화학공학소재연구정보센터
학회 한국화학공학회
학술대회 2015년 봄 (04/22 ~ 04/24, 제주 ICC)
권호 21권 1호, p.141
발표분야 공정시스템
제목 Variable selection in peptide QSRR model development: Metaheuristic optimization algorithms approach
초록 Quantitative Structure-Retention Relationships (QSRRs) present a useful model in prediction of retention time in Reversed Phase – Liquid Chromatography (RP-LC). In this work, five metaheuristic nature-inspired algorithms (Genetic Algorithm, Particle Swarm Optimization, Artificial Bee Colony, Firefly Algorithm, and Flower Pollination Algorithm) were applied in variable selection prior to development of theoretical QSRR models for 83 peptides originating from eight model proteins. They were coupled with three component – Partial Least Squares (PLS), and compared to classical interval PLS (iPLS). Selected subset size was varied, and algorithm parameter optimization was carried out. Fitness function used was Root Mean Square Error of Prediction (RMSEP). It was shown that for a population size of 450, and a mutation rate of 0.2 GA-PLS is superior. It yielded a low RMSEP of 5.534%, and outperformed the rest in terms of computational cost. The best model was further validated with a Y-randomization procedure and external validation on 106 peptides originating from Bacillus subtilis proteome.
저자 Petar Žuvela1, 유 준1, Katarzyna Macur2, Tomasz Bączek2
소속 1부경대, 2Medical Univ. of Gdańsk
키워드 variable selection; QSRR; RP-LC; metaheuristic optimization
E-Mail
원문파일 초록 보기