Drug side effect prediction based on the machine learning of small molecule-protein interaction profiles

2014 Fiscal Year Final Research Report

• Project/Area Number: 25730025
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Statistical science
• Research Institution: The Institute of Physical and Chemical Research
• Principal Investigator: SATO Tomohiro 独立行政法人理化学研究所, ライフサイエンス技術基盤研究センター, 研究員 (00595358)
• Project Period (FY): 2013-04-01 – 2015-03-31
• Keywords: 機械学習 / 予測モデル / ポリファーマコロジー / 副作用予測 / 構造記述子

Outline of Final Research Achievements

In this study, a novel method to predict drug adverse reactions (ADRs) based on machine leaning of small molecule-protein interaction profile was developed. At first, interactions between a compound and 329 proteins were predicted using molecular fingerprints. Then, the random forests models to predict 129 ADRs registered in SIDER2 drug side effect database were built based on the 329-dimensional interaction profile. Leave-cluster-out validation showed that the proposed method could maintain higher accuracy for compounds with low structural similarity to training data than the conventional prediction models directly using molecular fingerprint.

Free Research Field

創薬分子設計