The new AI tool captures the strategies of the best players in RNA video games
RNA molecules fulfill basic biological functions. In recent years there has been a strong interest in developing new RNA structures for the treatment of CRISPR for cancer treatment.
However, each RNA structure consists of a sequence of four blocks long, and determining the exact sequence needed to build a structure can be difficult to calculate.
In the new study, Koodli, Das, and colleagues conducted research on Eterna internet-based video games, a civil initiative to address computer problems in RNA design.
Eterna presents each player with the target RNA structure and the player tries to find an RNA sequence that allows the finished molecule to fold into the desired shape. Some players outperform the best computerized automatic methods to meet this challenge.
Using 1.8 million design decisions made by Eterna players, the researchers found an artificial neural network that captured some of the preferences and strategies of these experts.
This approach, called EternaBrain, can predict the selection of the best players with far greater accuracy than achieving random knowledge.
The advanced EternaBrain algorithm provides performance that is equal or better than the algorithm developed previously to solve Eterna challenges.