作者: Jed Zaretzki , Kevin M. Boehm , S. Joshua Swamidass
DOI: 10.1021/CI5005652
关键词:
摘要: Molecule and atom fingerprints, similar to path-based Daylight can substantially improve the accuracy of P450 site-of-metabolism prediction models. Only two chemical fingerprints have been used in metabolism prediction, so little is known about importance fingerprint parameters on site predictions. It possible that different might yield more accurate Here, we study if tuning specific data sets lead improved We measure impact 484 models nine isoform sets. Using a range search depths, path, circular, subgraph fingerprints. Two labelings, also, are considered, both standard SMILES labels also labeling marks ring bonds differently than nonring bonds, enabling ortho, para, meta positioning substituents be clearly encoded. Optimal ...