Cns Drug Enrichment Eval

cns-drug-enrichment-eval

Uni-QSAR: an Auto-ML Tool for Molecular Property Prediction — Gao et al. (2023) (arXiv:2304.12239, 2023)

What this evaluates

Tests the model's ability to classify CNS-active versus CNS-inactive drugs and enrich active compounds from large virtual screening databases. It probes generalization on small-sample molecular datasets using external validation.

Datasets

• CNS Drug Dataset — total 1057; splits: train (940), test (117)

Metrics

• AUC (primary) — range: [0, 1]
  • Area Under the ROC Curve for binary classification of CNS-active vs inactive drugs. Higher is better.

Input / output format

Input: Molecular fingerprints-based bit vectors and binary labels (CNS-active or CNS-inactive).

Output: Binary classification probability or label.

Scoring recipe

def score_auc(y_pred, y_true):
    auc = roc_auc_score(y_true, y_pred)
    return auc

Common pitfalls

• Relying solely on random splits without an external test set to assess generalization on unseen market drugs.
• Using traditional QSAR descriptors without leveraging 3D molecular representation learning for small-sample generalization.
• Skipping the initial fingerprint-based similarity search and rule-based filtering (Lipinski/Veber) before fine-grained screening.

Evidence (verbatim from paper)

We build CNS models using Uni-QSAR with 940 market drugs [72] (315 CNS-active and 625 CNS-inactive) as training data and evaluate models' performance using an additional external dataset of 117 market drugs [74] as testing data. The result in Fig. 5 shows that Uni-QSAR methods are more effective for the classification of CNS drugs versus non-CNS drugs. ... Validation AUC ... Testing AUC ...

Citation

@misc{gao2023uniqsar,
  title={Uni-QSAR: an Auto-ML Tool for Molecular Property Prediction},
  author={Gao et al. (2023)},
  year={2023},
  note={arXiv:2304.12239}
}

• arXiv: 2304.12239