Adaptive computing for In Silico recognition of hormone response elements

Abstract

An important step in understanding the mechanisms of gene expression regulation is recognizing DNA areas associated with regulation of transcription. Due to high diversity of transcription factors and mechanisms of their interaction with DNA targets, it is a challenging problem to establish an accurate model for computational prediction of functional regulatory elements in promoters of eukaryotic genes. A novel high-performance approach to recognition of symmetrically structured DNA motifs is described and tested by the example hormone response elements. For recognizing the motifs, we consider combined statistic modeling of the symmetric pattern. We also invent a highly specific two-phase neural architecture which exploits a different motif recognition paradigm. Hardware acceleration is proposed for resolving computational bottlenecks, and the hybrid tool is further used for analysis of hormone primary target genes. For the problem of accurate recognition of partially symmetric DNA motifs, we conclude that the developed hardware architecture is highly efficient for acceleration of computations, and makes the invented approach applicable for high-throughput and/or genome-wide analysis.