Machine learning for siRNA efficiency prediction: A systematic review

Abstract

Therapeutic applications of small interfering RNAs (siRNAs) have recently facilitated advancements in the biopharmaceutical industry, expanding opportunities for pharmacological intervention to targets previously deemed “undruggable.” Hence, determining rational design principles to inform the selection of effective siRNA sequences and appropriate chemical modifications has been a significant undertaking in the field. To accelerate the process of empirical siRNA design, machine learning (ML) techniques have been applied to the problem of siRNA efficacy prediction. This systematic review provides a comprehensive, yet succinct overview of advancements in this ML task by examining the evolution of model architectures trained to predict siRNA efficacy, features selected to represent individual samples and inform predictions, and the challenges associated with the use of ML in the context of therapeutic siRNA discovery. Consensus and conflict throughout the literature are discussed, promoting a nuanced understanding of this problem. Finally, the vast potential for future directions is addressed, supporting further research in computational biomedicine.