Modeling CRISPR-Cas13d on-target and off-target effects using machine learning approaches

We conducted a two-vector CRISPR/Cas13d proliferation screening experiment. The screening library contains 10,830 sgRNAs targeting 192 protein-coding genes and 234 lncRNAs, and the screening experiment was performed using a melanoma cell line A375. It provides a unique dataset to model Cas13d sgRNA efficiency and specificity. We designed a deep learning model, named DeepCas13, to predict the sgRNA on-target activity with high accuracy from sgRNA sequences and RNA secondary structures. The gRNA lentiviral library-infected cas13d-expressing A375 cells harvested in day 0 and day 28 were used for DNA extraction, gRNA enrichment by PCR amplification and deep sequencing.