500A Poster - 06. Regulation of gene expression
Thursday April 07, 2:00 PM - 4:00 PM

Authors:
Prasad Bandodkar ¹; Elle Rooney ²; Samiul Haque ²; Cranos Williams ²; Gregory Reeves ¹

Affiliations:
1) Texas A&M University; 2) North Carolina State University

Keywords:
m. computational models; n. networks

In most well-characterized Gene Regulatory Networks (GRNs), maps of GRNs remain incomplete. The GRN for regulating anterior-posterior (AP) patterning in early Drosophila embryos, is a well-studied network and most of the major components have been identified. However, the minor components of the network, responsible for compensatory regulation, remain elusive. Typical molecular techniques, such as a knockout, could result in several lost connections, swamping out compensatory regulation. Instead, we are using natural variation in the ~200 fly lines in the Drosophila Genetic Reference Panel (DGRP) to identify and characterize novel components of the network, using machine learning techniques.

The segmentation gene network that patterns the anterior-posterior axis in the early embryo is hierarchical and we are investigating the time period that is roughly mid-way between the start of nuclear cycle 14 and cellularization. During this time period, maternal coordinate genes such as Bicoid (Bcd), activate and regulate Gap genes such as Krüppel (Kr), and together they regulate pair-rule genes such as even-skipped (eve). The idea is to quantify the spatial expression patterns of the genes of interest to the segmentation GRN and use the subtle differences in spatial positions to map the cause back to specific regions in the genome and to the transcriptome. With 13 fly lines and imaging two genes- Kr and eve, searching within a 20 kb region of the genome, we were able to identify pangolin (pan) as a novel component of the AP patterning network. With data from ~200 fly-lines, we expect to have enough statistical power to identify several new components. We have built image analysis pipelines to reliably extract spatial gene expression data from a large number of stained embryos without manual supervision. Using unsupervised machine learning techniques over dataset from all ~200 fly lines and searching across the genome and the transcriptome, we expect to uncover novel components of the AP patterning network, or at the very least better characterize the existing network.