Biophysical protein ladders standards for benchmarking high-throughput studies

Proteins are specialized and efficient nanomachines that are essential to life and useful in biotechnology. The “parts lists” for these nanomachines are linearly encoded in DNA: Each protein comprises ~100-1000 components called “amino acids”; in the linear sequence of components, each position is occupied by one of twenty possible amino acids; and the linear sequence of amino acids folds into a three-dimensional structure that dictates a protein’s function. As such, changes in a protein’s DNA code can change which amino acid is present at a given protein position. Changes in some amino acids are catastrophic, breaking the protein machine, whereas many other changes are less damaging and instead modulate protein function. These modulating changes can be leveraged by protein engineers to create proteins with bespoke functions for biotechnology. However, although many amazing tools are available to modify protein sequences, the ability to reliably choose which protein changes produce desired functional effect(s) remains tantalizingly out of reach. The goal of this research is to solve this challenge. This project will advance NSF’s priorities in Artificial Intelligence and Biotechnology. Developing new proteins with predictable functions faces three key challenges. First, it is impossible to experimentally measure all possible amino acid changes in all possible proteins: For an example protein with 100 amino acids, thousands of single and millions of combinatorial changes are possible and, to date, ~203 million unique protein sequences have been deposited in databases. Second, despite decades of effort, current computational algorithms cannot accurately predict functional consequences for many single amino acid changes, much less the complex changes that arise from combinations of changes. Finally, current protein engineering approaches, such as directed evolution, do not provide the comprehensive information needed to develop improved computational algorithms. To leverage advances in AI for improving computer-based predictions, large and high-quality protein datasets are required for training and validating models. To that end, the Cellular Engineering Group at the National Institutes of Standards and Technology is developing methods and protocols for generating robot-assisted, high-quality, high-throughput datasets that accurately report outcomes for millions of variants for representative model proteins. To demonstrate that these datasets are reliable enough for developing AI computer models, each dataset must be calibrated using data generated from traditional biochemical and biophysical methods. However, using traditional methods to generate a set of even 10-20 protein calibration variants remains a time and resource bottleneck. The goal of this proposal is to develop and validate new and faster purification and functional assay protocols for generating the protein ladders containing partial loss-of-function variants needed to calibrate high-throughput results. In the process, this project will enable improved use of Artificial Intelligence algorithms for engineering proteins important to Biotechnology, with improved accuracy and efficiencies relative to current practices. This award is funded by a collaboration between NSF, the National Institute of Standards and Technology, and Schmidt Sciences. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2623040 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT,4082CYXXDB NSF TRUST FUND | Principal Investigator: Liskin Swint-Kruse | Institution: University of Kansas Medical Center Research Institute Inc, KANSAS CITY, KS | Award Amount: $349,436 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2623040 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2623040.html