RNA Editing Tool Predicts Target Performance With Pooled Screening, Machine Learning

In a paper appearing in Nature Biotechnology, a team from New York Genome Center, New York University, and Columbia University outline a deep learning and CRISPR-Cas13d-based RNA editing strategy for gauging the on-target gene expression effects of transcriptome engineering in human cells. With an approach dubbed "targeted inhibition of gene expression via gRNA design" (TIGER) that relies on CRISPR screening of pooled RNA targets, along with expression profiling and machine learning, the investigators assessed some 200,000 guide RNAs for CRISPR-Cas13d RNA editing on essential genes in human cell lines. With patterns detected for guide RNAs with mismatches or small insertions or deletions in the screen, they trained a machine learning model to predict on-target gene expression tweaks based on the sequence context and guide RNA selected, pointing to the possibility of calibrating gene expression in human cells. "Transcriptome engineering applications in living cells with RNA-targeting CRISPR effectors depend on accurate prediction of on-target activity and off-target avoidance," the authors write, noting that the current study suggests that "TIGER scoring combined with specific mismatches yields the first general framework to modulate transcript expression, enabling the use of RNA-targeting CRISPRs to precisely control gene dosage."