We’ve worked really hard these last couple of months to finalize some of our newest learning algorithms to address various problems in synthetic biology, Koopman operator theory, and systems biology. We submitted 5 papers to the ACC, fingers crossed!~
Here are five taglines to describe the five papers (as well as the name of the lead author of the work):
1 – A first-ever data-driven framework for programming the steady state of large transcriptional gene networks from time-series RNAseq data. (Hasnain, 2019)
2 – The first distributed Koopman learning algorithm, along with proofs of convergence for deepDMD. Our goal is to build data-driven models on systems with thousands of genes, eventually (Liu, 2019).
3 – The first phase-space stitching algorithm for assembling “local” estimates of Koopman operators into global Koopman operators. This result is useful when you have to fuse results of different experiments from different physiological conditions together (Pushpak, 2019).
4 – A new online Koopman learning algorithm for updating Koopman operators in real-time with streaming data (Sinha, 2019). These results are very useful if you’re doing continuous-sampling or fully automated biological experiments, like the ones run at Strateos/Ginkgo/Amyris.
5 – An extension of the Hankel-DMD algorithm, that addresses the acausality of Hankel-DMD, with a modification (causal-jump DMD) that now works with extended dynamic mode decomposition. We show it’s ability to predict growth curves of new bacterial strains in the absence of metabolic models, using only physiological conditions (Balakrishnan, 2019).