reproducing CRISPRGeneEffect score from read count using Chronos joint run

Hi Team,

Hope you are doing well. I’m currently testing Chronos to see if I can reproduce scores in “CRISPRGeneEffect” from read counts.

The post about the recent update (23Q2):
“The integrated CRISPR data (files beginning with “CRISPR”) is now being produced by a joint Chronos run rather than being integrated post hoc via Harmonia.”

I found that the score I generated using Chronos (Vignette.ipynb, I modified a little bit to match file format and input Avana and KY screens) was reasonably correlated with the 23Q2 score but not exactly the same. Especially the dynamic range of the score (i.e., KRAS) was clearly different ([0,-4] 23Q2, [0,-2] from the reproduced score).

I’m wondering if you used the same code on Github (GitHub - broadinstitute/chronos: Modeling of time series data for CRISPR KO experiments) or a different code. (I guess 2.0?)
If you used the same code, could you please share the pipeline or the code you ran for the new release, by any chance?
Thank you so much for your help and support!

Hi Eiru,

In fact no, the code on github is behind. We have a package of updates collectively we’re calling Chronos 2. I’m finalizing the edits and it should be ready to be pushed to the public repo in a couple weeks, hopefully.

Hi Joshua,

Good. Thank you for letting me know.
I look forward to seeing Chronos 2 published.