Hello, i am working on validating Drug-target screens to validate the hypothesis that the drug is inhibiting the gene product by examining that the gene ko and the drug have similar signatures (effects). however, i am lost to which dataset is more appropriate for building the correlation model between the two profiles as the CRISPRGeneEffect doesn’t contain the LogFold change in contrast to PRISM repurposing dataset.
We do share a raw log fold change per CRISPR guide, but that data is not ideal to use directly, as there are multiple known experimental artifacts (CN correction and guide efficacy being dominate ones of the top of my head) which should be corrected for to understand the effect of knocking out a gene.
As a result, the CRISPR gene effect scores are corrected via fitting the model used by Chronos and likely not completely equivalent to log fold changes. However, the spirit of the Gene Effect score is that it’s intended to be an interpretable score to capture amount of killing observed in a similar way as a log fold change measurement would.
In fact, in the portal we regularly look for evidence of drugs phenocopying gene knockouts by looking for high correlation between log fold change in drug screens and gene effect scores from CRISPR screens.
thanks,
Phil
thank you for the clarification, is there something similar to chronos but for drugs, like AUC\IC50 of drugs across different cell lines
Hi, sorry for interrupting, but how about checking the GDSC dataset?
Hi Song, thank you for the recommendation, but the main bottleneck with the dataset is the limited number of drugs. How about the sanger cell viability, is it comparable to the chronos score for genes?
Sorry, but can you clarify the ‘Sanger cell viability’ with the relevant link? In my opinion, the Chronos score and IC50 (or AUC) values will be difficult to compare directly, as they are computed in different ways and have distinct meanings.
it is called sanger-viability.csv with the following download link:
https://depmap.org/portal/data_page/?tab=allData&releasename=Sanger%20GDSC1%20and%20GDSC2&filename=sanger-viability.csv
wouldn’t the Chronos and IC50 (or AUC50) be correlated in drug-target pair with similar transcriptomic profile ?
