The abstract is read at the publisher; the summary is JClub's.
MCDBench provides 792 medium-chain dehydrogenase enzyme–substrate–cofactor combinations, including standardized inactive cases, enabling systematic evaluation of catalytic activity prediction across 36 alcohol substrates and multiple cofactors.
- Why it matters: Catalytic performance prediction remains difficult because structure–function relationships are complex and existing benchmarks lack rigorously validated negative examples.
- What they did: The authors curated MCDBench and benchmarked four activity-prediction models (DLKcat, UniKP, TurNuP, CataPro) using standardized conditions to test generalization across enzyme classes and substrate types.
- The result: Model correlations with experimental activities are modest under standardized conditions, and separating active from inactive catalytic space remains challenging, likely shaped by positive-data skew and limited validated negatives.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.