Advancing Multi-Omics Analysis Through Proteomics-Centric Tools and Methods

Le résumé fourni par la source

Proteomics—the comprehensive study of proteins—has become central to biomedical research, especially as it is increasingly integrated with other omics modalities such as transcriptomics, metabolomics, and post-translational modification profiling. This dissertation advances proteomics-centric multi-omics analysis through the development of novel computational tools and their application to cancer research. First, we introduce FragPipe-Analyst and FragPipeAnalystR, tools designed to enhance analysis from the FragPipe proteomics platform, offering functionalities such as missing value imputation, data quality control, clustering, differential expression, and pathway enrichment. The R package FragPipeAnalystR further supports site-specific analyses of post-translational modifications and enables advanced visualization. Both tools are open-source and broadly demonstrate utility in protein analysis tasks across several published datasets. We then illustrate the application of these tools in large-scale studies of lung adenocarcinoma (LUAD) and acute myeloid leukemia (AML). In LUAD, the study—part of the international and interdisciplinary CPTAC/ICPC program—yielded a comprehensive proteogenomic compendium reflecting the demographic, etiological, clinical, and molecular complexity of LUAD, thus addressing key gaps in understanding this heterogeneous and deadly disease. Focusing on genomic aberrations, we examined the proteogenomic consequences of comparatively uncommon RBM10 mutations. Our findings suggest that beyond previously reported impacts on EGFR inhibitor resistance, RBM10 alterations influence the immune microenvironment, with potential implications for cytotoxic and immune checkpoint inhibitor therapies. We also confirmed prior findings regarding ALK fusions and functionally validated the associated phosphorylation signatures. Additionally, we identified two LUAD subtypes with STK11 mutations—subtypes that display distinct genome instability and NRF2 signaling differences not fully explained by KEAP1 co-mutation, echoing recent transcriptomics-based reports. This molecular heterogeneity within STK11-mutated LUAD underscores opportunities for patient stratification and therapeutic optimization. In AML, we aimed to systematically characterize disease heterogeneity by identifying both cross-cutting and subtype-specific molecular features through integrated proteogenomic analysis. Employing similarity network fusion and multi-omic factorization, we established a protein-centric subtyping scheme (AML-8), revealing strong genotype-phenotype associations where mutations in NPM1, CEBPA, and RUNX1-RUNX1T1 translocations constitute the principal structure of subtypes. Further partitioning of NPM1-mutated and MDS-related AMLs was consistent with established cellular differentiation hierarchies, highlighting the interplay between genotype and differentiation status as the primary axes of AML heterogeneity. Compared to other classification frameworks, AML-8 underscores the dual influence of genetic drivers and cellular differentiation states. Additionally, in addition to refined characterization of the previous Mito-AML subtype, we uncovered a novel subtype with NPM1-mutations characterized by high expression of known oncogenic transcription factors (POU2F1, FOXC1, and HOXB8), raising questions about unique therapeutic vulnerabilities. We also identified MAP1A as a universal protein marker for primitive, venetoclax-responsive AML, emphasizing its robust diagnostic potential, while acknowledging the need for further validation as a predictive biomarker. Furthermore, we identified GATA2 and TCF12 as shared stemness transcription factors among primitive AML, suggesting conserved regulatory mechanisms contributing to stemness and disease persistence. Most importantly, this work highlights how proteomic-centric multi-omic analysis could reveal novel findings and improve clinical classification and treatment. Collectively, these cancer studies illustrate the capacity and versatility of the developed proteomics-centric tools and analysis frameworks to advance both fundamental understanding and clinical interpretation of multi-omics data. This dissertation lays the groundwork for future expansion of multi-omics integration methods and underscores the potential of such approaches to drive novel biological discoveries in complex diseases.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.