Based On Feature Fusion Protein Properties Prediction Of Multiple Points Of View

As an important material basis, protein constitute the organisms. The phenomenon of life is mainly realized through the structure and function of protein. It’s very importance for us to grasp the various attributes of proteins so that we can comprehend the functions of protein, understand various kinds of biochemical reactions, gene expression within the organisms and drug development.With the development of the Human Genome Project, the total amount of DNA sequences in the latest GenBank database published by the U.S so far has exceeded126.5billion base pairs, synchronously increased as the number of protein sequences. Life Sciences has stepped into the post-genome era substantially.Due to its weakness of high complexity, long experimental period and low efficiency, traditional biochemical experiments can not meet the urgent demand of which massive protein sequences presents. Therefor, protein attribute prediction from sequences becomes an important task and how to extract discriminative features is one of the most crucial aspects. Previous studies have shown that single-view feature can not reflect all the information of a protein, fusing multi-view features is considered as a promising route to improve prediction accuracy. In this paper, we propose a novel parallel framework for protein multi-view feature fusion:First, features from different views are parallely combined to form complex feature vectors; Then, we use the generalized principle component analysis (GPCA) for further feature extraction from the parallelly combined complex features, which lie in a complex space; Finally, the extracted features are used for prediction. Experimental results on different benchmark datasets and machine learning algorithms verify the effectiveness of the method proposed in protein attribute prediction.