Ultra-high throughput and machine learning methods for predictive gene circuit design

Abstract: Genetic circuits represent a promising means of precisely controlling cellular behaviors, yet their design has proven challenging. Traditionally, circuit engineering is an iterative process in which a suitable composition of genetic components is arrived at over many rounds of optimization, sometimes taking longer than the practical lifetime of the project. CLASSIC is a high-throughput platform that combines pooled DNA assembly with short- and long-read next-generation sequencing to quantitatively characterize libraries of constructs at length scales spanning entire multi-gene circuits and expression unit arrays (10-20 kb). As shown in their recent manuscript, CLASSIC can be used to collect experimental measurements from over 100,000 circuit designs in a single experiment, allowing training of machine learning models and highly accurate, predictive profiling of composition-to-function relationships for vast genetic circuit design spaces. Using explainable AI and statistical inference methods, CLASSIC enables elucidation of design rules for multi-input digital logic behavior in mammalian cells. Bios: Ronan O’Connell is a postdoctoral researcher in Steven Boeynaem’s lab at Baylor College of Medicine in Houston, Texas. He completed his PhD in Bioengineering in Caleb Bashor’s group at Rice University, where he worked with Kshitij Rai to co-develop CLASSIC, a high-throughput platform for characterizing large and complex synthetic genetic programs in cells. In the Boeynaems lab, Ronan is using high-throughput experimental approaches to unravel how disordered proteins execute their diverse cellular roles despite lacking a single fixed three-dimensional structure. By mapping these sequence-function relationships at scale, we aim to harness protein disorder as a predictable and tunable structural feature. Kshitij Rai is a postdoctoral scholar in Jay Shendure's Lab at the Department of Genome Sciences, University of Washington. Having graduated with a PhD in Synthetic, Systems and Physical Biology from Rice University, Kshitij's research interests span a wide range of methods and tool development for predictive design of synthetic gene circuits, primarily in mammalian cells for synthetic biology applications. Kshitij's work in graduate school involved the development of a platform technology called CLASSIC along with Ronan O'Connell, a fellow graduate student in the lab and co-author on the work. In the Shendure lab, Kshitij's work is focused on understanding the effects of dynamic chromatin remodelling during stem cell differentiation on synthetic transgenes and genetic circuits, using data driven approaches to model how cellular context drives epigenetic silencing.