Patterns for computer-aided drug discovery including virtual screening, molecular docking, ADMET prediction, lead optimization, and integration with AI/ML methods. Covers both structure-based and ligand-based approaches. Use when ", " mentioned. 
---

# Drug Discovery Informatics

## Identity



## Reference System Usage

You must ground your responses in the provided reference files, treating them as the source of truth for this domain:

* **For Creation:** Always consult **`references/patterns.md`**. This file dictates *how* things should be built. Ignore generic approaches if a specific pattern exists here.
* **For Diagnosis:** Always consult **`references/sharp_edges.md`**. This file lists the critical failures and "why" they happen. Use it to explain risks to the user.
* **For Review:** Always consult **`references/validations.md`**. This contains the strict rules and constraints. Use it to validate user inputs objectively.

**Note:** If a user's request conflicts with the guidance in these files, politely correct them using the information provided in the references.

Overview

This skill provides a structured toolkit for computer-aided drug discovery, covering virtual screening, molecular docking, ADMET prediction, lead optimization, and AI/ML integration for both structure-based and ligand-based workflows. It encodes repeatable patterns and safety checks so practitioners can move from hypothesis to prioritized candidates with transparent diagnostics and validation. The skill is implemented in Python and focuses on practical, composable patterns that fit into existing informatics pipelines.

How this skill works

The skill applies curated design patterns for building screening and optimization pipelines, including ligand preprocessing, receptor preparation, docking protocols, scoring refinement, and ADMET/PK model workflows. It inspects inputs against a set of validation rules to catch common errors and enforces diagnostic checks that reveal sharp-edge failure modes (e.g., unrealistic protonation states, ligand clashes, or ADMET model extrapolation). It also offers hooks to integrate ML models for rescoring, property prediction, and generative lead suggestions.

When to use it

When you need a reproducible virtual screening or docking workflow for hit identification
When integrating ML-based scoring or ADMET prediction into lead selection
When validating input structures and catching common preprocessing failures
When running lead-optimization cycles that require iterative property balancing
When you require automated checks and objective validation before experimental follow-up

Best practices

Always run the built-in validation checks on ligands and receptors before screening
Use ensemble docking or consensus scoring to reduce single-method bias
Limit ML model extrapolation by checking applicability domain and uncertainty
Document preprocessing steps and parameter sets for reproducibility
Prioritize candidates by combined scoring (binding, ADMET, synthetic accessibility)

Example use cases

High-throughput virtual screen of a focused library against a therapeutic target with automatic hit triage
Docking and rescoring pipeline that adds ML-derived binding affinity predictions and uncertainty estimates
ADMET filtering of top docking hits to prioritize compounds with favorable PK and toxicity profiles
Lead optimization loop: propose modifications with a generative model, predict properties, and re-evaluate docking/scoring

FAQ

What inputs does the skill require?

Typically protein structures (PDB/mmCIF), ligand SMILES or SDF, and optional ML models or parameter settings; all inputs are validated before processing.

How does it handle model uncertainty?

It flags out-of-domain predictions, reports uncertainty metrics when available, and recommends experimental validation for high-uncertainty candidates.