# Predictability features in cell line subsets

**URL:** <https://forum.depmap.org/t/predictability-features-in-cell-line-subsets/1419>\
**Category:** Q&A\
**Created:** [May 24, 2022, 7:18am UTC](https://forum.depmap.org/t/predictability-features-in-cell-line-subsets/1419 "2022-05-24T07:18:52Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![rnoble](https://avatars.discourse-cdn.com/v4/letter/r/3d9bf3/32.png) [@rnoble](https://forum.depmap.org/u/rnoble)\
**Post date:** [May 24, 2022, 7:18am UTC](https://forum.depmap.org/t/predictability-features-in-cell-line-subsets/1419/1 "2022-05-24T07:18:53Z")

</div>

Is there a way to get relative importance scores for features in a selected collection of cell lines? It’s obviously possible to check for correlations between dependencies on a given gene and features in this way but it would be really useful to be able to generate correlation and relative importance values for a given disease type for example. I appreciate this isn’t currently a feature in the DepMap portal but any suggestions of how to do this outside of DepMap would be very interesting.

Thanks

---

<div class="post-metadata">

**Author:** ![Joshua\_Dempster](https://yyz1.discourse-cdn.com/flex035/user_avatar/forum.depmap.org/joshua_dempster/32/19_2.png) [@Joshua\_Dempster](https://forum.depmap.org/u/Joshua_Dempster)\
**Post date:** [May 31, 2022, 1:06pm UTC](https://forum.depmap.org/t/predictability-features-in-cell-line-subsets/1419/2 "2022-05-31T13:06:27Z")

</div>

You could train a model using only those cell lines, but for any given disease type your model will be quite underpowered. I would stick to correlation measures.