Manolis Kellis
Location: (Cambridge, MA)
Personal Research Web Page: http://mit.edu/manoli
Keywords: Computational Biology, Genomics, Epigenomics, Regulatory Motifs, Regulatory Networks, BioInformatics, Cancer, Development, Small RNAs, Evolutionary Genomics, Phylogenomics, Phylogenetics, Comparative Genomics
Posted on: Thursday, June 4th, 2009
Broad Research Area: AI / Machine Learning / Robotics / Vision, Databases / Information Retrieval / Data Mining, Information Systems / Information Science, Other, Scientific/Medical Informatics, Theory / Algorithms
Research Interests:
The CompBio group at MIT is sitting with the Computer Science and Artificial Intelligence Lab, and is closely affiliated with the Broad Institute of MIT and Harvard. The lab consists of machine learning and algorithms graduates, with strong bacxkground and interest in biological systems. Our research spans three areas:
(1) Genome interpretation: We use comparative genomics and epigenomics datasets to discover all functional elements in the human genome, including protein-coding genes, large and small non-coding RNAs, enhancers, promoters, and regulatory motifs. We use multi-variate HMMs, Conditional Random Fields, SVMs, and related machine learning techniques for integrating diverse sources of data for each of these elements.
(2) Gene regulation: We study the sequence signals that are responsible for gene regulation during health, disease, and embryo development. This involves regulatory motif discovery using EM, Gibbs Sampling, word enumeration and refinement, discovering informative combinations of regulatory motifs, and motif grammars. We use these to construct pre- and post-transcriptional regulatory networks and study the dynamics of information flow through biological systems. Lastly, we have used probabilistic models to discover recurrent combinations of chromatin marks, or chromatin states, and used these to understand the epigenomic dynamics of multiple cell types in human, and various developmental stages in fly.
(3) Genome evolution: We study the mechanisms and principles of gene and genome evolution. We use phylogenomics techniques to infer gene duplication and loss patterns over thousands of gene families spanning dozens of genomes, by learning common properties of gene trees encoded as gene-rate and species-rate distribution priors, and topology priors based on birth-death processes. We have applied such methods to uncover a whole-genome duplication in yeast and vertebrates, to discover pathogenic genes in Candida species, and to uncover regions under positive and negative selection in the human genome.
In each of these areas, we work closely with experimental collaborators, both for the generation of large-scale genomic and epigenomic datasets, and for the directed validation of our computational predictions in a truly interdisciplinary environment.
Contact Information:
email obfuscated - click to reveal; 617-253-2419
