Protein submitochondrial localization from integrated sequence representation and SVM-based backward feature extraction.

Journal: Molecular bioSystems
Published Date:

Abstract

Mitochondrion, a tiny energy factory, plays an important role in various biological processes of most eukaryotic cells. Mitochondrial defection is associated with a series of human diseases. Knowledge of the submitochondrial locations of proteins can help to reveal the biological functions of novel proteins, and understand the mechanisms underlying various biological processes occurring in the mitochondrion. However, experimental methods to determine protein submitochondrial locations are costly and time consuming. Thus it is essential to develop a fast and reliable computational method to predict protein submitochondrial locations. Here, we proposed a support vector machine (SVM) based approach for predicting protein submitochondrial locations. Information from the position-specific score matrix (PSSM), gene ontology (GO) and the protein feature (PROFEAT) was integrated into the principal features of this model. Then a recursive feature selection scheme was employed to select the optimal features. Finally, an SVM module was used to predict protein submitochondrial locations based on the optimal features. Through the jackknife cross-validation test, our method achieved an accuracy of 99.37% on benchmark dataset M317, and 100% on the other two datasets, M1105 and T86. These results indicate that our method is economic and effective for accurate prediction of the protein submitochondrial location.

Authors

  • Liqi Li
    Department of General Surgery, Xinqiao Hospital, Third Military Medical University, Chongqing 400037, China. yanghuaxq@163.com.
  • Sanjiu Yu
  • Weidong Xiao
  • Yongsheng Li
    School of Interdisciplinary Medicine and Engineering, Harbin Medical University, Harbin, 150081, China. liyongsheng@ems.hrbmu.edu.cn.
  • Wenjuan Hu
  • Lan Huang
  • Xiaoqi Zheng
  • Shiwen Zhou
  • Hua Yang