Protein Structure Prediction Using Hybrid Neural Network and Fuzzy Inference System

This work presents a method based on an adaptive neuro-fuzzy inference system (ANFIS) for modeling protein secondary structure prediction which aims at acquiring the unknown structure information of target protein directly from its sequence data which is available. The number of input variables and inference rules are commonly too large, sometimes even huge, to make the model building feasible. To overcome these defects a two-phase process is employed in our model. In the first phase, the selection of number and position of the fuzzy sets of initial input variables can be determined by employing a fuzzy clustering algorithm; and in the second phase the more precise structural identification and optimal parameters of the rule-base of the ANFIS are achieved by an iterative GA updating algorithm. An experiment on three-state secondary structure prediction of protein is reported briefly and the performance of the proposed method is evaluated. The results indicate an improvement in design cycle and convergence to the optimal rule-base within a relatively short period of time, however, at the cost of little decrease in accuracy. keyword: bioinformatics, protein structure prediction, ANFIS