Prediction of success for polymerase chain reactions using the Markov maximal order model and support vector machine

• Published in: Journal of theoretical biology Vol. 369; pp. 51 - 58
• Main Authors: Li, Chun, Yang, Yan, Fei, Wenchao, He, Ping-an, Yu, Xiaoqing, Zhang, Defu, Yi, Shumin, Li, Xuepeng, Zhu, Jin, Wang, Changzhong, Wang, Zhifu
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 21.03.2015
• Subjects: Amplicon; DNA template; Humans; Markov Chains; Models, Theoretical; Polymerase Chain Reaction - methods; Polymerase chain reactions; Support Vector Machine; Support vector machines; Support vector machines; Polymerase chain reactions; Markov chains; Amplicon; DNA template
• ISSN: 0022-5193; 1095-8541; 1095-8541
• DOI: 10.1016/j.jtbi.2015.01.017

Polymerase chain reaction (PCR) is hailed as one of the monumental scientific techniques of the twentieth century, and has become a common and often indispensable technique in many areas. However, researchers still frequently find some DNA templates very hard to amplify with PCR, although many kinds of endeavors were introduced to optimize the amplification. In fact, during the past decades, the experimental procedure of PCR was always the focus of attention, while the analysis of a DNA template, the PCR experimental subject itself, was almost neglected. Up to now, nobody can certainly identify whether a fragment of DNA can be simply amplified using conventional Taq DNA polymerase-based PCR protocol. Characterizing a DNA template and then developing a reliable and efficient method to predict the success of PCR reactions is thus urgently needed. In this study, by means of the Markov maximal order model, we construct a 48-D feature vector to represent a DNA template. Support vector machine (SVM) is then employed to help evaluate PCR result. To examine the anticipated success rates of our predictor, jackknife cross-validation test is adopted. The overall accuracy of our approach arrives at 93.12%, with the sensitivity, specificity, and MCC of 94.68%, 91.58%, and 0.863%, respectively. [Display omitted] •Focus on a DNA template, the PCR experimental subject itself.•Estimate the number of occurrence for a k-word in a binary sequence by means of the Markov maximal order model.•The feature vector reflects information on both the expected probabilities of k-words and the order relation among them.•Develop a tool to predict whether a PCR is successful or not before a real PCR experiment is done.