Experimental exploration of a ribozyme neutral network using evolutionary algorithm and deep learning.

A neutral network connects functional genotypes within a fitness landscape and provides accessible paths for evolution. RNA neutral networks have been extensive studied theoretically, but no empirical evidence of such a network has been found. In this work, we use a deep learning-guided evolutionary algorithm to explore the sequence space of an RNA ligase ribozyme. High-throughput screening was used to evaluate the activities of ribozyme mutants over eight generations. We discover that one of the mutants identified that differs from the wildtype by 16 mutations is connected with the wildtype by an extensive network of neutral paths.