Search

Membrane proteins are encoded by \sim 30\% in the genome and function importantly in the living organisms. Previous studies have revealed that membrane proteins' structures and functions show obvious cell organelle-specific properties. Hence, it is highly desired to predict membrane protein's subcellular location from the primary sequence considering the extreme difficulties of membrane protein wet-lab studies. Although many models have been developed for predicting protein subcellular locations, only a few are specific to membrane proteins. Existing prediction approaches were constructed based on statistical machine learning algorithms with serial combination of multi-view features, i.e., different feature vectors are simply serially combined to form a super feature vector. However, such simple combination of features will simultaneously increase the information redundancy that could, in turn, deteriorate the final prediction accuracy. That's why it was often found that prediction success rates in the serial super space were even lower than those in a single-view space. The purpose of this paper is investigation of a proper method for fusing multiple multi-view protein sequential features for subcellular location predictions. Instead of serial strategy, we propose a novel parallel framework for fusing multiple membrane protein multi-view attributes that will represent protein samples in complex spaces. We also proposed generalized principle component analysis (GPCA) for feature reduction purpose in the complex geometry. All the experimental results through different machine learning algorithms on benchmark membrane protein subcellular localization datasets demonstrate that the newly proposed parallel strategy outperforms the traditional serial approach. We also demonstrate the efficacy of the parallel strategy on a soluble protein subcellular localization dataset indicating the parallel technique is flexible to suite for other computational biology problems. The software and datasets are available at: http://www.csbio.sjtu.edu.cn/bioinf/mpsp. [ABSTRACT FROM PUBLISHER]

Transmembrane helices (TMH) identification is one of the most important steps in membrane protein structure prediction. Existing TMH predictors tend to pursue accurate computational models without carefully considering the interpretability of these models and thus act as a black box. In this paper, a novel TMH predictor called SOMRuler with excellent interpretability while possessing high prediction accuracy is presented. The SOMRuler uses a self-organizing map (SOM) to learn helices distribution knowledge, which is encoded in the codebook vectors of the trained SOM, from the training samples. Human interpretable fuzzy rules are then extracted from the codebook vectors of the trained SOM. By extracting fuzzy rules from the learned knowledge rather than the original training samples, on the one hand, the computational burden of extracting fuzzy rules can be greatly reduced; on the other hand, the reliability of the extracted rules can also be enhanced since noise contained in the original samples can be smoothened by the learning procedure of SOM. The validity of the fuzzy rules extracted by SOMRuler is qualitatively and quantitatively analyzed. Experimental results on the benchmark dataset show that the SOMRuler outperforms most existing popular TMH predictors and is flexible to suite for a wide variety of problems in bioinformatics. The SOMRuler software is implemented by Java and Matlab and is available for academic use at: http://www.csbio.sjtu.edu.cn/bioinf/SOMRuler/. [ABSTRACT FROM PUBLISHER]