A ranking-based method to predict protein function in terms of gene ontology

In the post genomic era, protein function prediction has become the focus of many researches. Methods that predict function from the high-throughput experimental screening have gained popularity due to the reduced cost of conducting genome-wide functional screening. In this work we predict the function of uncharacterized proteins by ranking the functions based on the protein interaction data with the help of gene ontology terms. We used Neighbor counting, Chi-square, S-weight, and FS-weight methods to predict the functions of a set of 100 un-annotated proteins. Among them, the ZNF24 and VDP proteins, which were predicted to be involved in transcription regulation and vesicle-mediated transport, respectively, obtained the best results. In confirmation, our results were mostly confirmed by Pfam and WoLF Psort shows possible reliability of our method, and it therefore can be applied in protein function prediction when the prior knowledge such as sequences or structures is still not available.