Prediction in protein folding has been named one of the breakthroughs of artificial intelligence in 2024. As if to confirm this, scientists today reported that they have developed AI-assisted antivenoms for the deadliest snake venoms, mambas, sea snakes and rattlesnakes. Deep machine learning in dialogue with researchers helped create proteins that protect the body from deadly toxins, which was confirmed by tests on mice. And this is just the beginning.
Image source: Kate Zvorykina/Ella Maru Studio
Today, antivenoms are created through a complex and expensive method of immunizing animals. Animals, such as horses, are injected with snake venom and wait for antibodies to appear. The antibodies are then extracted from the collected blood in a centrifuge, resulting in serum—essentially a drug. All this takes weeks and months. Machine learning promises to radically reduce the time it takes to synthesize proteins to neutralize toxins. And it will, of course, be cheaper and safer for the victims.
The study was conducted by the University of Washington School of Medicine (UW Medicine) in collaboration with colleagues from the University of California Medical Institute for Protein Engineering and the Technical University of Denmark. Looking ahead, we note that the university has filed a patent application for an antidote developed with the help of AI.
The problem of snake bites affects over 2 million people every year. According to the World Health Organization, more than 100,000 of them die, and 300,000 suffer serious complications and even become disabled as a result of limb deformities, amputations or other consequences. Sub-Saharan Africa, South Asia, Papua New Guinea and Latin America are among the places where venomous snake bites pose the greatest public health problem.
The research team focused on finding ways to neutralize the venom derived from certain elapids. Elapids are a large group of venomous snakes, including cobras and mambas, that live in the tropics and subtropics. The venom of this group includes the so-called three-finger toxin (3FTx). This chemical damages body tissue, killing cells, and also interrupts signals between nerves and muscles, causing paralysis and death.
AI helped create new proteins that interact with toxins and bind them, preventing them from reaching the receptors of living organism cells. By interacting with the program, scientists received instructions for the synthesis of thermostable proteins with a high ability to bind to toxins. The virtually synthesized proteins matched the deep learning computer design almost exactly at the atomic level.
In laboratory conditions, the developed proteins effectively neutralized all three subfamilies of three-finger toxins tested. When administered to mice, the engineered proteins protected the animals from what could have been fatal effects on them.
Engineered proteins have many benefits. They can be produced with consistent quality using recombinant DNA technologies rather than by immunizing animals. In addition, the new proteins developed against snake toxins are small in size compared to antibodies. Their smaller size can provide greater tissue penetration to quickly counteract toxins and reduce damage.
In addition to opening up new ways to develop antidotes, the researchers believe computer-aided design techniques could be used to develop other drugs. Such methods could also be used to find cures for so-called neglected diseases that afflict the poorest countries in the tropics, for which there is usually no money.