Method For Predicting The Amphipathic Helix-PAHT

Amphipathic helices are a protein secondary structure in which one face contains hydrophobic residues and the other contains hydrophilic ones.Amphipathic helices were first found in apolipoproteins and then in many transmembrane proteins.They play an importance role in membrane remodeling and vesicular transport.Amphipathic helices in a transmembrane protein are often parallel to the membrane and,so their physical properties must be hydrophilic on one surface and hydrophobic on the other.Amphipathic helices are commonly found in transmembrane proteins and an amphipathic helix is separated into a face with hydrophilic residues and another face with hydrophobic residues.Amphipathic helices are involved in many functions,such as sensing membrane curvature and interacting with other membrane proteins.Because of the importance of the functions of amphipathic helices,the prediction of amphipathic helices from protein sequences has become increasingly important.Many methods have been proposed for solving this problem,such as Helical Wheel,Hydrophobic Moment,Heliquest,AmphipaSeek.Helical Wheel and Hydrophobic Moment provide scoring functions for evaluating candidate amphipathic helices in protein sequences,and these functions are based on projection and calculation of protein sequence.But they lack a function for predicting amphipathic helix regions from protein sequences.Heliquest calculates the physicochemical properties and amino acid composition of an alpha helix and screens databank to identify protein segments possessing similar features.AmphipaSeek predict amphipathic helices by a pattern recognition SVM with a dedicated kernel.But they failed to achieve high accuracy in prediction due to the complexity of the problem.We propose a new method for predicting amphipathic helices from protein sequences.We project the three-dimensional structure of a candidate amphipathic helix onto a plane,and the HNP model is established to describe the spatial structure of the protein alpha helix.On the basis of the HNP model,the HNP parameters are abstracted to describe the amphipathic feature.Then calculate the 9-dimensional features of each protein subsequence,and finally input the features into the trained support vector machine(SVM)model to predict whether this protein region is an amphipathic helix.We trained and tested the model using five protein sequences with annotated amphipathic helices.Experimental results showed that the model achieved high accuracy for amphipathic helix predication.In this paper,we present PAHT,a new machine learning method for amphipathic helix prediction.PAHT takes whole protein sequences and the secondary structure information of the protein as input and reports amphipathic helix and non-amphipathic helix,and non-helix regions in the sequences.We tested the performance of PAHT on five protein sequences and demonstrated that it achieved higher accuracy than other existing tools.We implemented PAHT into a usable website through SpringBoot and Vue,in which the Java program calls the SVM model through RPC,and finally we display the prediction results of the protein amphipathic helix.