Database and software development for protein-nucleic acid structure predication

描述（申请人提供）：我们的长期目标是蛋白质-核酸（NA，即RNA和DNA）结构预测，最终目标是合理的药物设计。蛋白质-RNA相互作用在RNA加工、基因表达和病毒复制中起关键作用。筛选抑制或促进特异性蛋白质-RNA结合的化合物以及设计与癌细胞表面抗原或其他疾病相关蛋白质靶点结合的短RNA序列（RNA适体）在药物开发中具有深远的应用。蛋白质-DNA相互作用对于转录、DNA损伤修复和细胞凋亡是必不可少的。因此，蛋白质-NA结构的可靠预测模型将对理解基本生物学过程和合理设计治疗干预措施产生深远的影响。然而，尽管广泛的生物医学意义的问题，蛋白质-RNA结构预测的计算研究仍然非常有限。关键瓶颈之一是缺乏实验确定的蛋白质-RNA复合物结构的大型训练和测试数据集。在这个项目中，我们建议通过构建严格的基准数据集，并首次为算法开发（包括参数训练），评估和系统改进提供灵活的诱饵，为蛋白质-RNA结构预测社区建立平台服务。我们还建议开发一种新的统计评分框架，通过从基准中提取分子相互作用信息并考虑分子的灵活性来预测蛋白质-RNA结构。包括灵活诱饵和软件在内的数据集将免费分发给学术界。该方法将被推广到蛋白质-DNA结构预测。我们还将通过预测RNA适体与前列腺特异性膜抗原（PSMA）的结合来测试我们算法的预测能力。我们的预测和合理设计的PSMA抑制RNA适体将通过实验测定进行彻底测试。我们预测和设计RNA适体的能力，结合癌细胞表面抗原，如PSMA具有高亲和力和特异性，将有很大的潜力，为靶向癌症诊断和治疗。

DESCRIPTION (provided by applicant): Our long-term goal is protein-nucleic acid (NA, namely RNA and DNA) structure prediction, for the ultimate goal of rational drug design. Protein-RNA interactions play critical roles in RNA processing, gene expression and viral replication. Screen of compounds that inhibit or promote specific protein-RNA binding and design of short RNA sequences (RNA aptamers) that bind to cancer cell-surface antigens or other disease-associated protein targets have profound applications in drug development. Protein-DNA interactions are essential for transcription, DNA damage repair and apoptosis. Reliable predictive model for protein-NA structures will thus have a far-reaching impact on understanding the fundamental biological processes and on rational design of therapeutic interventions. However, despite the widespread biomedical significance of the problem, computational studies on protein-RNA structure prediction remain very limited. One of the key bottlenecks is lack of large training and testing data sets of experimentally determined protein-RNA complex structures. In this project, we propose to establish a platform service for the protein-RNA structure prediction community by constructing rigorous benchmarking data sets and for the first time, flexible decoys for algorithm development (including parameter training), assessment and systematic improvement. We also propose to develop a new statistical scoring framework for predicting protein-RNA structures by extracting the molecular interaction information from the benchmarks and by accounting for molecular flexibility. The datasets including flexible decoys and the software will be freely distributed to the academic community. The methods will be generalized to protein-DNA structure predictions. We will also test the predictive power of our algorithms by predicting RNA aptamer binding to prostate-specific membrane antigen (PSMA). Our prediction and rational design of PSMA-inhibiting RNA aptamers will be tested thoroughly through experimental assays. Our ability to predict and design RNA aptamers that bind to cancer cell-surface antigens such as PSMA with high affinity and specificity will have great potential for targeted cancer diagnostics and therapy.