DeepGO Protein Function prediction.


DeepGO is a novel method for predicting protein functions using protein sequences and protein-protein interaction (PPI) networks. It uses deep neural networks to learn sequence features and PPI embeddings learned with knowledge graph representation learning method and hierarchically classifies proteins with GO classes.

The tool only needs protein sequence as an input. It takes network embeddings of the most similar sequence in our database found using BLAST search. The source code of the model is available on GitHub repository

The REST API is available for submitting sequences programmatically.


Examples:

>sp|P80386|AAKB1_RAT 5'-AMP-activated protein kinase subunit beta-1 OS=Rattus norvegicus GN=Prkab1 PE=1 SV=4 MGNTSSERAALERQAGHKTPRRDSSGGTKDGDRPKILMDSPEDADIFHTEEMKAPEKEEF LAWQHDLEVNEKAPAQARPTVFRWTGGGKEVYLSGSFNNWSKLPLTRSQNNFVAILDLPE GEHQYKFFVDGQWTHDPSEPIVTSQLGTVNNIIQVKKTDFEVFDALMVDSQKCSDVSELS SSPPGPYHQEPYISKPEERFKAPPILPPHLLQVILNKDTGISCDPALLPEPNHVMLNHLY ALSIKDGVMVLSATHRYKKKYVTTLLYKPI


MPYKLKKEKEPPKVAKCTAKPSSSGKDGGGENTEEAQPQPQPQPQPQAQSQPPSSNKRPSNSTPPPTQLSKIKYSGGPQIVKKERRQSSSRFNLSKNRELQKLPALKDSPTQEREELFIQKLRQCCVLFDFVSDPLSDLKFKEVKRAGLNEMVEYITHSRDVVTEAIYPEAVTMFSVNLFRTLPPSSNPTGAEFDPEEDEPTLEAAWPHLQLVYEFFLRFLESPDFQPNIAKKYIDQKFVLALLDLFDSEDPRERDFLKTILHRIYGKFLGLRAYIRRQINHIFYRFIYETEHHNGIAELLEILGSIINGFALPLKEEHKMFLIRVLLPLHKVKSLSVYHPQLAYCVVQFLEKESSLTEPVIVGLLKFWPKTHSPKEVMFLNELEEILDVIEPSEFSKVMEPLFRQLAKCVSSPHFQVAERALYYWNNEYIMSLISDNAARVLPIMFPALYRNSKSHWNKTIHGLIYNALKLFMEMNQKLFDDCTQQYKAEKQKGRFRMKEREEMWQKIEELARLNPQYPMFRAPPPLPPVYSMETETPTAEDIQLLKRTVETEAVQMLKDIKKEKVLLRRKSELPQDVYTIKALEAHKRAEEFLTASQEAL