EternaBrain: Automated RNA design through move sets and strategies from an Internet-scale RNA videogame

Emerging RNA-based approaches to disease detection and gene therapy require RNA sequences that fold into specific base-pairing patterns, but computational algorithms generally remain inadequate for these secondary structure design tasks. The Eterna project has crowdsourced RNA design to human video game players in the form of puzzles that reach extraordinary difficulty. Here, we demonstrate that Eterna participants’ moves and strategies can be leveraged to improve automated computational RNA design. We present an eternamoves-large repository consisting of 1.8 million of player moves on 12 of the most-played Eterna puzzles as well as an eternamoves-select repository of 30,477 moves from the top 72 players on a select set of more advanced puzzles. On eternamoves-select, we present a multilayer convolutional neural network (CNN) EternaBrain that achieves test accuracies of 51% and 34% in base prediction and location prediction, respectively, suggesting that top players’ moves are partially stereotyped. Pipelining this CNN’s move predictions with single-action-playout (SAP) of six strategies compiled by human players solves 61 out of 100 independent puzzles in the Eterna100 benchmark. EternaBrain-SAP outperforms previously published RNA design algorithms and achieves similar or better performance than a newer generation of deep learning methods, while being largely orthogonal to these other methods. Our study provides useful lessons for future efforts to achieve human-competitive performance with automated RNA design algorithms.Author summary: The design of RNA sequences that fold into target structures is a computationally difficult task whose importance continues to grow with the advent of RNA-based therapeutics and diagnostics. This paper reports a new approach stemming from the Eterna massive open laboratory, a project that crowdsources RNA design to >250,000 ‘players’ on the internet. The efforts of Eterna participants have led to the accumulation of nearly 2 million moves that lead to successful in silico solutions on difficult puzzles, many of which are only solvable by humans. Inspired by recent advances in automated game playing, we discovered that these moves are sufficiently stereotyped so that a neural network can be trained to predict moves with accuracy significantly higher than random guessing. The resulting method EternaBrain allows solution of new RNA design problems when used to predict complete series of moves rather than just single moves. Further improvement comes from heuristic strategies that are well known amongst the Eterna community but not described in prior publications on automated RNA design. EternaBrain appears highly complementary to other emerging next-generation RNA design methods based on neural-network and game playing approaches, suggesting new routes for automated methods to emulate human experts in RNA design.