A New Combination Strategy as Applied in Predicting Chromatographic Retention Times of Oligonucleotides at a Range of Temperatures from 30 °C to 80 °C

Abstract

We develop a new combination strategy to predict retention times of oligonucleotides in ion-pair reversed-phase high performance liquid chromatography. The key step of the strategy is to use score of generalized base properties (SGBP) combined with auto cross covariance (ACC) to resolve the feature representation of nucleic acids. This representation characterize physicochemical, quantumchemical, topological, spatial structural features, etc., and their neighboring effect between bases at a certain distance apart in a sequence. The next step is to use the variables selected by genetic algorithm (GA) to construct prediction models of retention times of oligonucleotides based on support vector machine (SVM). Accordingly, GA-SVM models give different prediction performance using different input descriptors resulting from different step lengths in the ACC transformation, indicating the neighboring effect between bases should not be neglected in the features related to the chromatographic retention of oligonucleotides. As a whole, the GA-SVM predictors obtained from more than 20 training samples can produce satisfying performance in predicting the chromatographic retention of oligonucleotides at a range of temperatures (30°C, 40°C, 50°C, 60°C and 80°C), respectively. The present approach based on the SGBP-ACC-GA-SVM combination shows great application prospect in the field of separation and analysis science, bioinformatics and proteomics.