Manolis Kellis
Associate Professor
Computer Science and Electrical Engineering Department
Massachusetts Institute of Technology
Regulatory and Systems Genomics of Complex Traits
While the methodological framework for relating genotype directly to
disease has been well established through the development of
genome-wide association studies (GWAS), the incorporation of
functional annotations, epigenomic information, and intermediate
molecular phenotypes is still a great challenge. In this talk, I’ll
describe some of our work seeking to address these challenges. (1) In
the context of ENCODE and the Roadmap Epigenomics program, I will
describe methods for integration of functional genomics datasets into
chromatin state annotations, linking of regulators to enhancer regions
and their target genes, and prediction and validation of causal
regulators using massively parallel reporter assays. (2) In the
context of disease network analysis, I will present three new methods
(a) for inference of regulatory networks from diverse functional
genomics datasets of gene expression, chromatin state, transcription
factor binding, and regulatory motif conservation; (b) for
deconvolution of direct and indirect effects in general
correlation-based and mutual information-based networks; and (c) for
integration of disease-associated variants in regulatory networks to
predict additional disease-associated genes. (3) In the context of the
Genotype-Tissue Expression (GTEx) project, I will present a new method
for discovering SNPs affecting multi-tissue expression patterns, by
learning expression modules jointly across individuals, and
discovering quantitative trait loci (QTLs) that alter the module
membership network between individuals, or network QTLs. (4) Lastly,
in the context of a longitudinal study of Alzheimer’s Disease (AD) in
a cohort of 750 individuals, I will present methods for studying the
interplay between genotype, DNA methylation patterns in the brain,
chromatin state, and Alzheimer’s disease using independent component
analysis and non-negative least squares for meQTL discovery, MWAS, and
genome-wide randomizations. Overall, our results suggest that global
regulatory changes are associated with complex disease, and suggest a
general methodology for integration of genetic and epigenetic
variation to uncover their systems-level effects.
