EAGER: Cloud-based analysis of mass spectrometry proteomics data

蛋白质进行广泛的关键活动，包括信号传导，DNA复制，基因调控，免疫反应等串联质谱是唯一的高通量方法来表征这些蛋白质在复杂的生物混合物。该项目将产生算法，有可能大大提高世界各地蛋白质组学科学家解释串联质谱数据的能力，从而为这些基本生物过程提供关键见解。由于蛋白质是细胞中的主要功能分子，因此这些见解是我们对生命的科学理解的基础。该提案的关键智力贡献在于为蛋白质组学开发新的机器学习方法。具体而言，该项目将开发用于光谱特征提取的新方法，包括深度神经网络自动编码和低维表示学习，学习允许有效空间数据访问方法的深度相似性度量，自动编码在光谱和实验压缩中的应用，蛋白质组数据的子模块总结方法，以及通过数据嵌入利用网络效应的分类和排名方法。然后，这些工具将用于蛋白质组学元数据推断、基于云的肽和蛋白质鉴定以及类似蛋白质组学实验的检索和排名。因此，这项研究将产生一种更加强大的数据驱动方法，用于联合解释大量质谱数据集，从而为科学家提供有价值的新工具来深入了解蛋白质功能。

Proteins carry out a vast range of critical activities including signaling, DNA replication, gene regulation, immune response, etc. Tandem mass spectrometry is the only high-throughput way to characterize these proteins in complex biological mixtures. This project will produce algorithms with the potential to dramatically improve the ability of proteomics scientists around the world to interpret their tandem mass spectrometry data, thereby providing critical insights into these fundamental biological processes. Because proteins are the primary functional molecules in the cell, such insights are foundational to our scientific understanding of life.The key intellectual contributions in this proposal lie in the development of novel machine learning methods for proteomics. Specifically, the project will develop novel methods for spectrum feature extraction, including deep neural network autoencoding and low-dimensional representation learning, learning deep similarity metrics that allow efficient spatial data access methods, applications of autoencoding for spectrum and experiment compression, submodular summarization methods on proteomic data, and classification and ranking methodologies that take advantage of network effects via a data embedding. These tools will then be used for proteomic metadata inference, cloud-based peptide and protein identification, and retrieval and ranking of similar proteomic experiments. The research will hence yield a profoundly more powerful, data-driven approach to jointly interpreting massive mass spectrometry data sets, thereby giving scientists valuable new tools to glean insights into protein function.