Identifying the missing protein in human proteome by structure and function prediction

After the completion of human genome project, the proteome research becomes one of the center problems in post-genomics era. The Human Protein Project aims to identify at least one protein product from each of the human protein-coding genes by using experiment methods. However there are still many proteins without experimental evidence which become one of the major challenges in chromosome-centric human proteome project. Taking into consideration of the complexity of detecting these missing proteins by using proteomics approach, here we provide the structure and function of these missing proteins by bioinformatics methods. 616 “uncertain” missing proteins are extracted from the neXtProt database and the structure and function of these missing proteins are predicted by using state-of-the-art software I-TASSER and COFACTOR respectively. A comprehensive evaluation shows that the results are in good consistent with many manually curated annotations from well-established databases and other mass spectrum datasets. There are 188 foldable proteins (I-TASSER C-Score larger than -1.5) without using any homologous template, which may be native gene-coding proteins. The Gene Ontology function prediction results are in good agreement with the manual annotation from neXtProt database, and also the confidence scores are well correlated with the evaluation metrics with Pearson correlation coefficient of 0.65. The data are deposited into Human Proteome Structure and Function database (HPSF) which can provide valuable references about the missing proteins. The HPSF database is publicly available at http://zhanglab.ccmb.med.umich.edu/HPSF/.