A new scoring framework for selecting structural models

描述（由申请人提供）：可靠和高效的能量评分功能对于准确的蛋白质结构预测，蛋白质设计和计算机辅助药物发现至关重要。不幸的是，这样的能量评分函数仍然存在。可能最成功的评分函数类型是基于统计潜力（也称为基于知识）的评分函数。尽管取得了显著的成功，但这些评分函数存在以下问题：1)对其成对势能函数的推导过于简化；2)只考虑（低能量）原生结构，而忽略（高能）非原生结构。因此，这些评分函数很难从大量的诱饵（即非原生）结构中识别原生结构。例如，通常发现基于统计电位的评分函数在预测蛋白质-配体结合模式方面成功率相对较低，并且在虚拟数据库筛选中失败。在这个项目中，我们提出了一种新的能量评分函数，用于预测蛋白质结构和蛋白质与RNA， DNA或配体的相互作用。我们基于统计力学的方法的新颖之处在于：}1)包括非原生状态/结构以获得更好的构象采样；2)使用新颖的迭代方法严格推导有效的成对势函数。我们将使用已知的不同集合测试和改进我们的新得分函数。在这个项目中开发的所有源代码和可执行文件将免费提供给公众。为了直接验证我们的方法，我们与实验人员建立了密切的合作关系，研究一种新型抗癌药物PRIMA-1的作用机制。这项生物信息学驱动的研究可能会导致治疗和/或预防人类乳腺癌的潜在治疗应用。我们的初步结果表明，我们的新能量评分函数具有良好的性能。我们的初步研究还发现了一种新的有效药物，可以显著杀死人类乳腺癌细胞。我的生物信息学专业知识与我的合作者的生物化学和癌症研究专业知识的协同结合，为我们找到PRIMA-1的分子靶点铺平了道路，希望能找到治疗和/或预防人类乳腺癌的新型抗肿瘤药物。

DESCRIPTION (provided by applicant): Reliable and efficient energy scoring functions are vitally important for accurate protein structure prediction, protein design and computer-aided drug discovery. Unfortunately, such energy scoring functions still remain at large. Probably the most successful type of scoring functions is the statistical potential-based (also referred to as knowledge-based) scoring functions. Despite achieving significant success, these scoring functions suffer from 1) oversimplified derivation of their pairwise potential energy functions and 2) sole consideration of (low-energy) native structures while ignoring (high-energy) non-native structures. Consequently, these scoring functions have difficulty in discerning native structures from a large ensemble of decoy (i.e., non-native) structures. For instance, statistical potential-based scoring functions were usually found to have relatively low success rates in predicting protein-ligand binding modes and failed in virtual database screening. In this project we propose to derive a new type of energy scoring functions for predicting protein structures and protein interactions with RNA, DNA, or ligands. The novelty of our statistical mechanics-based approach is two-fold:} 1) including the non-native states/structures for better conformational sampling, and 2) using a novel iterative method to rigorously derive the effective pairwise potential functions. We will test and refine our new scoring functions using known diverse sets. All the source codes and executables developed in this project will be freely available to the public. To directly test our methods, we have established closed collaborations with experimentalists on studying the mechanism of a novel anti-cancer agent PRIMA-1. This bioinformatics-driven study may lead to potential therapeutic application for treatment and/or prevention of human breast cancer. Our preliminary results show promising performance of our new energy scoring functions. Our preliminary studies have also identified a new potent agent that dramatically kills human breast cancer cells. The synergetic combination of my bioinformatics expertise with my collaborators' biochemical and cancer research expertise paves our way to find molecular target(s) of PRIMA-1 with the hope of identifying novel anti-tumor agents for treatment and/or prevention of human breast cancer.