Using Sequence-Predicted Contacts to Guide Template-free Protein Structure Prediction
Using Sequence-Predicted Contacts to Guide Template-free Protein Structure Prediction
复制标题
DOI：
10.1145/3307339.3342175
复制
发表时间：
2019-09
期刊：
影响因子：
--
通讯作者：
Ahmed Bin Zaman;Prasanna Parthasarathy;Amarda Shehu
中科院分区：
文献类型：
--
作者：
Ahmed Bin Zaman;Prasanna Parthasarathy;Amarda Shehu
无模板蛋白质结构预测寻求以降低原子间能量的方式组织连续键合的氨基酸的三维结构。现在很好地理解，能量函数是生物活性结构的不可靠指南。这种认识提出了关于能量功能的适当作用和利用的问题。最近的工作建议采用氨基酸接触形式的互补信息。在这里，我们推进这一工作，并利用多目标优化来研究原子间能量和基于接触的评分的各种组合。不同数据集上的评估表明，在无模板蛋白质结构预测的多目标优化设置中，将接触信息与能量函数相结合的优越性。
Template-free protein structure prediction seeks three-dimensional structures that organize the serially-bonded amino acids in ways that lower interatomic energy. It is now well understood that energy functions are unreliable guides towards biologically-active structures. This realization raises questions on the proper role and utilization of energy functions. Recent work suggests employing complementary information in the form of amino-acid contacts. Here, we advance this line of work and leverage multi-objective optimization to investigate a variety of combinations of interatomic energy and contact-based scoring. Evaluation on diverse datasets demonstrates the superiority of combining contact information with energy functions in a multi-objective optimization setting for template-free protein structure prediction.