Protein folding rate prediction integrating multi-level structural information

Studying protein folding can not only drive the great development of life sciences, but also provide tremendous help for human disease prevention and treatment, and has great application value in the fields of medicine and bioengineering. This article uses the BP neural network model to predict the rate of protein folding, and provides an effective way to find the key factors of protein folding kinetics. The main research contents are as follows: (1) Optimization of the neural network model. This article selected 4 types of optimizers and 36 types of activation function combinations to assess the performance of the neural network model in predicting the rate of protein folding. From the results, it is more accurate and fast to predict the rate of protein folding when using the Adam optimizer and Sigmoid and Tanh function combinations as the parameters of the neural network model. (2) The influence of chain length in primary structure information on prediction accuracy is studied and compared. From the prediction results, it was found that the prediction accuracy was higher when using the effective chain length of the protein than using the protein chain length. (3) Study the effect of different separation cutoffs on protein folding rates. The results show that when the separation cutoff value lcut of the contact order is 3, the most accurate prediction value can be obtained.