Wound healing is a systems-level program governed by complex inter and intracellular networks. After wounding, receptor activation causes intracellular signaling cascades inducing downstream transcriptional changes specific to the dose and identity of the ligand. Although these networks have been rigorously studied, establishing a direct mapping between signaling and gene expression is still an open question. I test the hypothesis that transcription factor dynamics carry significant, stimulus-specific information for downstream transcription by combining live cell imaging, high-throughput transcriptomics (MERFISH), and deep learning-based variational inference. Coupling measurements of NF-kB dynamics with MERFISH will produce a rich, high dimensional data set that cannot be fully understood with traditional approaches. Deep networks trained to perform variational inference can also estimate the mutual information between high dimensional vectors. By combining these novel approaches, I will quantify the contribution of NF-kB dynamics to the transcriptional encoding of environmental information.
Authors: Evan Maltz, Robert Foreman, Adewunmi Adelaja, Alexander Hoffmann, Roy Wollman