Every year 100,000 people worldwide die from venomous snakebites, and three times as many suffer permanent disabilities. A team at the University of Washington has just published groundbreaking work using deep learning to assist in the design of new snake antivenoms, by designing proteins de novo to bind neurotoxins. It is exciting to see a real-life example of artificial intelligence (AI) being used to identify promising therapeutic candidates. Although AI has tremendous potential for drug design, there are so far few examples of proven success, in contrast to the many successful applications of AI for analysing medical imaging data and for disease diagnosis. In the present research the highest affinity toxin-binders provided up to 100% protection from otherwise lethal doses of toxins in vivo.
Interestingly, the authors have filed a provisional patent application for both the “design and composition” of the proteins created. In a previous edition of M&C’s annual AI Report, our analysis revealed that the number of published patent applications relating to AI-based drug discovery was at the time surprisingly low. We speculated that this might have been due to applicants choosing to protect only the compounds per se, in order to keep details of the underlying discovery platform confidential. This can be an effective strategy given the finite term of patent protection and as such platforms are prone to frequent changes over time.
In the present case, however, there is enormous variation in the composition of different snake venoms, even within members of the same species. In the US, for example, Mohave rattlesnakes carry types of toxin with ten-fold differences in lethality. In cross-treatment experiments, the neutralising effect of the top candidate molecules in the paper was shown to be completely specific to the target toxin used in the design phase, so it would be impractical to patent all of the potentially effective candidate molecules that could be generated by the approach. The design of new antivenoms is therefore a good example of when complementary patent protection for the method or platform used to identify the candidates might be the only practical approach. Such protection might become even more important in the future, as medicine advances from a one-size-fits-all approach to the design and delivery of more personalised treatments, so that each patient might receive a different therapeutic agent.
诺贝尔奖得主大卫·贝克和蒂莫西·帕特里克·詹金斯领导的一项开创性研究展示了创新的计算机设计蛋白,这些蛋白能够中和致命的蛇毒,带来了更安全、高效且成本低廉的治疗新可能。 A groundbreaking study led by Nobel Laureate David Baker and Timothy Patrick Jenkins introduces innovative, computationally designed proteins that can neutralize lethal snake venom toxins, offering potential for safer, more effective, and cost-efficient treatments.