DeepFunc: A Deep Learning Framework for Accurate Prediction of Protein Functions from Protein Sequences and...

Area of Science:

• Computational Biology
• Bioinformatics
• Machine Learning in Genomics

Background:

• Understanding molecular life relies on annotating protein functions.
• High-throughput sequencing generates vast protein data, but manual annotation remains limited (approx. 1%).
• Experimental function annotation is costly, time-consuming, and cannot match the pace of sequence data generation.

Purpose of the Study:

• To develop an accurate computational method for predicting protein functions.
• To overcome the bottleneck of manual protein function annotation.
• To leverage both sequence-derived and network-derived information for enhanced prediction.

Main Methods:

• A novel deep learning framework, DeepFunc, was developed.
• Protein sequence information (domains, families, motifs from InterPro) is encoded into a sparse binary vector.
• This vector is combined with topological information from protein-protein interactions (PPIs) and functional linkages, then processed by a deep neural network.

Main Results:

• DeepFunc demonstrated superior performance on a benchmark dataset compared to existing methods.
• On the Critical Assessment of protein Function Annotation algorithms (CAFA) 3 dataset, DeepFunc achieved the highest Fmax of 0.54 and AUC of 0.94.
• The framework effectively integrates sequence and network data for accurate protein function prediction.

Conclusions:

• DeepFunc offers a highly accurate and efficient computational solution for protein function prediction.
• The proposed deep learning framework significantly advances the field of bioinformatics and functional genomics.
• DeepFunc's performance highlights the potential of integrating diverse data types for biological discovery.