Online Service Of Functionally Similar Proteins Detection And Protein-protein Interaction Prediction

Protein-protein interactions are inner protein complex or just momentarily. The dynamicchanges of protein-protein interactions is very important for the molecular function andmetabolism of the organism. A lot of protein-protein interactions make up the protein-proteininteraction network. The prediction and analysis of protein-protein (network) count a lot forrevealing the mysteries of life.A pipeline was explored to detect functionally similar protein within species basing onprotein-protein interaction network. Here employed graph theory as computing model, a kindof modified shortest path as algorithm, protein-protein interaction network and proteinsequence as input, and a lot of functionally similar proteins are extracted. Compared withtraditionally functionally similar proteins detection method, sequence alignment, we foundthat a lot proteins which are functionally similar but with low sequence similarity are detected.We tested this method with six species, downloaded from two databases, and the resulted wasevaluated with GO TO. And we also found that there is no clear boundary between thehomologues and the convergent evolutions proteins. What different from functionally similarprotein across different species is that functionally similar protein within species are unlike tobe immunological repellent, so they would have great potential applications in the fields ofmedicine and industry.Protein-protein interaction network is made of protein-protein interactions. And theprotein-protein interaction prediction is widely used and has been a common technique whichis also a hot research field. And, now, there are a lot of theories are developed forprotein-protein interaction prediction. But every theory has its own emphasis, and the resultsof single theory would go with high false negative rate, so adoption of several methods wouldalways be the choice. But the currently available methods would always expecting complexdata as input, and it is a hard work to use them, what’s more, some computing models have nocodes or developed pipeline available. So protein-protein prediction software that take simpledata as input are urgently needed to be developed. Here a binding profile based method thatonly sequence information are needed was developed, and this is a great advancementtowards this target. Protein-protein (network) interaction prediction is computing density, a lot of methodshave no available software. Though some software running under command line aredeveloped, but this simple software always repel general users from a distance. But kindlydesigned software with beautiful user interface always could not carry so many computingworks. Here we developed an online platform which includes binding profile based,domain-motif interactions from structural topology based and interolog based protein-proteininteraction prediction, and we adopted a high performance server. As a result the userexperience and high computing density would not be a problem again.In the end, the proteome of wheat was used to evaluate the performance of this platform,and the two main function, protein-protein interaction (network) prediction andprotein-protein interaction network based functionally similar proteins detection withinspecious, could be used together.