Collaborative Research: FET: Small: De Novo Protein Scaffold Filling by Combinatorial Algorithms and Deep Learning Models

蛋白质测序在鉴定蛋白质功能、分析蛋白质相互作用和表征翻译后修饰方面起着重要作用。尽管最近在蛋白质测序和组装方面取得了进展，但许多目前可用的组装蛋白质都是草稿形式。即使将自顶向下和自底向上的测序方法相结合，所组装的蛋白质序列仍然存在许多空白。换句话说，在特定蛋白质的测序步骤结束时，更有可能看到与间隙分开的contigs（称为支架）。因此，一个重要但也是自然的组合问题是将缺失的氨基酸填充到支架中以获得完整的蛋白质序列。在本项目的新框架下，从头蛋白质测序将极大地推动蛋白质功能和结构鉴定的研究和临床实践。该项目将为研究人员获取抗体序列信息提供强大的计算工具，这对构建抗体数据库具有重要价值。这种跨学科的研究也为各个层次的学生，特别是代表性不足的非洲裔美国学生提供各种培训项目，帮助他们从开放和跨学科的角度追求高质量的研究。取得的新进展将纳入本科/研究生课程。研究结果将通过期刊出版物、会议、开源软件发布、教程和研讨会演讲等方式进行传播。在本项目中，研究者将通过两个相关阶段研究基于质谱的从头蛋白支架填充问题。首先，在没有适当参考的情况下，研究人员将分析自上而下和自下而上的串联质谱来构建蛋白质支架。这些方法包括通用全局优化、动态规划和图算法，也可以处理小的蛋白质突变（其中一些氨基酸的质量发生了变化）。其次，研究人员将使用深度学习方法，如卷积神经网络与长短期记忆相结合、卷积去噪自编码器、变压器模型等，完成第一步自上而下和自下而上串联质谱分析获得的蛋白质测序的最后一步。该项目将形成蛋白质支架填充的组合和深度学习方法的新框架，以及相应的开源软件。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Protein sequencing plays an important role in identifying protein functions, analyzing protein-protein interactions, and characterizing post-translational modifications. Despite the recent progress in protein sequencing and assembly, many of the currently available assembled proteins come in a draft form. There are still many gaps in the assembled protein sequences even if one combines top-down and bottom-up sequencing methods. In other words, at the end of the sequencing step for a specific protein, it is more likely to see contigs separated with gaps (which is called a scaffold). Hence, an important but also natural combinatorial problem is to fill the missing amino acids into a scaffold to obtain a complete protein sequence. With the new framework produced by this project, de novo protein sequencing will greatly advance the research and clinical practice of identifying the function and structure of proteins. The project will provide researchers with powerful computational tools for obtaining the sequence information of antibodies, which is extremely valuable for the construction of antibody databases. This interdisciplinary research also provides various training projects to students at all levels, particularly to underrepresented African American students, and helps them to pursue high quality research from an open-minded and cross-disciplinary perspective. New advances achieved will be integrated into undergraduate/graduate curricula. The results will be disseminated through journal publications, conferences, open-source software release, tutorials, and seminar talks.In this project, the investigators will study the mass spectrometry-based de novo protein scaffold filling problem by two related phases. Firstly, the investigators will analyze the top-down and bottom-up tandem mass spectrometry to construct the protein scaffold without a proper reference. The methods include general global optimization, dynamic programming, and graph algorithms, which can also handle small protein mutations (where the mass of some amino acid changes). Secondly, the investigators will use deep learning methods, such as combined convolutional neural network and long short-term memory, convolutional denoising autoencoder, and transformer models to finish the last step of protein sequencing obtained by top-down and bottom-up tandem mass spectrometry analysis at first step. The project will result in a new framework of combined combinatorial and deep learning methods for protein scaffold filling, and a corresponding open-source software.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.