Research
My graduate research in the Levine Lab and postdoctoral work in the Bussemaker Lab have focused on the computational modeling of gene expression regulation, with an emphasis on the inference of accurate sequence-to-affinity models from high-throughput protein-DNA interaction data, specifically PBM and SELEX-seq data (the latter of which I co-developed).
A significant challenge is that the binding data gathered by these two assays are drastically different from each other. PBM (and other microarray) assays produce real-valued and roughly normally distributed protein-occupancy values, while SELEX-seq (and other high-throughput sequencing) assays produce integer-valued, poisson-distributed sequence read counts. To accommodate these two disparate and noisy data types, I use robust linear and non-linear regression techniques, and poisson regression techniques, respectively, to infer the accurate sequence-to-affinity models.
My work has resulted in the following specific advances:
- Development of an algorithm, FeatureREDUCE, that emerged as the top performer in a recent benchmark study comparing to all known PBM analysis algorithms (26 in total).
- Discovery that binding by a common cofactor evokes latent differences in DNA binding specificity between the developmental HOX proteins - the result of a recent collaboration.