With the development of next generation sequencing techniques, it is fast andcheap to determine protein sequences but relatively slow and expensive toextract useful information from protein sequences because of limitations oftraditional biological experimental techniques. Protein function prediction hasbeen a long standing challenge to fill the gap between the huge amount ofprotein sequences and the known function. In this paper, we propose a novelmethod to convert the protein function problem into a language translationproblem by the new proposed protein sequence language "ProLan" to the proteinfunction language "GOLan", and build a neural machine translation model basedon recurrent neural networks to translate "ProLan" language to "GOLan"language. We blindly tested our method by attending the latest third CriticalAssessment of Function Annotation (CAFA 3) in 2016, and also evaluate theperformance of our methods on selected proteins whose function was releasedafter CAFA competition. The good performance on the training and testingdatasets demonstrates that our new proposed method is a promising direction forprotein function prediction. In summary, we first time propose a method whichconverts the protein function prediction problem to a language translationproblem and applies a neural machine translation model for protein functionprediction.
展开▼
