A quick rundown of how the various projects with which I’m involved are going:
- I’ve just completed (and tested) our analysis pipeline for classifying foci for the morphology project. This is for those markers with phenotypes we are expecting to find, and that we treat as a supervised learning problem. So from a (relatively small) hand assembled training set, we can score all the objects in the screen relatively quickly. I’ve never worked on a project that generates data of this magnitude before (hundreds of gigs per screen, around 30 screens, some in multiple replicates!). What still remains is to figure out how to handle the unsupervised case: screens that will contain morphological mutants that we don’t know anything about a priori, and will thus have to find in some way. In collaboration with our amazing system admin, Jeff Liu, and helpful PIs Charlie & Zhaolei, we have cleared some data storage obstacles. Finally, with help from folks from the Broad Institute, we’re on the verge of working out some image analysis problems.
- I’ve been stalled on the mass spectrometer noise model project for some time now. This is mostly due to my limited availability. I’ve had more pressing demands on my time (course work, thesis drafts/defense, other projects). I’ve been sitting on some promising ideas for this project for a while now, and am excited to start work on it soon.
- I’ve agreed to play a minor role in Renqiang and Jingjing’s Bayesian modeling of microRNA mechanism project. Simply put, they want to change the method used to infer the parameters of their model from Gibbs sampling (slow, not so exciting) to some variational inference techniques (newer, more exciting). The model and results have already been published in JCB (see Renqiang’s website for a preprint here)