Predicting Protein Folding Rates Based On Torsion Parameters

Protein folding is commonly defined as the process by which a protein’s structure assumes its functional shape or conformation, and protein folding rate is an important parameter to measure how fast protein folding. In the past decades, many methods have been put forward to predict the proteins folding rates. Generally speaking, these methods can be divided into three categories: model based on the topology of protein, model based on the chain length of protein, and model based on amino acid composition and its phychemical properties.In this paper, we put forward a new parameter, termed average contact dihedral angle number, to predict the folding rates. It indirectly reflects three aspects of information, namely topology, chain length and amino acid composition. Using a training dataset including 103 proteins, the correlation coefficient between the experimental and predicted folding rates with jackknife test is 0.762. And we also used a non-overlapping test dataset including 32 proteins, which are independent of our training dataset, to test the robustness. Comparison with five web-based server methods,the results indicates that the prediction accuracy and robustness of our model are better than these methods.In 2010,Luo and Zhang proposed the dynamical contact order, which is based on the moment of inertia and the torsion potential energy of the polypeptide chain between contact resides, to predict the folding rates of proteins. However, the calculation of moment of inertia is based on regarding the whole amino acid residue as a material point, which results in failing to meticulously reflect the differences of side chains for different amino acid residues. In fact, the sequence of side chains determines all that is unique about a particular protein, including its biological function and its specific three-dimensional structure. Especially, if the size of protein is approximatively equal, the specificity of protein side chains will directly reflect the difference in the folding rates. In order to reflect the difference of side chains, we redefined the contact is the non-hydrogen atoms between different amino acid residues. Meanwhile, we made a new definition of the axis of rotation and its corresponding atoms for 20 amino acid residues. On the basis of it, we put forward two computational models of moment of inertia based on atomic group, that is, order moment of inertia and contact order moment of inertia. For the total moment of inertia, the correlation coefficients between them and the experimental folding rates based on 103 proteins are-0.636 and-0.740, respectively. Last, based on contact order moment of inertia for 103 proteins, the correlation coefficient between the experimental and predicted folding rates with jackknife test is 0.768.By data analysis, there are three main conclusions: for two-state proteins, the topology has an important influence on the protein folding rates; for two-state proteins, the chain length is the main determinant of the protein folding rates; if the chain length is approximately equal,the topology of side chain is the uppermost determinant of the protein folding rates for two-state proteins.