Enabling precision oncology through multimodal data integration and visualization

The first theme of the thesis focuses on genomic variant interpretation, highlighting tools that streamline the annotation and classification of variants. Chapter 2 presents Genome Nexus, a tool designed to aggregate and harmonize data from multiple genomic databases, making it easier for researchers and clinicians to interpret cancer variants at scale. Genome Nexus integrates data from resources like Ensembl, ClinVar, and OncoKB, offering users a centralized platform for comprehensive variant annotation. By standardizing and linking these datasets, Genome Nexus improves the accuracy and efficiency of identifying actionable mutations, addressing a critical bottleneck in precision oncology workflows.

Chapter 3 expands on this theme by introducing reVUE, a resource for curating and classifying the protein effects of variants of unexpected efffects (VUEs). This tool addresses a common challenge in cancer genomics: interpreting the functional impact of complex variants not accurately predicted by popular bioinformatics tools, such as the Variant Effect Predictor (VEP). For example, actionable inframe driver alterations in KIT in gastrointestinal stromal tumors are inaccurately classified as splice alterations by VEP, potentially causing patients not to receive imatinib treatment. Through expert literature curation and data mining, reVUE provides a systematic way to capture these corner cases, addressing a critical need for routine clinical sequencing.

The second theme of the thesis shifts toward integrating multimodal datasets that include genomics and other data types and enhancing their usability for the broader cancer research community. Chapter 4 describes innovations in cBioPortal, one of the most widely used platforms for visualizing and analyzing cancer data. Enhancements to cBioPortal include new functionality for analyzing longitudinal datasets, exemplified by its application to the AACR Project GENIE Biopharma Collaborative dataset. The cBioPortal allows researchers to easily use statistical tools in the browser to explore drug responses, resistance mechanisms, and clinical outcomes.

Chapter 5 addresses how the Human Tumor Atlas Network (HTAN) has leveraged advanced profiling techniques—such as single-cell sequencing, spatial transcriptomics, and multiplex imaging—to generate high-resolution tumor maps. These datasets were systematically organized and made accessible through user-friendly interfaces, enabling researchers to explore tumor microenvironments and cellular heterogeneity with unprecedented detail. This chapter highlights the critical role of data sharing and standardization in accelerating discoveries across the cancer research ecosystem.

Finally, Chapter 6 showcases the power of spatial data integration, focusing on a case study in colorectal cancer. By combining spatial transcriptomics and genomic data within cBioPortal, the study identified immune-infiltrated colorectal cancers with significantly improved progression-free survival. This chapter underscores the potential of multimodal approaches to unravel the complexities of tumor biology and improve patient stratification for targeted therapies.