CHEE Seminar: Ramya Kumar
Monday, December 4, 2023 – 3:00 p.m.
Ramya Kumar, PhD
Assistant Professor of Chemical Engineering
Colorado School of Mines
“Causal Machine Learning Reveals Payload Specific Polymer Design Criteria for pDNA and RNP Delivery”
Speech & Hearing Building, Room 205
Social Hour immediately following the seminar in Old Engineering 157 (Graduate Student Lounge) at 4:00 p.m.
ABSTRACT
To improve the affordability, safety and accessibility of therapeutic nucleic acids, we need to replace engineered viral vectors with synthetic materials. Owing to their exquisite tunability and precision, polymers have emerged as promising class of biomaterials that can resolve challenges impeding the clinical translation of gene editing platforms. However, our ability to rapidly realize targeted polymer properties for specific therapeutic goals is handicapped by intractably vast polymer design spaces. Consequently, polymeric vector development has been plagued by inefficient trial-and-error based approaches and low discovery rates. Here, I will present an information-driven workflow for polymeric vector discovery that (1) accelerated the discovery of a highly efficient polymeric vehicle for genome editing payloads, that outperformed four state-of-the-art commercial transfection reagent (2) identified payload-specific structure-function relationships correlating polymer attributes to cellular toxicity, editing efficiency and payload uptake, informing the synthesis of subsequent polymer libraries. To reveal the physicochemical drivers of gene delivery performance, SHapley Additive exPlanations (SHAP) were computed for nine polyplex features, and a causal model evaluated the average treatment effect of the most important features selected by SHAP. Our machine learning interpretability and causal inference approach derives structure–function relationships underlying delivery efficiency, polyplex uptake, and cellular viability and probes the overlap in polymer design criteria between RNP and pDNA payloads. While pDNA delivery demands careful tuning of polycation protonation equilibria while RNP payloads are delivered most efficaciously by polymers that deprotonate cooperatively via hydrophobic interactions. These payload-specific design guidelines will inform further design of bespoke polymers for specific therapeutic contexts. The union of combinatorial polymer design, parallelized experimentation, and machine learning can be leveraged to establish a powerful polymer discovery pipeline. The workflow discussed herein demonstrates the possibility of translating statistically derived polymer design principles to therapeutically useful materials.
BIOSKETCH
Dr. Ramya Kumar obtained her BE (Hons.) in chemical engineering from BITS Pilani, India, and her Ph.D. in chemical engineering at the University of Michigan, Ann Arbor. At Michigan, she received a Rackham Predoctoral fellowship, the Procter & Gamble Team Innovation award, and the Richard & Eleanor Towner Prize for creative and innovative teaching. Kumar is also an ACS PMSE Future Faculty awardee. She completed her postdoctoral training at the University of Minnesota, Twin Cities where she developed materiomics workflows to accelerate polymeric vector. In January 2022, Kumar began her independent career as an assistant professor in the Department of Chemical Engineering at the Colorado School of Mines. Her lab applies controlled radical polymerization, surface-initiated polymerization, and statistical modeling to develop novel biomaterials. At Mines, her group’s research is funded by the National Institutes of Health through an R21 award and by the National Science Foundation (through the GRF).