NEW YORK (GenomeWeb) – A new study from Stanford University suggests that CRISPR/Cas9- and short hairpin RNA-based genetic screens yield comparable results — and with a new informatics tool developed by the study's authors, combined screening results can outperform either method on its own.
Led by Stanford Professor Michael Bassik and first author David Morgens, the scientists ran both short hairpin RNA (shRNA) knock-down and CRISPR/Cas9 knock-out screens to identify essential genes in a human cancer cell line for which 217 genes had already been deemed essential and 947 which had been deemed nonessential.
"Both shRNA and CRISPR/Cas9 screens had very high performance in the detection of essential genes," the authors wrote in a study published today in Nature Biotechnology. Each method found more than 60 percent of the "gold standard" set of essential genes; however, the methods found sets of essential genes that the other did not and both suffered from false positives not found in the gold-standard set. The gold-standard essential genes came from a 2014 study led by University of Toronto professor Jason Moffat.
But the authors suggested that by combining the data sets using a newly described statistical framework, they could find a greater percentage of essential genes and reduce the false-positive rate.
In recent years, CRISPR screening has exploded in popularity. As reported by GenomeWeb, firms have seen demand for CRISPR screening equal or eclispse demand for RNAi screens.
The new study from Bassik's team suggests that shRNA screening can pick up on things that CRISPR/Cas9 can't, and that combining the results from the two methods could reveal information that neither could on its own.
The authors noted the different methods identified different biological categories of genes. For example, the CRISPR/Cas9 screen found essential genes in the respiratory chain, while the shRNA screen found essential genes in the chaperonin-containing T-complex. This could be explained simply by the lack of effective reagents towards a particular gene or by off-target effects, but it could also be true that for certain genes, "a small loss in gene product via knockdown leads to a completely different phenotype than a large loss via knockout," the authors wrote.
To make more sense of the screening data, the researchers developed an empirical Bayesian framework for statistical analysis that incorporates data from both shRNA and CRISPR/Cas9 screens, which they dubbed the Cas9 high-Throughput maximum Likelihood Estimator (casTLE).
"For each gene, casTLE combines measurements from multiple targeting reagents to estimate a maximum effect size as well as a P value associated with that effect," the authors explained. They validated casTLE on previous data sets from RNAi, CRISPR deletion, and CRISPR interference and activation screens.
The authors reported that by combining the data from their new shRNA and CRISPR/Cas9 screens, they recovered more than 85 percent of "gold-standard" essential genes with a false-positive rate of approximately 1 percent.
The study also noted that while heterogeneity has been well-established in results from shRNA libraries, it was present to some degree in CRISPR/Cas9 screening, perhaps because Cas9 can frequently introduce in-frame indels rather than a true knock-out.
The software to run casTLE is available for free download from Morgen's Bitbucket page.