Feasibility of Using a High-speed AI-driven Experimental Platform to Develop an Optimized DNA Template for Cell-free Protein Expression

We propose to demonstrate the feasibility of using a high-speed artificial intelligence (AI)-driven experimental platformto optimize the DNA template for cell-free protein synthesis (CFPS) to maximize protein expression. Our platform isdesigned to rapidly find an optimal DNA sequence among millions of possible sequences through repeated designbuild-test cycles: AI is used to design the DNA template for each consecutive iteration based on the results fromprevious cycles, CFPS is used to build the protein from the DNA template, and testing is used to determine theprotein yield. Using published datasets, we will run simulations to show that our platform can find optimal sequencesamong >250,000 possibilities based on experimental testing of only about 1% of those sequences. As a proof-ofconceptcase, we plan to optimize a ~30-nucleotide expression cassette region encompassing the sequences fromthe start of the ribosomal binding site through codon 5 of the protein coding sequence. It has been historicallyintractable to optimize cassette regions over ~10-nt because the search space if simply too large for comprehensiveexperimental exploration. Our proposal aims to demonstrate the feasibility of optimizing larger cassette and codingregions through tight coupling of computational and experimental evaluation.