Receptor.AI and Sethera Will Use Each Experiment’s Results to Select the Next Peptide Series
Receptor.AI and Sethera Will Use Each Experiment’s Results to Select the Next Peptide Series
On 3 August, Receptor.AI and Sethera announced a research partnership to discover drugs against challenging therapeutic targets. Sethera will generate peptide screening results in the laboratory, and Receptor.AI will use those data to select the next series of molecules for synthesis and testing.
A peptide is a short chain of amino acids. These molecules can bind protein targets that offer few suitable interaction sites for small molecules. To become a drug candidate, a peptide must remain bound to its intended target, retain its shape, and sometimes cross the cell membrane. Chemists therefore optimize both the amino acid sequence and the shape of the complete molecule.
Sethera creates encoded libraries. The tag makes it possible to identify the sequence of each peptide that produces a screening hit. One or more chemical bridges close the chain into a ring and determine its shape. The laboratory selects variants against a chosen target and collects data on their sequences, structures, frequency after selection, and activity.
Under the program announced by the companies, Receptor.AI will combine these data with physics based modeling, develop hypotheses about binding, and rank candidate series. The partners will synthesize the selected series, test it in the laboratory, and feed the results into the next computational round. In this process, the model determines which molecules receive the time, reagents, and capacity available for the next experiment.
In July, PeptAI reported that its agent had changed the subsequent peptide search after binding was measured in the laboratory. Receptor.AI and Sethera will begin with one undisclosed target that is difficult to address through drug development. The companies will compare this cycle with conventional selection based on screening and assay results. They will examine whether the initial hits are confirmed, whether the candidates are selective for the target, and whether candidate series advance. They will assess the AI by whether it helps select a more useful next experiment.
In each cycle, the library supplies many physical variants, the laboratory measures their properties, and the computational analysis proposes the next series. The partnership links the algorithm’s decision to a result that can be tested in vitro.