Probing the Missing Human Proteome: A Computational Perspective

The missing human proteome comprises predicted protein-coding genes with no credible protein level evidence detected so far and constitutes ∼18% of the human protein coding genes (neXtProt release 19/9/2014). The missing proteins may be of pharmacological interest as many of these are membrane receptors, thus requiring comprehensive characterization. In the present study, we explored various computational parameters, crucial during protein searches from tandem mass spectrometry (MS) data, for their impact on missing protein identification. Variables taken into consideration are differences in search database composition, shared peptides, semitryptic searches, post-translational modifications (PTMs), and transcriptome guided proteogenomic searches. We used a multialgorithmic approach for protein detection from publicly available mass spectra from recent studies covering diverse human tissues and cell types. Using the aforementioned approaches, we successfully detected 24 missing proteins (22-PE2, 1-PE4, and 1-PE5). Maximum of these identifications could be attributed to differences in reference proteome databases, exemplifying use of a single standard database for human protein detection from MS data. Our results suggest that search strategies with modified parameters can be rewarding alternatives for extensive profiling of missing proteins. We conclude that using complementary spectral data searches incorporating different parameters like PTMs, against a comprehensive and compact search database, might lead to discoveries of the proteins attributed so far as the missing human proteome.