A question about differences between Top Co-dependencies and Predictability data on CRISPR

Q&A
1

Hi DepMap team,

I have a question about differences between Top Co-dependencies and Predictability data on CRISPR.

For example, NFE2L2, there are CRISPR Top Co-dependencies results such as SLC33A1, TAPT1, ADPGK…
However, CRISPR Predictability showed different results such as AKR1C1, TRIM16L…

I think the results are from same database, CRISPR (DepMap Public 23Q2+Score, Chronos), right?
I just wonder how could get the predictability results.
you said that (see details for model definitions.) in “i tab” located next to “Predictability”, but I cannot find the details.

Thanks.

2

The “Co-dependencies” and “Predictability” results come from two different analyses. (But both are using the same CRISPR dataset)

Co-dependencies are simply the top correlations of dependencies as measured by Pearson correlation.

However, the “predictability” tab refers to trying to predict dependencies using other omics features. This was done by creating random forest models. You can see the details of the methodology by clicking the “Information about this page” link in the top left.
Screenshot 2023-06-26 at 9.31.36 AM

Thanks,
Phil

3

Thanks for kind reply.

Best,
Taegeun Bae