Developability has been recognized as a crucial driver for clinical success of antibodies. The traditional “function first, developability second” screening paradigm is sequential and not optimal, which frequently results in less developable candidates that require additional engineering.
We advocate a parallel approach where developability is considered alongside function when selecting candidates. This is made possible by our in silico developability assessment platform, with fast turnaround for minimal cost.
· Comprehensive suite of computational models, including both physics-based and AI-based
· AI models trained on thousands of internal data points; achieved state-of-the-art (SOTA) performance on multiple developability prediction tasks
· Thoroughly validated and optimized in numerous internal projects
· Quick turnaround (minimum of 1 week); only antibody sequences needed
· High-throughput (HTP) version for fast evaluation of up to 1000 candidates
· Low-throughput (LTP) version for detailed, fine-grained analysis of selected candidates
36 hits that had passed binding and functional screening were analyzed by XcelDevTM Silico. 6 properties of each antibody were predicted and scored on a scale of 0-1 (displayed in color gradient shown below).
An overall developability score was calculated to rank all 36 hits. To verify the effectiveness of the ranking, the top 7 (in the red box) were expressed and subject to a battery of developability assays. All 7 performed well in these assays, their Tm, SEC, AC-SINS and HIC results shown below.