蛋白质结构预测综述

Survey on protein structure predication

蛋白质结构预测对于从分子层面理解蛋白质的生物功能具有重要意义。本研究从同源建模、自由建模等经典方法以及深度学习这几个方面来阐述蛋白质结构预测方面的进展。已知结构蛋白质模板数量的增加、序列比对等算法对信息提取能力的提升以及片段拼接技术的应用使得同源建模在预测蛋白结构的能力大大提升。域分割和片段分割技术及并行计算策略的应用使得自由建模方法在预测远程氨基酸接触能力不断提升。深度学习技术与以上经典方法的结合提升了蛋白结构预测的准确性和速度,但是对于没有同源性蛋白结构的预测,仍然存在巨大的挑战。

The prediction of protein structure has provided important underpinnings for understanding the biological functions of protein at molecular level.Herein the development of protein structure predication is reviewed from the aspects of homology modeling,ab initio prediction and deep learning.Because of the increasing number of protein templates with known structure,the improved performance of sequence alignment algorithm for information extraction and the application of fragment assembly technology,the ability of homology modeling for predicting protein structure is greatly increased.The prediction of long-distance amino-acid contact using ab initio method is more and more precise for the application of regional partition and fragment segmentation and the strategy of parallel computing in ab initio method.Deep learning combined with the above-mentioned classical technologies can promote the accuracy and speed of protein structure prediction,but there is still a great challenge when the protein has no homologous proteins.