5,795 Molecules with Acetylcholinesterase Inhibition Data for Predictive Modeling

5,795 small molecules were compiled from BindingDB to support machine learning models predicting acetylcholinesterase (AChE) inhibitory activity. The dataset was curated by R.U. Laskar and published on figshare in April 2026. It includes multiple molecular representations, such as physicochemical descriptors and graph-based structures, used to evaluate fifteen predictive models.

Use Cases

• Training hybrid machine learning models for AChE inhibition prediction based on physicochemical descriptors.
• Benchmarking graph neural networks against traditional models based on graph-based molecular structures.
• Developing late-fusion predictive frameworks based on the integration of descriptor-based and graph-based representations.

Strengths

• A curated dataset of 5,795 molecules provides a substantial foundation for model development.
• The study evaluated fifteen predictive models, including twelve individual and three hybrid models, providing a performance benchmark.
• The best-performing hybrid model achieved an r² of 0.7400 ± 0.0138 across ten independent splits, indicating stable performance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is small in scale at 412.2 KB, limiting the scope of analysis.

Provenance

Source

BindingDB

Collection Method

Curated compilation of experimentally measured bioactivity data for small molecules targeting human AChE.

Time Range

null

Freshness

Last updated 2026-04-15 09:41:37; freshness should be verified.

Geography

null