Bioinformatics Algorithms for Protein Interactions and Applications

蛋白质通过与其他蛋白质相互作用，形成蛋白质复合物，或与其他分子（如核酸和小配体）相互作用，驱动大多数细胞过程。蛋白质相互作用鉴定的实验技术仍然昂贵，劳动密集，耗时，并且存在高水平的假阳性预测。在计算机上准确预测这些相互作用是生物信息学的一个基本问题。我提出的研究计划涉及开发算法和工具，用于高度准确，可解释和适用的蛋白质相互作用预测。这项工作将影响生物学家确定、理解和使用蛋白质相互作用的能力。我建议构建的模型将通过web服务器广泛可用并易于访问。经过适当培训的用户将能够免费下载并运行高通量预测软件。高可信度的计算机预测可以大大减少生物学家的湿实验室工作，因为他们可以将精力集中在那些最有可能成功的相互作用上。了解这些相互作用将有助于解释蛋白质功能、构建PPI网络和研究疾病机制。这些应用程序可以被其他人用来开发新的治疗方法，例如，通过药物筛选、重新利用和副作用识别来对抗艾滋病毒、SARS-CoV-2等病毒。

Proteins drive most cellular processes by interacting with other proteins, to form protein complexes, or with other molecules such as nucleic acids and small ligands. Experimental techniques for protein interaction identification remain expensive, labor intensive, time consuming and suffer from high levels of false-positive predictions. Accurately predicting these interactions in silico is a fundamental problem in bioinformatics. My proposed research program concerns developing algorithms and tools for highly accurate, interpretable, and applicable protein interaction prediction. This work will impact the biologists' ability to determine, understand, and use protein interactions. The models I propose to build will be made widely available and easily accessible via web servers. Users with proper training will be able to freely download and run the software for high throughput prediction. High confidence in-silico predictions can dramatically reduce the wet lab work of biologists, as they can focus their efforts on those interactions that are most likely to be successful. Understanding the interactions will help to annotate protein functions, construct PPI networks, and study disease mechanisms. The applications presented can be used by others to develop novel therapeutics, e.g., against viruses like HIV, SARS-CoV-2, through drug screening, repurposing, and side effect identification.