AI Drug Safety Screening Cuts Costs by 43%

MindWalk Holdings Corp. has published a white paper revealing significant cost and performance improvements for its LensAI™ immunogenicity screening. The study demonstrates that running LensAI on Amazon EC2 Hpc8a instances, powered by 5th Gen AMD EPYC processors, resulted in up to a 43% lower cost per sequence compared to Intel configurations. This advancement addresses a critical bottleneck in AI-driven drug discovery, where the rapid generation of candidate molecules previously outpaced safety screening capabilities.

The enhanced efficiency means that crucial immunogenicity assessments, which identify potential patient immune responses to therapeutic antibodies, can now be conducted practically at scale. This early-stage screening is vital for flagging high-risk candidates before significant investment is made, thereby preventing expensive late-stage program failures. The benchmark results show a public antibody database being screened in under five hours.

This performance leap is attributed to the architecture of LensAI and the capabilities of AMD EPYC processors. The software's design allows it to leverage high core counts effectively and benefits from specific processor instructions. MindWalk's ongoing engineering collaboration with AMD has focused on optimizing both AMD's accelerators and server processors for their Bio-Native AI applications, ensuring scalability without code rewrites.

The cost reductions and speed increases achieved with LensAI are not confined to immunogenicity screening alone. These benefits are expected to extend to other LensAI workflows running on MindWalk's ReefIQ™ biological context layer. This enhanced efficiency directly impacts MindWalk's margins as enterprise agreements shift towards recurring revenue, enabling the company to absorb greater workflow volume without a proportional rise in compute costs.