Technical Report 薬剤活性予測の改良のための化合物フィンガープリントの比較解析
Comparative analysis of molecular fingerpirnts toward improvement of drug activity prediction

松山, 祐輔  ,  石田, 貴士

2017-BIO-49 ( 5 )  , pp.1 - 7 , 2017-03-16
Selection of feature vector used for machine learning drug activity prediction is an important problem. A number of molecular fingerprints have been proposed until now, there is no fingerprint that provides the best performance in any target protein, and any task. In this study, we conducted comparative analysis on drug activity prediction performance using many molecular fingerprints, and analyzed the similarity between molecular fingerprints. We also analyzed the change in prediction performance when using multiple fingerprints in detail. As a result, it is considered that their molecular fingerprints pick up dome important features of compounds separately to some extent, and that using multiple molecular fingerprints would contribute to performance improvement.

Other information