Past Project: Drosophila Image Processing
In spring 2016, I stepped out of my condensed matter comfort zone and into the world of systems biology. I worked in an interdisciplinary group at Clark which includes members of the Math department, Biology department, the BioInformatics and Computational Biology Program, and the Physics department (i.e. me). One of the group's major goals is to develop a thermodynamic model for gene experession in everyone's favorite fruit fly, Drosophila. Drosophila have a very small and well-documented genome and are easy to breed in the large quantities necessary for fitting parameters with little uncertainty, making them an ideal organism for this type of research.

My role was to automate the process extracting spatial and temporal gene expression data from images of stained Drosophila embryos for the purpose of tuning the model's parameters. I developed a short macro in the NIH's Java-based image processing software ImageJ to perform the initial processing steps, including the creation of a binary mask for each image, a task that before was done by hand representing a severe bottleneck in the analysis. I also expanded and refined the MATLAB pipeline used for the rest of the analysis, including image alignment and, most importantly, background subtraction. The particular gene stained for in the images below is expressed as a set of stripes along the DV (dorsal-ventral) axis of the embryo.
There is a nonzero background signal in areas where we expect the gene to have no expression (left half of the embryo in the image on the left). By fitting this background signal to an ellipsoid, we are able to eliminate it allowing careful spatial quanitification of the gene expression in the expressing region.
