# \#multivi

**URL:** https://discourse.scverse.org/tag/multivi/15.md

[Latest](https://discourse.scverse.org/latest.md) · [Categories](https://discourse.scverse.org/categories.md) · [Tags](https://discourse.scverse.org/tags.md)

---

## [Issue with multiVI get\_normalized\_expression and get\_normalized\_accessibility](https://discourse.scverse.org/t/issue-with-multivi-get-normalized-expression-and-get-normalized-accessibility/3878)

<div class="topic-metadata">

**Author:** [@andywangzhou](https://discourse.scverse.org/u/andywangzhou)\
**Replies:** 5\
**Last updated:** [December 7, 2025, 7:22am UTC](https://discourse.scverse.org/t/issue-with-multivi-get-normalized-expression-and-get-normalized-accessibility/3878 "2025-12-07T07:22:37Z")

</div>

Hi developers and the community, I’m getting this issue running the MultiVI as instructed in the tutorial: imputed\_expression = model.get\_normalized\_expression() \>\>\> imputed\_expression …

---

## [Question about complex experiment](https://discourse.scverse.org/t/question-about-complex-experiment/3864)

<div class="topic-metadata">

**Author:** [@dackjames](https://discourse.scverse.org/u/dackjames)\
**Replies:** 1\
**Last updated:** [November 25, 2025, 7:39pm UTC](https://discourse.scverse.org/t/question-about-complex-experiment/3864 "2025-11-25T19:39:01Z")

</div>

Hi everyone, I’m starting a fairly complex experiment that includes 10x single-cell, single-nucleus, and 10x multiome (snRNA-seq + scATAC-seq from the same nucleus), with control and treatment groups spread across all t…

---

## [MultiVI questions on Integrating Multiple Multiome Datasets](https://discourse.scverse.org/t/multivi-questions-on-integrating-multiple-multiome-datasets/3786)

<div class="topic-metadata">

**Author:** [@DineshRavindraRaju](https://discourse.scverse.org/u/DineshRavindraRaju)\
**Replies:** 1\
**Last updated:** [September 25, 2025, 12:33pm UTC](https://discourse.scverse.org/t/multivi-questions-on-integrating-multiple-multiome-datasets/3786 "2025-09-25T12:33:02Z")

</div>

Hi all, I have a couple of questions regarding the tool. First, thank you for providing both the tool. For multiple multiome samples, is it better to use Cell Ranger aggr without normalization to integrate modalities …

---

## [multiVI speed and cpu consumption](https://discourse.scverse.org/t/multivi-speed-and-cpu-consumption/3714)

<div class="topic-metadata">

**Author:** [@bxxu](https://discourse.scverse.org/u/bxxu)\
**Replies:** 2\
**Last updated:** [August 30, 2025, 9:16pm UTC](https://discourse.scverse.org/t/multivi-speed-and-cpu-consumption/3714 "2025-08-30T21:16:19Z")

</div>

Hi developers, I am using multiVI (from scvi-tools 1.3.2) to analyze the processed pbmc10k data following the tutorial. For some reason, training is really slow, it takes 15 seconds for 1 epoch. In addition, multiVI is c…

---

## [Does MultiVI support using tile matrix directly instead of peak calling for ATAC input?](https://discourse.scverse.org/t/does-multivi-support-using-tile-matrix-directly-instead-of-peak-calling-for-atac-input/3701)

<div class="topic-metadata">

**Author:** [@Captain\_Pam](https://discourse.scverse.org/u/Captain_Pam)\
**Replies:** 2\
**Last updated:** [July 9, 2025, 4:26pm UTC](https://discourse.scverse.org/t/does-multivi-support-using-tile-matrix-directly-instead-of-peak-calling-for-atac-input/3701 "2025-07-09T16:26:22Z")

</div>

Hi all, I’m working with 10x multiome data and have processed the ATAC and RNA modalities separately: For ATAC, I used tile matrix (e.g., 5kb bins) and trained a peakVI model. For RNA, I used scVI. Now, I’m planning …

---

## [Protocol for model optimization (currently focused on MultiVI)](https://discourse.scverse.org/t/protocol-for-model-optimization-currently-focused-on-multivi/305)

<div class="topic-metadata">

**Author:** [@mvinyard](https://discourse.scverse.org/u/mvinyard)\
**Replies:** 3\
**Last updated:** [May 14, 2025, 9:53pm UTC](https://discourse.scverse.org/t/protocol-for-model-optimization-currently-focused-on-multivi/305 "2025-05-14T21:53:35Z")

</div>

Hi there - first, I am simply blown away by the suite of tools you have assembled. It is truly awe-inspiring and sets the bar for any other python / comp-bio developers. My question (and please assist me if I have place…

---

## [\[suggestion\] what would be the appropriate pipeline to perform joint embedding of GEX and ATAC?](https://discourse.scverse.org/t/suggestion-what-would-be-the-appropriate-pipeline-to-perform-joint-embedding-of-gex-and-atac/3275)

<div class="topic-metadata">

**Author:** [@yojetsharma](https://discourse.scverse.org/u/yojetsharma)\
**Replies:** 1\
**Last updated:** [October 17, 2024, 1:09am UTC](https://discourse.scverse.org/t/suggestion-what-would-be-the-appropriate-pipeline-to-perform-joint-embedding-of-gex-and-atac/3275 "2024-10-17T01:09:33Z")

</div>

I have processed my snRNA of multiome using scanpy and snATAC using snapatac2. Both have been annotated as well. I would like to perform joint embedding of rna and atac but the var attributes are different and so are th…

---

## [Scvi.data.organize\_multiome\_anndatas with two big anndata objects](https://discourse.scverse.org/t/scvi-data-organize-multiome-anndatas-with-two-big-anndata-objects/2247)

<div class="topic-metadata">

**Author:** [@niuyw](https://discourse.scverse.org/u/niuyw)\
**Replies:** 1\
**Last updated:** [May 3, 2024, 8:46pm UTC](https://discourse.scverse.org/t/scvi-data-organize-multiome-anndatas-with-two-big-anndata-objects/2247 "2024-05-03T20:46:47Z")

</div>

Hi, I am trying to use MultiVI with two different anndata objects for paired scRNA and scATAC. In the organize\_multiome\_anndatas step, I was following the tutorial from the singl-cell best practice adata\_paired = ad.co…

---

## [Integration of Multiple Multiome Datasets](https://discourse.scverse.org/t/integration-of-multiple-multiome-datasets/1685)

<div class="topic-metadata">

**Author:** [@jjuhh](https://discourse.scverse.org/u/jjuhh)\
**Replies:** 5\
**Last updated:** [March 6, 2024, 9:45pm UTC](https://discourse.scverse.org/t/integration-of-multiple-multiome-datasets/1685 "2024-03-06T21:45:43Z")

</div>

Thank you for developing a bunch of amazing tools in the single-cell field! I encountered an issue related to the integration of multiple multiome datasets. As I understand, totalVI is specialized in CITE-seq(protein +…

---

## [How to compute the integration metrics](https://discourse.scverse.org/t/how-to-compute-the-integration-metrics/2066)

<div class="topic-metadata">

**Author:** [@Jerry391](https://discourse.scverse.org/u/Jerry391)\
**Replies:** 1\
**Last updated:** [February 16, 2024, 5:58pm UTC](https://discourse.scverse.org/t/how-to-compute-the-integration-metrics/2066 "2024-02-16T17:58:19Z")

</div>

Hi, I want to ask how to compute integration metrics, such as ARI, normalized mutual information, graph connectivity, batch LISI (iLISI), cell-type LISI (cLISI) and so on? Cause I can’t find the related code on GitHub an…

---

## [MultiVI size factor for multiple modalities](https://discourse.scverse.org/t/multivi-size-factor-for-multiple-modalities/1915)

<div class="topic-metadata">

**Author:** [@joshchiou](https://discourse.scverse.org/u/joshchiou)\
**Replies:** 0\
**Last updated:** [November 24, 2023, 11:49pm UTC](https://discourse.scverse.org/t/multivi-size-factor-for-multiple-modalities/1915 "2023-11-24T23:49:48Z")

</div>

In the methods from the MultiVI paper, both the expression and accessibility distributions contain ℓc - a term that captures cell-specific biases. In scvi.model.MULTIVI.setup\_anndata, I believe setting size\_factor\_key ov…

---

## [Nonsense UMAP when including categorical covariates in a MULTIVI model](https://discourse.scverse.org/t/nonsense-umap-when-including-categorical-covariates-in-a-multivi-model/1713)

<div class="topic-metadata">

**Author:** [@bsierieb1](https://discourse.scverse.org/u/bsierieb1)\
**Replies:** 1\
**Last updated:** [August 28, 2023, 9:02pm UTC](https://discourse.scverse.org/t/nonsense-umap-when-including-categorical-covariates-in-a-multivi-model/1713 "2023-08-28T21:02:14Z")

</div>

Hi, I am working with 10X multiome data. I first built an scvi and a peakvi model on the RNA and ATAC portion separately to get a general sense of each modality. Both worked fine, but I also discovered that it was absol…

---

## ["ValueError: cannot specify integer \`bins\` when input data contains infinity" in multi sample data](https://discourse.scverse.org/t/valueerror-cannot-specify-integer-bins-when-input-data-contains-infinity-in-multi-sample-data/1229)

<div class="topic-metadata">

**Author:** [@Behzad](https://discourse.scverse.org/u/Behzad)\
**Replies:** 6\
**Last updated:** [June 27, 2023, 10:36pm UTC](https://discourse.scverse.org/t/valueerror-cannot-specify-integer-bins-when-input-data-contains-infinity-in-multi-sample-data/1229 "2023-06-27T22:36:44Z")

</div>

I have few samples and merged them all (so the adata has 6 samples in it) and followed the scanpy tutorial without any problem until I reached to the point where I had to extract highly variable genes using this command: …

---

## [Dataset integration and analysis](https://discourse.scverse.org/t/dataset-integration-and-analysis/1342)

<div class="topic-metadata">

**Author:** [@mdbabumiamssm](https://discourse.scverse.org/u/mdbabumiamssm)\
**Replies:** 3\
**Last updated:** [May 3, 2023, 5:33pm UTC](https://discourse.scverse.org/t/dataset-integration-and-analysis/1342 "2023-05-03T17:33:11Z")

</div>

Hi everyone, I am working on multiple GEO datasets on a specific blood cancer. My intention is to download cell ranger processed raw data (as available) from GEO and then pre-process, integrate, and use the datasets for…

---

## [totalVI, peakVI, multiVI with scRNA-seq and scATAC-seq data](https://discourse.scverse.org/t/totalvi-peakvi-multivi-with-scrna-seq-and-scatac-seq-data/296)

<div class="topic-metadata">

**Author:** [@Z\_Huang](https://discourse.scverse.org/u/Z_Huang)\
**Replies:** 3\
**Last updated:** [March 9, 2023, 5:53pm UTC](https://discourse.scverse.org/t/totalvi-peakvi-multivi-with-scrna-seq-and-scatac-seq-data/296 "2023-03-09T17:53:07Z")

</div>

Hi community, I am new to scvi-tools and I went through all of the tutorials. I learned peakVI consumes scATAC-seq data for DNA analysis; totalVI consumes CITE-seq data for RNA and protein analysis. Here is my question…

---

## [3 sets of CITE-Seq data, how to concatenate or make single file for further analysis?](https://discourse.scverse.org/t/3-sets-of-cite-seq-data-how-to-concatenate-or-make-single-file-for-further-analysis/1116)

<div class="topic-metadata">

**Author:** [@mdbabumiamssm](https://discourse.scverse.org/u/mdbabumiamssm)\
**Replies:** 0\
**Last updated:** [February 7, 2023, 10:17pm UTC](https://discourse.scverse.org/t/3-sets-of-cite-seq-data-how-to-concatenate-or-make-single-file-for-further-analysis/1116 "2023-02-07T22:17:02Z")

</div>

Hi everyone , I have 3 sets of similar CITE-Seq data, i performed pre-processing individually using Muon+scanpy. Now, I want to make them as single file and perform comparative analysis. Any help ? Thanks in advance. …

---

## [Anndata.concatenate() with two 10x multiome datasets?](https://discourse.scverse.org/t/anndata-concatenate-with-two-10x-multiome-datasets/989)

<div class="topic-metadata">

**Author:** [@mkarikom](https://discourse.scverse.org/u/mkarikom)\
**Replies:** 2\
**Last updated:** [December 29, 2022, 6:10pm UTC](https://discourse.scverse.org/t/anndata-concatenate-with-two-10x-multiome-datasets/989 "2022-12-29T18:10:33Z")

</div>

I have a 10x Multiome data set (GSE199994) and ran scvi.data.read\_10x\_multiome() on each of 10 batches (1 batch per patient). But when I try adataconcat = adata1.concatenate(adata2), the issue seems to be that there are…

---

## [How to dynamically mask categorical\_covariate\_keys during training?](https://discourse.scverse.org/t/how-to-dynamically-mask-categorical-covariate-keys-during-training/840)

<div class="topic-metadata">

**Author:** [@mkarikom](https://discourse.scverse.org/u/mkarikom)\
**Replies:** 3\
**Last updated:** [October 19, 2022, 4:55pm UTC](https://discourse.scverse.org/t/how-to-dynamically-mask-categorical-covariate-keys-during-training/840 "2022-10-19T16:55:14Z")

</div>

I run setup\_anndata(categorical\_covariate\_keys=\['key1','key2',...\]), and need a way to use these covariates selectively in training\_step() (manual optimization). Is there a way to do this? Thanks!

---

## [Clarification on unpaired data smoothing during MultiVI \`training\_step()\`?](https://discourse.scverse.org/t/clarification-on-unpaired-data-smoothing-during-multivi-training-step/811)

<div class="topic-metadata">

**Author:** [@mkarikom](https://discourse.scverse.org/u/mkarikom)\
**Replies:** 0\
**Last updated:** [October 7, 2022, 7:53pm UTC](https://discourse.scverse.org/t/clarification-on-unpaired-data-smoothing-during-multivi-training-step/811 "2022-10-07T19:53:20Z")

</div>

@Tal\_Ashuach, thanks providing your fascinating MultiVI preprint! I just have a few questions related to the method described and the current implementation in scvi-tools. From the last paragraph of 4.1 in the preprint,…

---

## [multiVI and totalVI modal integration question](https://discourse.scverse.org/t/multivi-and-totalvi-modal-integration-question/743)

<div class="topic-metadata">

**Author:** [@yulijia](https://discourse.scverse.org/u/yulijia)\
**Replies:** 0\
**Last updated:** [September 15, 2022, 12:55pm UTC](https://discourse.scverse.org/t/multivi-and-totalvi-modal-integration-question/743 "2022-09-15T12:55:46Z")

</div>

Hi, Recently, I am using scvi-tools to generate multi-modal integrated datasets. I have a question about the MULTIVI.get\_latent\_representation and TOTALVI.get\_latent\_representation function. I followed the user guide …

---

## [Advice regarding complex multimodal integration/analysis strategy](https://discourse.scverse.org/t/advice-regarding-complex-multimodal-integration-analysis-strategy/647)

<div class="topic-metadata">

**Author:** [@amagen](https://discourse.scverse.org/u/amagen)\
**Replies:** 0\
**Last updated:** [July 28, 2022, 3:19pm UTC](https://discourse.scverse.org/t/advice-regarding-complex-multimodal-integration-analysis-strategy/647 "2022-07-28T15:19:12Z")

</div>

Hi all, I have a couple of human cancer types, each with different modalities (scRNAseq,CITEseq,MERFISH for cancer type #1 and scRNAseq,CITEseq,scATACseq,multiome for cancer type #2), some have different technology vers…

---

## [How to impute the modality (like ATAC) when I only have the other modality (RNA) using a pre-trained MultiVI?](https://discourse.scverse.org/t/how-to-impute-the-modality-like-atac-when-i-only-have-the-other-modality-rna-using-a-pre-trained-multivi/580)

<div class="topic-metadata">

**Author:** [@Hongru-Hu](https://discourse.scverse.org/u/Hongru-Hu)\
**Replies:** 5\
**Last updated:** [July 1, 2022, 8:10pm UTC](https://discourse.scverse.org/t/how-to-impute-the-modality-like-atac-when-i-only-have-the-other-modality-rna-using-a-pre-trained-multivi/580 "2022-07-01T20:10:15Z")

</div>

hI, just wondering how to impute the modality (like ATAC) when I only have the other modality (RNA) using a pre-trained MultiVI? If the built-in function does not work for this case, what would the flow I need to pass th…

---

## [Labels\_key not available for scvi.model.MULTIVI.setup\_anndata()](https://discourse.scverse.org/t/labels-key-not-available-for-scvi-model-multivi-setup-anndata/309)

<div class="topic-metadata">

**Author:** [@zhenzuo2](https://discourse.scverse.org/u/zhenzuo2)\
**Replies:** 3\
**Last updated:** [March 1, 2022, 8:03pm UTC](https://discourse.scverse.org/t/labels-key-not-available-for-scvi-model-multivi-setup-anndata/309 "2022-03-01T20:03:30Z")

</div>

scvi.model.MULTIVI.setup\_anndata(adata, batch\_key='orig.ident', labels\_key = 'scpred\_prediction') Traceback (most recent call last): File "\<stdin\>", line 1, in \<module\> TypeError: setup\_anndata() got an unexpected keyw…

---

## [Reference mapping with MultiVI](https://discourse.scverse.org/t/reference-mapping-with-multivi/252)

<div class="topic-metadata">

**Author:** [@JGM](https://discourse.scverse.org/u/JGM)\
**Replies:** 0\
**Last updated:** [January 25, 2022, 8:40am UTC](https://discourse.scverse.org/t/reference-mapping-with-multivi/252 "2022-01-25T08:40:23Z")

</div>

First of all, thank you for a great set of tools. I was wondering if it is possible to create a mapping/scArches-like framework using the MultiVI model? Optimally, queries could be added using either both or only one of…

---

## [Integrate multiple samples for paired multi-omics data](https://discourse.scverse.org/t/integrate-multiple-samples-for-paired-multi-omics-data/279)

<div class="topic-metadata">

**Author:** [@zhenzuo2](https://discourse.scverse.org/u/zhenzuo2)\
**Replies:** 1\
**Last updated:** [February 15, 2022, 12:54am UTC](https://discourse.scverse.org/t/integrate-multiple-samples-for-paired-multi-omics-data/279 "2022-02-15T00:54:07Z")

</div>

Hi, Are there ways to merge paired data before the model traning? I can merge ‘Gene Expression’ but not ‘peaks’ from multiple samples.

---

## [MULTIVI training fails before first epoch](https://discourse.scverse.org/t/multivi-training-fails-before-first-epoch/236)

<div class="topic-metadata">

**Author:** [@MoritzTh](https://discourse.scverse.org/u/MoritzTh)\
**Replies:** 1\
**Last updated:** [December 16, 2021, 4:21pm UTC](https://discourse.scverse.org/t/multivi-training-fails-before-first-epoch/236 "2021-12-16T16:21:20Z")

</div>

Hi, I am trying to integrate mostly paired scRNA and scATAC data following your MULTIVI tutorial. Creating the mvi anndata and setting up the model with scvi.model.MULTIVI.setup\_anndata(adata\_mvi, batch\_key='modality') …

---

## [MultiVi with multiple GPUs and CPUs](https://discourse.scverse.org/t/multivi-with-multiple-gpus-and-cpus/190)

<div class="topic-metadata">

**Author:** [@DanieleMuraro](https://discourse.scverse.org/u/DanieleMuraro)\
**Replies:** 2\
**Last updated:** [December 2, 2021, 12:04pm UTC](https://discourse.scverse.org/t/multivi-with-multiple-gpus-and-cpus/190 "2021-12-02T12:04:40Z")

</div>

Dear scvi-tools developers and community, I was wondering if it is possible to run MultiVi with multiple GPUs and CPUs. In case it is, how can I set up the number of GPUs and CPUs for use of MultiVi? Thank you for you…