RESEARCH USE
Where Chemprop fits
Chemprop trains message-passing neural networks on labelled molecular or reaction data for classification, regression, multitask learning, and related prediction workflows. Define data cleaning and splitting before training, especially when related chemical scaffolds could leak across partitions. Record the major Chemprop version, features, hyperparameters, seeds, and evaluation metrics.
Research tasks
- Train molecular or reaction property models
- Compare data splits, ensembles, and hyperparameter settings
- Produce predictions, uncertainty estimates, or learned representations
What to evaluate before use
- Performance depends on training-domain coverage, label quality, and the split design; out-of-domain chemistry requires separate assessment.
- Chemprop 1 and 2 differ materially in interfaces and defaults. Reproducible work should state the major version and follow the matching documentation.
Verification note
This entry summarizes the tool's role without assessing scientific accuracy or endorsing its outputs. Features and terms can change; consult the official source before adopting it for consequential work.
molecular propertiesmessage passing
Last verified: 2026-09-04
Source: official documentation ↗