qsar-modeling
Builds QSAR / QSPR models using chemprop D-MPNN, MolFormer, Uni-Mol, ChemBERTa, random forest baselines, and Gaussian processes with explicit handling of OECD 5 principles, applicability domain (kNN, leverage, conformal prediction, Mahalanobis), scaffold-balanced splits, ensemble uncertainty, calibration (Platt, isotonic), feature importance (SHAP, atomic attribution), and prospective validation. Use when building target-specific predictive models from in-house bioassay data, ADMET endpoints, or selectivity profiles.Version Compatibility
Reference examples tested with: chemprop 2.0+ (major API change from 1.x), RDKit 2024.09+, scikit-learn 1.4+, MAPIE 0.8+ (conformal prediction), shap 0.44+, pytorch 2.1+.
Before using code patterns, verify installed versions match. If versions differ:
- Python:
pip show <package>to check signatures - CLI:
chemprop train --help
If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.
QSAR Modeling
Hard rules
- No fabricated citations. Every cited work must resolve to a verifiable
- No claim without provenance. Every quantitative or factual claim
- No silent failure. Every script invocation, API call, or tool use must declare its exit status and what to do on non-zero. A skill that silently swallows errors is a violation.
When to use
Load this skill when the user asks a question that matches its declared
trigger conditions (see the frontmatter description for the most common
ones). Do not load it for questions outside its scope — defer to the
appropriate ORS skill instead.
When NOT to use
Do not load this skill if the question is in a sibling skill's domain
(see ## Cross-references), if the user explicitly asks for a different
tool, or if the task is outside the skill's declared category.
Model Taxonomy
| Model | Architecture | Use case | Fails when |
|---|---|---|---|
| Random Forest + ECFP4 | Classical baseline | Small data (<200 compounds) | Saturates at ~AUC 0.85 |
| chemprop D-MPNN | Directed message passing | Modern default; 100-10k compounds | Very small datasets (<100) |
| MolFormer | Transformer (87M params) | Large public data | Compute overhead |
| Uni-Mol | 3D-aware transformer | 3D-relevant endpoints | Requires 3D conformers |
| Gaussian Process + ECFP4 | Probabilistic | Active learning | O(N^3) scaling |
Decision: For 200-10k compounds, chemprop 2.0 D-MPNN is the modern standard. For <200 compounds, Random Forest + ECFP4 is competitive.
chemprop 2.0 Training (CLI)
Goal: Train a chemprop D-MPNN ensemble with scaffold-balanced split.
chemprop train \
--data-path data.csv \
--task-type classification \
--save-dir model_dir \
--molecule-featurizers rdkit_2d_normalized \
--num-folds 5 \
--ensemble-size 5 \
--epochs 50 \
--batch-size 128 \
--split scaffold_balanced \
--split-sizes 0.8 0.1 0.1
OECD 5 Principles
- Defined endpoint: specific bioassay, units, threshold definitions
- Unambiguous algorithm: reproducible code, fixed seeds
- Defined applicability domain (AD): where the model is valid
- Appropriate statistical validation: external test set, cross-validation
- Mechanistic interpretation: biological/chemical rationale
Applicability Domain Methods
| Method | Definition | Pro |
|---|---|---|
| Ensemble variance | Std across N-model predictions | Built-in to chemprop |
| kNN distance | Mean Tanimoto to k nearest in training | Easy to interpret |
| Leverage | Hat matrix diagonal | Statistical |
Operational rule: Set --ensemble-size 5 at training; at predict time, flag predictions with ensemble std > P95 as out-of-AD.
References
- Yang et al., J. Chem. Inf. Model. 59:3370 -- chemprop scaffold split.
- OECD -- QSAR validation principles.
Related Skills
- chemoinformatics/molecular-descriptors - Features
- chemoinformatics/molecular-standardization - Critical upstream
- chemoinformatics/admet-prediction - ADMET QSAR"
Cross-references
Other skills in this category:
- admet-prediction
- conformer-generation
- covalent-design
- docking-rescoring
- free-energy-calculations
- generative-design
- molecular-descriptors
- molecular-io
- molecular-standardization
- pharmacophore-modeling
- pose-validation
- protac-degraders
- retrosynthesis
- scaffold-analysis
- shape-similarity
- similarity-searching
- substructure-search
- virtual-screening
Changelog
- 1.1.0 (migration) — Bulk-migrated to v0.4.0 schema: canonical metadata block, base Hard rules, Cross-references. Body content unchanged; author should review and fill in any domain-specific extensions to the Hard rules.
- 1.0.0 — Initial release.