Several Predictions for Protein Secondary Structure by Multiple-Trained Hidden Markov Models

Predicting the secondary structure of a protein is important in many fields of medicine and bioinformatics. Many researchers have tried to obtain the highest possible prediction accuracy to reduce the cost, time, and effort in the prediction process instead of using special techniques such as X-ray or magnetic resonance. This paper presents a proposed method for several predictions of protein secondary structure using the Hidden Markov Model (HMM) multiple training and testing. We used a data set of more than 9,000 sequences from different families extracted by X-ray, ensuring the impartiality of the work. The data set was divided into two groups, the first for training and the second for testing. The data is initialized before it is entered into the program designed by matlab15b, where it is transformed from a set of sequences into matrices of different sizes. The proposed method can train several sequences ranging from 5 to 100 sequences once and test each with 150 test sequences once. The advantage of this method is that it allows choosing the best training sequences from among several training sequences, obtaining an accuracy of 58%-78%, and reducing time and effort compared to the default methods.