Asimov is building a system at a Boston University laboratory to test whether a therapeutic protein can be manufactured
Asimov is building a system at a Boston University laboratory to test whether a therapeutic protein can be manufactured
On July 24, Asimov announced that it will design therapeutic protein variants computationally, produce them at Boston University's DAMP laboratory, and measure the properties of the finished product. The company plans to feed these results back into the model that selects the next sequence.
A therapeutic protein begins with an amino acid sequence, but the work does not end there. The protein must still be produced in cells, purified, and tested. This stage sometimes reveals that a design that performed well computationally is difficult to turn into a product.
In its July 24 announcement, Asimov describes the process: the company designs protein variants, the laboratory produces them and measures properties related to manufacturability, and the model receives the results. The result of each experiment will provide data for selecting the next batch of variants. This allows protein manufacturability to inform sequence selection before the next experimental cycle begins.
DAMP is Boston University's Design, Automation, Manufacturing, and Processes Laboratory. It already performs remotely commissioned experiments involving biological materials, chemicals, and liquids. Asimov will establish a facility there for work on therapeutic proteins and will become one of the laboratory's first users.
The BU laboratory is part of the Programmable Cloud Laboratory Test Bed network. On July 22, the U.S. National Science Foundation awarded the network 380 million dollars over four years to support 20 laboratory nodes; Astera Institute is contributing up to 20 million dollars. BU will receive up to 20 million dollars for its node. The network is intended to let laboratories share experimental methods and results.
Asimov is connecting protein design with what happens when a protein is actually manufactured. A protein that can be produced and measured gives the model an example to inform its next selection; a protein with poor manufacturing properties helps the model reject similar sequences earlier.