The ability to computationally create single-chain antibody fragments and other antibody-like proteins circumvents many problems encountered in traditional monoclonal antibody creation in animals, leading to more effective biosensors and clinical antibodies.

The capability to select the analyte binding affinity of a designed protein is critical for the development of biosensors, in which the binding affinity must be matched to the dynamic range of the analyte as opposed to being merely as tight as possible. While computational design often focuses on creating the tightest possible binding, information collected in this dataset about mutations that alter binding affinity in non-optimized sequences will provide information often lacking in published datasets. Furthermore this knowledge will enable adjustments when external design constraints impose non-favorable mutations, such as with the removal of disulfide linkages in our phase-changing scFvs.

Our initial target will be hIL-6, a clinical biomarker for a number of important diseases, including ovarian cancer, lung cancer, and cytokine storm in sepsis, covid, and Car-T syndrome in patients undergoing cancer immunotherapy. Increased affinity to the cytokine target will improve sensitivity and expand dynamic range in our sensor platform. In addition to hIL-6, we intend to expand on this approach to target several additional cytokines, eventually enabling the design of high-affinity scFvs to cytokine targets outside of our dataset. We will create datasets that will enable the development of new AI approaches to predictively engineer scFvs and other loop-presenting proteins as sensing elements for a diverse set of cytokines relevant to human health.