Search

Integrative approaches that simultaneously model multi-omics data have gained increasing popularity because they provide holistic system biology views of multiple or all components in a biological system of interest. Canonical correlation analysis (CCA) is a correlation-based integrative method designed to extract latent features shared between multiple assays by finding the linear combinations of features–referred to as canonical variables (CVs)–within each assay that achieve maximal across-assay correlation. Although widely acknowledged as a powerful approach for multi-omics data, CCA has not been systematically applied to multi-omics data in large cohort studies, which has only recently become available. Here, we adapted sparse multiple CCA (SMCCA), a widely-used derivative of CCA, to proteomics and methylomics data from the Multi-Ethnic Study of Atherosclerosis (MESA) and Jackson Heart Study (JHS). To tackle challenges encountered when applying SMCCA to MESA and JHS, our adaptations include the incorporation of the Gram-Schmidt (GS) algorithm with SMCCA to improve orthogonality among CVs, and the development of Sparse Supervised Multiple CCA (SSMCCA) to allow supervised integration analysis for more than two assays. Effective application of SMCCA to the two real datasets reveals important findings. Applying our SMCCA-GS to MESA and JHS, we identified strong associations between blood cell counts and protein abundance, suggesting that adjustment of blood cell composition should be considered in protein-based association studies. Importantly, CVs obtained from two independent cohorts also demonstrate transferability across the cohorts. For example, proteomic CVs learned from JHS, when transferred to MESA, explain similar amounts of blood cell count phenotypic variance in MESA, explaining 39.0% ~ 50.0% variation in JHS and 38.9% ~ 49.1% in MESA. Similar transferability was observed for other omics-CV-trait pairs. This suggests that biologically meaningful and cohort-agnostic variation is captured by CVs. We anticipate that applying our SMCCA-GS and SSMCCA on various cohorts would help identify cohort-agnostic biologically meaningful relationships between multi-omics data and phenotypic traits.

Various individual omics data allow researchers to gain insights on the biological process of the human body, the mechanisms of complex disease, and the development of therapeutics. In recent years, a large amount and various types of omics data have been generated, and more biology has been uncovered by integrating multiple types of omics data. My dissertation presents three projects involving single omics, two omics, and more omics data into analysis, all of which bring new understanding to the molecular mechanism of phenotypes of interest. In my first project, utilizing only genomics data, I follow up and extend the Raffield et al.’s (2018) research by involving a larger sample of under-represented individuals to study gene-gene interaction between a ~3.7k length deletion in the hemoglobin α gene and sickle cell traits on the risk of anemia and chronic kidney disease. About ten times as many African American and Hispanic/Latino participants have substantially enhanced the statistical power for detecting and qualifying the interaction effects. This project reflects the essentiality of the genomic analysis for underrepresented individuals, especially for high-impact variants, as well as the importance of studying gene-gene interactions in understanding complex diseases. In my second project, I utilize proteomics and genomics data to study the genetic determinants for 21 inflammation-related proteins. This project enlarges our knowledge regarding the genetics underlying inflammatory biomarkers which could further elucidate mechanisms of a battery of inflammation induced or related diseases. Finally, in my third project, I improve the sparse multiple canonical correlation analysis proposed by Witten et al. (2009). Specifically, my improvements lie in two aspects: first, I adopt the Gram-Schmidt algorithm to improve orthogonality of inferred canonical vectors; second, I extend the supervised method to allow more than two omics data types, while the original implementation only allows two omics data. With my modified method, I found a strong association between blood cell counts and top canonical vectors from proteomic data, which has important implications for association studies involving proteomics data from blood. In summary, my dissertation projects studying various types of omics data have not only led to useful findings, but also present areas where the field can benefit from the development and application of statistical/computational frameworks for analyzing multi-omics data. We anticipate more biological insights can be gained by efficiently analyzing multi-omics data.

Additional file 1: Figure S1. The phylogenetic tree of 108 novel taxon candidates. The phylogenetic tree was constructed with the 16S rRNA gene sequences of each strain using MEGA7 [80] under the neighbor-joining method. The bootstrap value is 1000. The ANIs between adjacent taxa on the tree was calculated using OrthoANI OTA solfware [80] and list in the panel (red color). The names of 102 novel-taxon candidates that were later identified to represent novel taxa were colored in blue, while the 6 candidates that were later determined to be new strains of known species were colored in grey and the 16S rRNA gene identity to the known species were listed in the brackets.

Additional file 16: Supplementary Methods. The documentation of detailed recipes and references of basic media used in this study.

Inflammation biomarkers can provide valuable insight into the role of inflammatory processes in many diseases and conditions. Sequencing based analyses of such biomarkers can also serve as an exemplar of the genetic architecture of quantitative traits. To evaluate the biological insight, which can be provided by a multi-ancestry, whole-genome based association study, we performed a comprehensive analysis of 21 inflammation biomarkers from up to 38,465 individuals with whole-genome sequencing from the Trans-Omics for Precision Medicine (TOPMed) program. We identified 22 distinct single-variant associations across 6 traits - E-selectin, intercellular adhesion molecule 1, interleukin-6, lipoprotein-associated phospholipase A2 activity and mass, and P-selectin - that remained significant after conditioning on previously identified associations for these inflammatory biomarkers. We further expanded upon known biomarker associations by pairing the single-variant analysis with a rare variant set-based analysis that further identified 19 significant rare variant set-based associations with 5 traits. These signals were distinct from both significant single variant association signals within TOPMed and genetic signals observed in prior studies, demonstrating the complementary value of performing both single and rare variant analyses when analyzing quantitative traits. We also confirm several previously reported signals from semi-quantitative proteomics platforms. Many of these signals demonstrate the extensive allelic heterogeneity and ancestry-differentiated variant-trait associations common for inflammation biomarkers, a characteristic we hypothesize will be increasingly observed with well-powered, large-scale analyses of complex traits.