A new computational framework that harnesses the power of machine learning could provide actionable insight to help prevent contact model failures and revolutionise safe freezing of vital medicines and vaccines, say scientists. This innovative study has supported for the identification of a new cryoprotectant molecule that can prevent ice crystal growth during freezing and thawing, an achilles heels in cryopreservation.
Realizing the Potential of Cryopreservation
Rapid freezing is essential for the preservation of many life-saving treatments, including vaccines and fertility materials as well blood donations, and cancer therapies. The researchers call these molecules — which are used in a process called “freezing” to help the treatments work — “cryoprotectants.
Or these therapies would need to be used in a limited and unforesightful fashion or, even worse, not available for immediate use in the future at all due to the lack of cryopreservation. Nevertheless, the present trial-and-error approaches in identifying new cryoprotectants are very expensive and time-consuming which is a bottleneck for exploration in this area.
Experts from the University of Warwick and the University of Manchester have managed to come up with an innovative solution by making use of machine learning. Their new framing device for the code should enable hundreds of molecules to be tested by computer, reducing the discovery process and opening the way towards better cryoprotectants.
Combining Molecular Simulations and Machine Learning
You cannot just take a dataset, sprinkle it with a machine learning algorithm and interpret the output as science —”Prof Gabriele Sosso, from Warwick, who led the research. This particular example suggested use of it by a team as part of their suite of tools, and was successful because, among other things, it neatly complemented molecular simulations and most crucially could be used in the context of well-done experiments.
With a computer model, the researchers guided through massive libraries of chemical compounds to identify those that would make for the best cryoprotectants. For the team, this data-driven method is a huge leap from conventional trial-and-error strategies, which allows the group to concentrate on regions where human imagination and experience are still needed.
Dr. Matt Warren, the Ph. MIT D. student, who led the project, says its machine learning-powered technique has promise He noted that “for years, we’ve labored long hours to collect data in the lab, and it’s both exhilarating and encouraging to finally have a machine learning model it at our disposal for data-driven predictions of cryoprotective activity.
Conclusion
The search for a new type of cryoprotectant molecule that inhibits the formation of ice crystals during freezing and thawing is thus an important advancement in the field of cryopreservation. This advance could be important for identifying new cryoprotectants more quickly, and it might help to repurpose molecules that are already known to slow or halt ice growth. Using machine learning, the researchers have established a new paradigm for discovering cryoprotectants that could produce more powerful and broadly applicable life-saving therapies.
