UCLA

Researchers continue to struggle in deciphering the underlying molecular machinery of complex, multifactorial, and comorbid medical disorders. Integrating multiple layers of data –from genomic to exposomic – and evaluating their combinatorial effect on the phenome can mitigate limitations of simple differential analyses and ultimately help uncover causal factors.

In my dissertation work, I specifically focus on the integration of transcriptomic data with other data types that have a high clinical translatability such as phenomic and radiomic characteristics. I apply a multi-layered transcriptome-phenome-radiome integrative framework to two use case scenarios to demonstrate its benefits and drawbacks.

For use case scenario 1, I perform a multi-level analysis of RNA sequencing collected from in-house human placental decidual samples of various modes of parturition in late-stage pregnancy. I highlight differences in gene expression, co-expression, and alternative splicing and identify tissue- and labor-specific enrichment. I then incorporate dense prognostic and maternal and fetal phenomic information to derive genes and biological processes associated with premature and ceased labor. I demonstrate how an integrative framework successfully allows us to extract biologically relevant information that would have otherwise been missed through hypothesis-driven or monolayer differential analysis. For use case scenario. 2, I generate isoform-level information from RNA sequencing collected from The Cancer Genome Atlas (TCGA) GBM tumors. Using additional layers of the transcriptome, I filter for tumor-enriched genes to subtract microenvironment effects. I then incorporate 2 forms of quantitative morphologic radiomic features to extract exon inclusion-radiophenotype correlates. Through functional annotation, I highlight the underlying biological differences between tumor phenotypes. I demonstrate how an integrative framework provides exploratory insights into the biology of a GBM tumor yet fails to reveal significant associations due to data quality and analytical limitations.

The potential applications of a multi-layered and clinically-informed integration of the transcriptome, phenome, and radiome extend far beyond the immediate rejoice of joining systems biology efforts in the integration of “big data”. Through a synergistic coupling of functional molecular indexes, phenotypic characterization, and dense prognostic traits, it enables an in-depth and comprehensive investigation of multifactorial disorders. In the process, it uses a converged data- and hypothesis-mediated approach to balance the benefit of a comprehensive analysis approach and an elaborate mechanistic depiction of etiology. By incorporating individual-level information (from phenomic and radiomic traits) into population-level findings (from transcriptomic analyses), it poses as a promising contributor to the personalized and precision medicine initiatives of modern medicine.