ATUM's HuAbGPS design tool begins by grafting heavy and light chain CDRs onto human germline frameworks (step 1 below). We then identify amino acid differences between the original (murine) framework and the new human framework. To find the murine framework amino acids essential for proper presentation of the CDRs within the human framework context, we create a DoE matrix where murine framework amino acids are systematically tested in different combinations. The result of this antibody engineering process is a functional antibody containing the minimal number of murine framework amino acids whose heavy and light chain sequences are closer to human than to any other species.

ATUM's AntibodyGPS® design platform enables the simultaneous optimization of different critical antibody attributes including affinity, aggregation, stability and expression.

First, potential amino acid changes within the CDRs are ranked based on evolutionary conservation and physico-chemical properties. Up to 96 variant antibodies are then designed to incorporate the most highly ranked substitutions so that each substitution is represented in multiple variants.

The antibody variants are then expressed, and their properties are measured. Machine learning is used to analyze the data and determine the effect of each amino acid substitution on each of the measured properties as well as quantifying the epistatic effect on each residue. The graph below shows the effect of 72 different changes in the affinity, thermostability and expression level (titer). Very few changes have a positive effect on all three desirable antibody attributes, which is a severe limitation in traditional library screening approaches. Because ATUM's GPS technology allows us to identify changes that separately improve affinity, titer and thermostability, we are able to combine amino acid substitutions to produce an antibody that is improved for all three, as shown in the graph below.