# Umap results not comparable

**URL:** <https://discourse.scverse.org/t/umap-results-not-comparable/1099>\
**Category:** Help\
**Created:** [February 2, 2023, 7:16pm UTC](https://discourse.scverse.org/t/umap-results-not-comparable/1099 "2023-02-02T19:16:55Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![Chokerino](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.scverse.org/chokerino/32/516_2.png) [@Chokerino](https://discourse.scverse.org/u/Chokerino)\
**Post date:** [February 2, 2023, 7:16pm UTC](https://discourse.scverse.org/t/umap-results-not-comparable/1099/1 "2023-02-02T19:16:55Z")

</div>

Hello,

I was using scanpy to cluster my data but i found the umap generated to not be comparable to the original umap library. Is there any way to debug this or get the same results on the UMAP library. Is there any additional normalization that scanpy does which I might be missing?

---

<div class="post-metadata">

**Author:** ![gtca](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.scverse.org/gtca/32/1199_2.png) [@gtca](https://discourse.scverse.org/u/gtca)\
**Post date:** [February 2, 2023, 9:03pm UTC](https://discourse.scverse.org/t/umap-results-not-comparable/1099/2 "2023-02-02T21:03:22Z")

</div>

Are there more details you could provide?

For some initial pointers, you should already get comparable results with using `umap.UMAP(...).fit_transform(adata.obsm["X_pca"])` and parameters used by `sc.tl.umap()` (some of them are also saved in `adata.uns["umap"]["params"]`).

---

<div class="post-metadata">

**Author:** ![Chokerino](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.scverse.org/chokerino/32/516_2.png) [@Chokerino](https://discourse.scverse.org/u/Chokerino)\
**Post date:** [February 2, 2023, 9:25pm UTC](https://discourse.scverse.org/t/umap-results-not-comparable/1099/3 "2023-02-02T21:25:50Z")

</div>

I went through the documentation and saw the PCA being performed. This was the main factor.