ProLanGO: Protein Function Prediction Using Neural Machine Translation Based on a Recurrent Neural Network.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Next-generation sequencing rapidly generates vast protein sequence data.
• Traditional experimental methods are slow and costly for determining protein functions.
• Predicting protein function is crucial to bridge the gap between sequence data and biological understanding.

Purpose of the Study:

• To develop a novel method for protein function prediction.
• To address the challenge of annotating large-scale protein sequence data.
• To explore the application of machine translation in bioinformatics.

Main Methods:

• Proposed a new protein sequence language (ProLan) and a protein function language (GOLan).
• Developed a neural machine translation model using recurrent neural networks.
• Translated ProLan to GOLan for function prediction.
• Blindly tested the method in the Critical Assessment of Function Annotation (CAFA 3) competition.

Main Results:

• Achieved good performance on training and testing datasets.
• Demonstrated the method's effectiveness in the CAFA 3 competition.
• Validated performance on proteins with functions annotated post-competition.

Conclusions:

• The proposed language translation approach is a promising direction for protein function prediction.
• Neural machine translation effectively addresses the challenge of annotating protein sequences.
• This method offers a scalable solution for functional genomics.