# Scvi model training after sub-clustering

**URL:** <https://discourse.scverse.org/t/scvi-model-training-after-sub-clustering/2220>\
**Category:** scvi-tools\
**Tags:** integration, scvi\
**Created:** [April 16, 2024, 5:35pm UTC](https://discourse.scverse.org/t/scvi-model-training-after-sub-clustering/2220 "2024-04-16T17:35:31Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Saksgoel](https://avatars.discourse-cdn.com/v4/letter/s/4da419/32.png) [@Saksgoel](https://discourse.scverse.org/u/Saksgoel)\
**Post date:** [April 16, 2024, 5:35pm UTC](https://discourse.scverse.org/t/scvi-model-training-after-sub-clustering/2220/1 "2024-04-16T17:35:31Z")

</div>

Hi. Thank you for these amazing tools.  
I have a very basic question regarding model training. After generating the model and plotting the map, if I need to sub cluster a population or remove a sample from the dataset, do I need to re-train the model??  
If not re-training can I use the dcvi-normalized values for the sub clustered group?

---

<div class="post-metadata">

**Author:** ![cane11](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.scverse.org/cane11/32/241_2.png) [@cane11](https://discourse.scverse.org/u/cane11)\
**Post date:** [April 25, 2024, 9:23pm UTC](https://discourse.scverse.org/t/scvi-model-training-after-sub-clustering/2220/2 "2024-04-25T21:23:16Z")

</div>

Generally you can just go ahead. I sometimes retrain, if I subset from an atlas object to e.g. T cells and need better resolution for T cell subsets. In general, I rerun the highly variable gene selection before training the second model.