What happened to Harakiri gene between 21Q2 and 21Q4?

Hello! :slight_smile:
When I updated my scripts with the new data from 21Q4, I found that HRK (8739) gene was mostly essential whereas it was mostly non-essential in 21Q2 (Chronos). And if I check the scores, it looks like this:

Do you have an idea on what caused this shift between the two releases? And which one to trust?

Hi abct,

In the Achilles screens, HRK is mostly not essential. In Sanger screens, it is strongly essential. Every quarter we release a new version of both the Achilles only dataset and also the Achilles + Sanger integrated dataset, harmonized with a package we’re calling Harmonia (main step: ComBat batch correction). The influence of the Sanger dataset shifts this gene to be common essential. Often with conflicting data it can be difficult to know which source to believe, but in this case HRK is unexpressed in many cell lines where the Sanger data says it is highly depleted. I’d treat it as not common essential and trust the Achilles only data instead.

Oh, yes, I used the Achilles for 21Q2 and Achilles + Sanger (with a filtering on Achilles in the plot) for the 21Q4. I didn’t expect that would play such a role!
Thanks a lot for answering!