47. DETECTING INDIVIDUAL TRANSCRIPTION SITES IN DROSOPHILA EMBRYOS

Department: Computer Science & Engineering
Faculty Advisor(s): Yoav Freund

Primary Student
Name: William M Beaver
Email: wbeaver@ucsd.edu
Phone: 917-783-5064
Grad Year: 2009

Abstract
Kosman developed a Fluorescent In Situ Hybridization (FISH) method that allows the visualization of mRNA concentration at sub-cellular resolution. In particular, his method allows counting of the active transcription sites in nuclei (0, 1 or 2 sites) for a few different genes simultaneously. This new method might give new insights on the dynamics of gene regulation during the spatial differentiation of cells in embryonic development.

The problem is that translating the multi-channel, confocal image stacks into per-cell quantitative measurements is prohibitively labor intensive. We developed a method through which the experimentalist teaches the computer how to segment individual nuclei and how to identify gene transcription sites inside individual nuclei. We completed the analysis on fifty confocal stacks from fifty different Drosophila embryos, all at approximately the same stage of development. Our method misses less than one out of every five hundred nuclei and we have no false detections. Using a similar segmentation method we identify the nascent transcription sites of four genes within each image stack. We achieve high accuracy for sna, vnd and sog, but we have relatively poor accuracy for rho. Upon further inspection we find the higher false positive rate for rho is primarily due to large concentrations of rho in the cytoplasm. These concentrated mRNA signals are also difficult, but not impossible, for human experts to discriminate, so we continue to look for ways to capture their expert knowledge during the learning process.

By collocating all the objects within an image stack, we construct a spatially oriented, relational model of nascent gene transcription. We are currently in the process of analyzing the gene activity patterns and relating them to biological questions such as measuring the stochastic variation in transcriptional activation in different nuclei in a developing tissue, and relating the levels of nascent transcription to the levels of cytoplasmic mRNA on a cell by cell basis.

This work is in collaboration with David Kosman and William McGinnis from the Section of Cell and Developmental Biology at UCSD.

« Back to Posters or Search Results