Based On Sequence Features Combined With The Nuclear Non-linear Regression To Predict Protein Folding Rate

Protein is a kind of important biological macromolecules and the main undertaker of life activity occupying a special place in organisms.It is also a basic organism that constitutes a cell.Proteins can be used to assemble themselves in a specific PH,temperature,and so on,the process of obtaining its functional structure and conformation is called protein folding.Protein folding rate is an important parameter to measure protein folding speed.Understanding protein folding rate has profound significance for exploring protein folding mechanism.In recent years,many researchers have given the way to predict protein folding rate and many different characteristic parameters based on these methods.This paper forecast the rate of protein folding based on sequence characteristics of combination and nuclear nonlinear regression.The main work includes the following two aspects:1.Select nine kinds of physical and chemical properties of amino acids based on the chemical and physical properties of amino acids and the characteristics of amino acid sequence structure in the role of different categories of proteins,and combining the LZ_c complexity of eigenvalues of the amino acid sequence with them to characterize proteins.Multidimensional eigenvector of protein is used to establish multivariate nonlinear regression model that are used for calculating the prediction value of folding rate of 83 proteins.We analyze the correlation between eigenvalues and corresponding folding rate among different combinations using Jack-knife testing method.The experimental results show that the prediction accuracy and the feasibility of multivariate nonlinear regression model are higher than linear regression model and multivariate nonlinear regression model has low computing complexity and convenience for operating.2.Because of those parameters'effects on the protein structure and function including the consensus hydrophobic of amino acids,the trend of the alpha helix,a power supply in alpha helix N terminal,the probability of backbone dihedral Angle,parameters of the tendency coefficient of metal combination,the average distance contact and the sequence complexity,we use the selection and fusion of these characteristic parameters to characterize proteins and get protein multidimensional eigenvector.Establish multivariate nonlinear regression model by experimental value of protein folding rate and the characteristics of the protein vector to calculate rate of protein folding prediction of the 29 two-state proteins and35 multi-state proteins.Checked by jack-knife,the folding rate of two-state protein and multi-state protein has a good correlation between the characteristic parameters.It is verified that these eigenvalues affect the folding rate of proteins with different folding types.