Systematic Functional Discovery Of Human Uncharacterized Proteins Through Multiple Resource Integration With Bayesian Approach

With the rapid development of high-throughput sequencing and mass spectrometry technologies,more and more previously defined missing proteins have been identified in human body.However,in contrast to the ability of large amount of omics data generation,the studies in characterizing protein functions are relatively slow.So far,about 3,000 human protein coding genes has not been characterized.Completely functional characterization of proteins is an urgent need,and the basis for undering mechanism studies of various biological processes and pathological conditions.Establishing the relationship between the characterized and uncharacterized proteins is the theoretical basis for computationally predicting the function of uncharacterized proteins.In this study,we firstly found some characteristics of uncharacterized proteins,including sparse interactive proteins,low evolutionary conservation,and variable conformation.Secondly,phyletic evolution profile,protein-protein interaction profile and gene expression profile were used to calculate the association strength among the proteins based on clustering analysis,thereby estimated the probability of the associations.In addition to the three strategies,we also built a Bayesian network model by integrating predictive scores from the three profiles to comprehensively predict the protein functions.To evaluate the performance of our proposed strategies for protein function characterization,we predicted the function of about 2,000 well-characterized proteins based on evolutionary relationship,interactive relationship,co-expression relationship,and the integrative approach.The three single-profile based strategy achieved 40%,45%and 32%accuracy.By integrating the results of those three different resources,our Bayesian network model successfully achieved 62%accuracy,this predictive accuracy is an ideal predictor for refractory proteins without homology sequence.Finally,we predicted the protein complexes and pathways that the uncharacterized proteins involved in based on protein association network,and found that some of these proteins could involve in specific protein complexes and disease related signaling pathways.Our results revealed that the uncharacterized protein could also play important roles in multiple biological processes.