Protein functional annotation of simultaneously improved stability, accuracy and false discovery rate achieved by a sequence-based deep learning

Abstract

Functional annotation of protein sequence with high accuracy has become one of the most important issues in modern biomedical studies, and computational approaches of significantly accelerated analysis process and enhanced accuracy are greatly desired. Although a variety of methods have been developed to elevate protein annotation accuracy, their ability in controlling false annotation rates remains either limited or not systematically evaluated. In this study, a protein encoding strategy, together with a deep learning algorithm, was proposed to control the false discovery rate in protein function annotation, and its performances were systematically compared with that of the traditional similarity-based and de novo approaches. Based on a comprehensive assessment from multiple perspectives, the proposed strategy and algorithm were found to perform better in both prediction stability and annotation accuracy compared with other de novo methods. Moreover, an in-depth assessment revealed that it possessed an improved capacity of controlling the false discovery rate compared with traditional methods. All in all, this study not only provided a comprehensive analysis on the performances of the newly proposed strategy but also provided a tool for the researcher in the fields of protein function annotation.

Keywords: annotation accuracy; deep learning; false discovery rate; prediction stability; protein function prediction.

Figure 1

The workflow of the deep learning algorithm (CNN) applied together with the sequence encoding technique proposed in this study.

Figure 2

Variations (the actual value of each method subtracted by the minimum value among four different methods) of four measurements among different protein function annotation methods: (A) MCC, (B) AC, (C) SE, (D) SP. On the right side, the statistical differences between any two methods were provided as the violin box plots. ^* indicated great difference of P < 0.05, and ^** denoted significant difference of P < 0.01. Detailed P-values are provided in Supplementary Table S1.

Figure 3

The EFs of different protein function annotation methods of the training and testing data sets (for all studied GO families) with the lowest similarity. On the right side, the statistical differences of the EFs between any two methods were provided as the violin box plots. ^* indicated great difference of P < 0.05, and ^** denoted significant difference of P < 0.01. Detailed P-values are provided in the Supplementary Table S2.