MHNpath: Training Data and Models for Tunable Synthesis Planning

Shivesh Prakash released this 10.1 GB collection of training data, trained models, and associated files for the MHNpath retrosynthetic planning tool in April 2026. The data supports a machine learning framework that prioritizes reaction templates and allows user tuning based on cost, temperature, and toxicity. It includes case studies on complex molecules like dronabinol and benchmarks against established pathways from PaRoutes.

Use Cases

• Training retrosynthetic prediction models based on reaction template data.
• Benchmarking synthesis planning algorithms against the included 'gold-standard' experimental pathways.
• Designing cost-effective and greener synthetic routes based on the tunable scoring system for cost, temperature, and toxicity.
• Exploring novel enzymatic pathways for complex molecules as demonstrated in the case studies.

Strengths

• Dataset size is 10.1 GB, indicating substantial content.
• Includes trained models (PT files) and benchmark data against experimentally validated pathways.
• Released under a permissive CC-BY-4.0 license for open use and modification.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

figshare, associated with the paper 'A User-Tunable Machine Learning Framework for Step-Wise Synthesis Planning'.

Collection Method

Likely generated for training and evaluating the MHNpath machine learning tool.

Time Range

null

Freshness

Last updated 2026-04-20 06:05:20; freshness should be verified.

Geography

null