Systematic Analysis of Single- and Multi-Target Compounds for Machine Learning

Two balanced datasets contain 15,142 multi-target and 15,081 single-target compounds, plus a subset of 1,828 diverse-target and 1,776 single-target compounds. The data was compiled by Christian Feldmann at the University of Bonn for a study published in Molecular Pharmaceutics. Each compound includes a SMILES representation, ChEMBL ID, UniProt target IDs, and a category label.

Use Cases

• Train classification models to distinguish single-target from multi-target compounds based on SMILES representations.
• Analyze structural patterns associated with diverse-target activity using the provided compound subset.
• Benchmark feature selection or data reduction techniques using the 'Y' tag indicating compound retention after random or nearest-neighbor removal.

Strengths

• Contains over 30,000 total compounds across two balanced datasets.
• Provides explicit compound counts: 15,142 MT-CPDs, 15,081 ST-CPDs, 1,828 DT-CPDs, and 1,776 ST-CPDs.
• Includes key identifiers like ChEMBL IDs and UniProt target IDs for external database linkage.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Row count for individual files is unknown, which may limit suitability assessment.

Provenance

Source

University of Bonn

Collection Method

Compiled for a machine learning study published in Molecular Pharmaceutics.