An Optimized Prediction Model Based on Feature Probability for Functional Identification of Large-Scale Ubiquitous Data

Recently, there is a growing interest in the sequence analysis. In particular, the next generation sequencing (NGS) technique fragments the base sequence and analyzes the functions thereof. Its essential role is to arrange pieces of the base sequence together based on sequencing and to define the functions. The organization of unarranged piece of sequence is one of the active research areas; moreover, definition of gene function automatically is a popular research topic. The previous studies about the automatic gene function have mainly utilized the method that automatically defines protein functions by using the similarities of base sequence or the disclosed database and the protein interaction or context free method. This study aims to predict the category of protein whose function was not defined after learning automatically with GO by extracting the characteristics of protein inside the cluster. This study conducts clustering by using the protein interaction that is generated by the similarities of base sequence under the assumption that the proteins inside the cluster have similar function. The proposed method is to show an optimized result in accordance with the option after finding the option value that can give the outperformed prediction of GO, which classifies the functions based on the IPR and keywords inside the same cluster as the unique features.