The Problem

Gene delivery needs a version update with the current modalities limited by a range of negative side effects. The last 50 developmental years of gene delivery technologies has seen iterative steps towards addressing these negative effects by making the agents compatible with humans. However, these modifications have so far not resolved the outstanding safety issues, especially upon initial dosing leading to antibodies that target these agents, which yields unwanted immunogenic responses as well as precluding necessary repeat dosing of RNA-based therapies. For example, in April 2024, Verve Therapeutics voluntarily halted a clinical trial using LNP delivery in the liver due to an abnormal increase in liver enzymes; and in June 2024, Pfizer halted its AAV gene therapy for a rare muscle disease due to a patient passing away in response to the delivery agent.

What if...

What if a machine learning platform could produce proteins able to deliver and redose RNAs anywhere in the human body without causing an immune reaction? We wanted to build that.

So What?

Our end-to-end workflow will revolutionize RNA delivery design, especially as the parts are modular, wherein the generation of a cargo, carrier, or targeting moieties can be mixed and matched with other delivery systems. In particular, a centralized design platform with key parameters for candidate selection will be a useful tool for both industry and researchers working in the gene therapy field where many of these components are sequestered, dramatically increasing time and resource costs; i.e. design of specific cell targeting is often done in isolation from therapeutic cargo and/or therapeutic carrier.

Contact

If interested in the code or video demo, please register your interest here.

Built With

• alphafold-v2
• alphafold-v3
• esm3-(open-and-large-models)
• nvidia/google-clould-compute
• python
• runpod