From: Narlikar, Leelavati (NIH/NLM/NCBI) [F]
Sent: Friday, May 01, 2009 10:30 AM
To: NLM/NCBI List ncbi-seminar
Subject: NCBI Seminar May 05 11:00 am, B2 NCBI Library, Building 38A 


Genome-wide discovery of enhancers involved in human embryonic heart 
development

NCBI Seminar May 05 11:00 am, B2 NCBI Library, Building 38A

Leelavati Narlikar
(Ovcharenko group)


Heart development requires the precise spatial and temporal expression of 
several genes, lack of which can cause severe cardiovascular 
abnormalities. A large fraction of gene regulation during development is 
mediated by the sequence-specific binding of transcription factors (TFs) 
to regulatory DNA elements. In this talk, I will focus on our work on 
identifying heart enhancers, i.e., DNA elements responsible for driving 
expression of genes in the developing heart.

We compiled a set of experimentally verified heart enhancers and a 
randomly drawn control set. Starting with over 700 sequence features based 
on known and de novo TF binding specificities, we learned a sparse linear 
classifier that distinguishes between the two sets. This classifier is 
superior to state-of-the-art methods; using a standard cross-validation 
procedure, we get an accuracy of 92% (measured as area under the ROC 
curve), while the closest competitor achieves only 72%. Furthermore, most 
of the features learned by our method like binding specificities of SRF, 
MEF2, ETS, SMAD, and GATA have been previously implicated in heart 
development. Using our classifier as a predictor, a genome-scan identified 
over 6000 novel heart enhancers in the human genome. Although the 
classifier used no gene-expression data during training, the predictions 
are associated with genes highly expressed in the heart. Furthermore, 
these predictions correlate with histone modification markers derived from 
ChIP experiments conducted in the mouse embryonic heart. Finally, 
preliminary in vivo tests of our predictions in mouse and zebra-fish 
models indicate a high validation rate of 75%.