Our lab develops scalable algorithms for regulatory genomics, i.e., how do various factors regulate gene expression. Here, we feel the data space is complex and multi-dimensional, apt to leverage the power of deep learning algorithms and their variants. To make sense of this data, we have been using both simpler, interpretable algorithms and more complex, expressive algorithms for mapping the interactome of genes with the goal being to map out the gene regulatory landscape. This has applications both in precision nucleotide-based therapeutics and in genome editing to characterize gene functions and for crop improvement.
As part of this we are mapping out the “language of the genome” using somewhat unconventional tools in this sub-domain such as natural language processing (NLP) and variational autoencoders to decipher the underlying wirings of the genome and discover the latent space that controls a lot of the interactions of the genome.
Current projects: