Developing a real-time proteogenomics pipeline

为了最大限度地提高癌症等复杂疾病的治疗效果，医疗方法需要精确地根据个人的情况量身定制。随着精准医学触手可及，人们比以往任何时候都更加关注能让我们更好地了解我们的身体、它们的差异和功能障碍的技术。我们身体中的每个细胞都包含三个基本的生物实体：基因组（DNA）包含包括基因在内的遗传信息，如果表达，基因被转录成转录本（RNA），转录本本身随后被翻译成蛋白质，蛋白质控制着细胞的大部分行为。在过去的几十年里，创新的技术使得读取一个人的基因组或转录组变得可靠、实惠和快速。然而，准确有效地捕获细胞或其蛋白质组的蛋白质景观仍然是一个挑战。蛋白质组通常用霰弹枪质谱法研究。在这种方法中，蛋白质被消化成小肽，这些小肽被碎片化，用质谱仪进行鉴定。然而，蛋白质组是一个复杂的环境，生物样品含有比质谱仪可能测量的更多的肽。因此，质谱仪只选择最丰富的片段和鉴定。这种策略可以防止片段化和检测可能携带重要生物信息的较少丰度的肽。为了解决这个问题，软件控制质谱仪的行为和实时调整多肽的选择正在出现。然而，在蛋白质组学分析中使用的次优理论框架仍然存在偏见。质谱仪产生蛋白质的足迹（称为质谱）。蛋白质组学数据的分析包括将每个足迹与正确的蛋白质联系起来。目前的方法依赖于分析样本中预先建立的所有可能蛋白质的列表，通常来自样本起源物种的已知基因组和基因。然而，最近的方法和偶然的发现强调了蛋白质的存在，它们逃避了基因注释。在人类中，数以万计的蛋白质从已知的基因组注释中缺失。因此，目前的蛋白质组学数据分析方法无法识别这些蛋白质。在这个NOVA-FRQNT-NSERC项目中，我们建议开发一个用于生物样品实时蛋白质基因组学分析的软件管道。我们将开发一种人工智能方法，通过质谱实时识别蛋白质；我们将使用基因组数据（DNA和/或RNA测序数据）来控制质谱仪的行为，以最大限度地确定生物样品中的蛋白质。我们将评估我们的管道在哺乳动物细胞和酵母经受或不经受氧化应激。氧化应激是一种已知的蛋白质组调节因子，在应激下多种非注释蛋白表达。此外，我们将使用复杂组织的单细胞测序来增加基于这种深入转录组学数据的蛋白质鉴定的敏感性。该计划将开发一种蛋白质组学方法，具有巨大的研究和临床应用潜力。此外，这将是一个独特的机会，让我们小组的学生同时获得前沿算法和实验室技术的跨学科经验，培养他们成为未来的混合计算和实验研究人员。

To maximize treatment efficacy in complex diseases such as cancer, medical approaches need to be precisely tailored to an individual's condition. With precision medicine within reach, the focus is more than ever on technologies enabling us to better understand our bodies, their differences and their dysfunctions. Each cell in our body contains three essential biological entities: the genome (DNA) contains the genetic information including genes, which are transcribed into transcripts (RNA) if expressed, which themselves are then translated into proteins, which control much of the cells' behaviour. Throughout the last decades, innovative technologies have made reading one's genome, or transcriptome, reliable, affordable and quick. However, the challenge is still ongoing to accurately and efficiently capture the protein landscape of cells or their proteome.The proteome is commonly investigated by shotgun mass spectrometry. In this approach, proteins are digested into small peptides that are fragmented to be identified using a mass spectrometer. However, the proteome is a complex environment and biological samples contain more peptides than a mass spectrometer can possibly measure. Thus, mass spectrometers select only the most abundant for fragmentation and identification. This strategy prevents the fragmentation and detection of less abundant peptides that may carry important biological information. To address this problem, software controlling mass spectrometers' behavior and adjusting the selection of peptides in real-time are emerging.Yet, these are still biased by the suboptimal theoretical framework used in proteomics analyses. The mass spectrometer yields footprints (called mass spectra) of proteins. The analysis of proteomics data consists in linking each footprint to the correct protein. Current methods rely on a pre-established list of all possible proteins in the analyzed sample, generally derived from the known genome and genes of the species from which the sample originates. However, recent methods and serendipitous discoveries have highlighted the existence of proteins, which have eluded gene annotations. In human, tens of thousands of proteins are lacking from known genome annotations. Therefore, these proteins can never be identified by current methods of analysis of proteomic data.In this NOVA-FRQNT-NSERC Program, we propose to develop a software pipeline for real-time proteogenomic analyses of biological samples. We will develop an artificial intelligence method to identify proteins by mass spectrometry in real-time; and we will control the mass spectrometer behavior using genomic data (DNA and/or RNA sequencing data) to maximize protein identifications in biological samples. We will evaluate our pipeline in mammalian cells and yeasts subjected or not to oxidative stress. Oxidative stress is a known regulator of the proteome and multiple non-annotated proteins are expressed under stress. Furthermore, we will use single-cell sequencing of a complex tissue to increase the sensibility of the identification of proteins based on this in-depth transcriptomic data. This program will develop a proteomics method that has a great potential for research and use in clinics. Furthermore, it will be a unique opportunity for students in our groups to gain interdisciplinary experience in cutting-edge algorithms and laboratory techniques concomitantly, training them as hybrid computational and experimental researchers of the future.