Indiana University-Purdue University Indianapolis (IUPUI) / Proteoforms are distinct protein molecule forms created by variations in genes, gene expression, and other biological processes. Many proteoforms contain multiple primary structural alterations, including amino acid substitutions, terminal truncations, and posttranslational modifications. These primary structural alterations play a crucial role in determining protein functions: proteoforms from the same protein with different alterations may exhibit different functional behaviors. Because top-down mass spectrometry directly analyzes intact proteoforms and provides complete sequence information of proteoforms, it has become the method of choice for the identification of complex proteoforms. Although instruments and experimental protocols for top-down mass spectrometry have been advancing rapidly in the past several years, many computational problems in this area remain unsolved, and the development of software tools for analyzing such data is still at its very early stage. In this dissertation, we propose several novel algorithms for challenging computational problems in proteoform identification by top-down mass spectrometry. First, we present two approximate spectrum-based protein sequence filtering algorithms that quickly find a small number of candidate proteins from a large proteome database for a query mass spectrum. Second, we describe mass graph-based alignment algorithms that efficiently identify proteoforms with variable post-translational modifications and/or terminal truncations. Third, we propose a Markov chain Monte Carlo method for estimating the statistical signi ficance of identified proteoform spectrum matches. They are the first efficient algorithms that take into account three types of alterations: variable post-translational modifications, unexpected alterations, and terminal truncations in proteoform identification. As a result, they are more sensitive and powerful than other existing methods that consider only one or two of the three types of alterations. All the proposed algorithms have been incorporated into TopMG, a complete software pipeline for complex proteoform identification. Experimental results showed that TopMG significantly increases the number of identifications than other existing methods in proteome-level top-down mass spectrometry studies. TopMG will facilitate the applications of top-down mass spectrometry in many areas, such as the identification and quantification of clinically relevant proteoforms and the discovery of new proteoform biomarkers. / 2019-06-21

Les protéoformes désignent toutes les formes sous lesquelles une protéine peut être présente. Cela inclut les formes portant des modifications post-traductionnelles, les isoformes et les formes résultant d’épissages alternatifs. Ces modifications peuvent influer sur la fonction d’une protéine, d’où l’intérêt de développer des méthodes sensibles, spécifiques et robustes de quantification de protéoformes pour une meilleure compréhension de mécanismes pathologiques ou la recherche de biomarqueurs. L’objectif de ce travail a été d’exploiter les avantages de la spectrométrie de masse à haute résolution pour la caractérisation et la quantification de protéoformes de protéines associées à des maladies neurodégénératives. Nous nous sommes principalement intéressés à deux protéines, la C9ORF72 et l’alpha-synucléine. Dans un premier temps, une méthode de quantification des deux isoformes de la C9ORF72 a été développée et appliquée à une cohorte de 43 cerveaux humains comprenant des cas de démences fronto-temporale (DFT) avec ou sans mutation C9ORF72. Les résultats obtenus montrent pour la première fois par spectrométrie de masse une diminution d’environ 50% de l’isoforme longue de la C9ORF72 en présence de la mutation. Dans une seconde étape, nous avons élargi l’analyse à la quantification multiplexe de 49 protéines cérébrales potentiellement impliquées dans les DFT. En parallèle, nous nous sommes intéressés aux formes tronquées de l’alpha-synucléine. Leur quantification a été réalisée par une approche top-down dans des tissus cérébraux et une approche par bottom-up dans le liquide céphalorachidien (LCR). Enfin, l’analyse a été étendue à la quantification multiplexe de l’alpha-synucléine et de la protéine tau du LCR. / Proteoforms describe the complexity of protein forms. This includes forms with post-translational modifications, isoforms and forms resulting from alternative splicing. These modifications can influence the function of a protein, hence the interest in developing sensitive, specific and robust methods for the quantification of proteoforms for a better understanding of pathological mechanisms or biomarkers discovery. The objective of this work was to take advantage of high-resolution mass spectrometry for the characterization and quantification of proteoforms associated with neurodegenerative diseases. We mainly focused on two proteins, C9ORF72 and alpha-synuclein. First, a method for quantifying the two C9ORF72 isoforms was developed and applied to a cohort of 43 human brains including cases of frontotemporal dementia (FTD) with or without C9ORF72 mutation. The results obtained show for the first time by mass spectrometry a decrease of about 50% of the long isoform of C9ORF72 in the presence of the mutation. Then, we extended the analysis to the multiplex quantification of 49 brain proteins potentially involved in FTD. In parallel, we focused on the truncated forms of alpha-synuclein. Their quantification was performed by a top-down approach in brain tissue and a bottom-up approach in cerebrospinal fluid (CSF). Finally, the analysis was extended to the multiplex quantification of alpha-synuclein and tau protein in CSF.