Poster Sessions

BIOINFO-18

Leila Taher

L. Taher, D. McGaugley, W. Miller, A. McCallion, I. Ovcharenko

Detecting Conserved Function of Evolutionarily Diverged Noncoding Elements

This work explores the occurrence of significant regulatory information in absence of sequence similarity. We modeled genomic sequences as arrangements of transcription factor binding sites (TFBSs), and aligned sequences of TFBSs instead of nucleotides. Our alignment model contemplates evolutionary events in regulatory sequences, e.g., matches, mismatches, and duplications of TFBSs. To train our model, we developed a strategy that reconstructs the TFBS structure of diverged sequences based on phylogenetic relationships among groups of species. Thereby, we constructed a set of 1,200 noncoding elements that have diverged between human and zebrafish but are likely to share the same ancestry. Our method correctly detects ancestral identity for over 75% of these elements embedded into 50kb background DNA. We evaluated the significance of the alignment scores by comparing true orthologs to sequences with similar GC-content, and enforced a false positive rate close to zero. Applying our method to a selected set of ~1000 human loci, we predicted 60 sequences that are very likely to share a common ancestor despite not being conserved between human and zebrafish. We are currently testing our predictions in transgenic zebrafish enhancer assays. This constitutes the first genome-wide computational method designed to reconstruct ancestral identity of diverged regulatory elements.