CRISPR-VAE: A Method for Explaining CRISPR/Cas12a Predictions, and an Efficiency-aware gRNA Sequence Generator

Abstract

Deep learning has shown great promise in the prediction of the gRNA efficiency, which helps optimize the engineered gRNAs, and thus has greatly improved the usage of CRISPR-Cas systems in genome editing. However, the black box prediction of deep learning methods does not provide adequate explanation to the factors that increase efficiency; rectifying this issue promotes the usage of CRISPR-Cas systems in numerous domains. We put forward a framework for interpreting gRNA efficiency prediction, dubbed CRISPR-VAE, that improves understanding the factors that increase gRNA efficiency, and apply it to CRISPR/Cas12a (formally known as CRISPR/Cpf1). We further lay out a semantic articulation of such factors into position-wise k-mer rules. The paradigm consists of building an efficiency-aware gRNA sequence generator trained on available real data, and using it to generate a large amount of synthetic sequences with favorable traits, upon which the explanation of the gRNA prediction is based. CRISPR-VAE can further be used as a standalone sequence generator, where the user has low-level control ability. The framework can be readily integrated with different CRISPR-Cas tools and datasets, and its efficacy is confirmed. The complete implementation of the methods can be found at github.com/AhmadObeid/CRISPR-VAE.