Home >

news Help

Publication Information


Title
Japanese: 
English:Cryptic site detection using machine learning based on mixed-solvent molecular dynamics simulations results 
Author
Japanese: 本野 千恵, 小関 準, 柳澤 渓甫, 工藤 玄己, 吉野 龍ノ介, 広川 貴次, 今井 賢一郎.  
English: Chie Motono, Jun Koseki, Keisuke Yanagisawa, Genki Kudo, Ryunosuke Yoshino, Takatsugu Hirokawa, Kenichiro Imai.  
Language English 
Journal/Book name
Japanese: 
English: 
Volume, Number, Page        
Published date Oct. 22, 2024 
Publisher
Japanese: 
English: 
Conference name
Japanese: 
English:Asia & Pacific Bioinformatics Joint Conference 2024 
Conference site
Japanese:沖縄 
English:Okinawa 
Abstract Cryptic sites are not detectable in static protein structures however emerge by conformational change based on structural fluctuations and/or interaction with a small compound. Binding small compounds to the sites could control protein function allostery. Thus detection of such sites leads to the expansion of drug discovery for undruggable targets. The mixed-solvent molecular dynamics (MSMD) relying on chemical fragments mixed with water molecules can search for cryptic sites based on the accumulation of fragments (hotspots) on the protein surface. It is, however, still difficult to predict whether a hotspot matches cryptic site or not. We developed a support vector machine (SVM)-based method to discriminate hotspots mapping cryptic sites using the features obtained from MSMD simulations using six probes with various chemical properties (benzene, isopropanol, phenol, imidazole, acetonitrile, and ethylene glycol); those features are probe occupancy, physico-chemical and dynamical characteristics of the protein surfaces around the hotspots. Using 61 hotspots matching cryptic sites and 125 hotspots obtained from proteins having no cryptic sites, 5-fold cross-validation of our SVM models yielded an accuracy ~0.80. Our approach makes it possible to identify cryptic sites more effectively by a combination of hotspot detected based on multiple chemical fragments and protein dynamic features.

©2007 Tokyo Institute of Technology All rights reserved.