Integrating genomic data to build more predictive models of biology.
Fine, Alexander.
2019
-
By measuring
multiple characteristics of a biological system at once, one can get a more complete
picture of how a perturbation affects that system. Single-omic measurements sometimes
fail to correlate well with other genome-wide measurements, demonstrating the value of
measuring a biological system at multiple levels. However, knowing which aspects are
informative and how to best integrate ... read morethem into a comprehensive model is a challenge
that persists within the field of systems genetics. Here, we utilize induced and
naturally-occurring perturbations across multiple biological systems to build more
predictive models of biology across three distinct projects. For each project, we
integrated a combination of genetic, epigenetic, transcriptional, and cellular
measurements to better understand their role in the complex biological system, gaining a
more comprehensive understanding than could be accomplished through examining any single
measurement alone. First, we disrupted normal germ cell differentiation by knocking out
the key meiotic protein PRDM9, which deposits epigenetic modifications throughout the
genome to select sites for meiotic recombination. We integrated cellular and molecular
phenotypes to determine how cytological and transcriptional programs remain coupled in
case of developmental arrest. We found that these two developmental programs became
uncoupled in the perturbed system, demonstrating the need to pair transcriptional
analyses with classical molecular biology in order to put gene expression data within
their proper context. Next, we employed genetic variation within and between the
C57BL/6J and CAST/EiJ genomes to test the effect on binding affinity of variants within
the binding site of the PRDM9 long zinc-finger array. We showed that, in addition to
their individual effects, combinations of certain base-pairs across the binding site of
the zinc-finger had epistatic effects on binding affinity. Finally, we integrated
genetic, epigenetic, and transcriptomic data from hepatocytes of nine diverse strains of
mice to build a multi-omic model of gene expression. Our work quantified the
inter-dependency among genetic, epigenetic, and gene expression variation across these
strains. These projects used existing and novel tools for data integration and analysis
to go beyond the scope of any single-omic analysis, demonstrating the power and
necessity of multi-omic analyses in the field of modern
genetics.
Thesis (Ph.D.)--Tufts University, 2019.
Submitted to the Dept. of Genetics.
Advisors: Gregory Carter, and Jennifer Trowbridge.
Committee: Mary Ann Handel, Steven Munger, and Amy Yee.
Keywords: Genetics, Computer science, and Biology.read less - ID:
- t722hp35m
- To Cite:
- TARC Citation Guide EndNote
