yomix 1.7.0

yomix: an interactive tool to explore low dimensional embeddings of omics data

LATEST RELEASE: see https://pypi.org/project/yomix/

DOCUMENTATION: https://perrin-isir.github.io/yomix/

Yomix is an interactive tool to explore low dimensional embeddings of omics data.

As illustrated in the GIF below, users can explore embeddings—typically generated using dimensionality reduction techniques such as UMAP, t-SNE, TriMap, or VAEs. Within these visualizations, users can identify clusters of interest and use a lasso tool to define specific subsets. Yomix can then compute gene signatures for these subsets almost instantaneously. For instance, it can identify gene signatures that best distinguish subset A from the rest of the data, differentiate subset A from subset B, or find genes whose expression levels are most strongly correlated with a user-defined direction within subset A.

INSTALL

In a python virtual environment, do:

pip install yomix

Then try the tool with:

yomix --example

To use it on your own files:

yomix yourfile.h5ad

where yourfile.h5ad is an anndata object saved in h5ad format (see anndata - Annotated data), with at least one .obsm field of dimension 2 or more (the low dimensional embedding).

When there are many samples in the dataset, the --subsampling option can be passed to improve reactiveness:

yomix --subsampling N yourfile.h5ad

It randomly subsamples the dataset to a maximum number of N samples. For example:

yomix --subsampling 5000 yourfile.h5ad

Other option: INSTALL FROM SOURCE

git clone https://github.com/perrin-isir/yomix.git

We recommand to create a python environment with micromamba, but any python package manager can be used instead.

cd yomix

micromamba create --name yomixenv --file environment.yaml

micromamba activate yomixenv

pip install -e .

Then try the tool with:

yomix yomix/example/pbmc.h5ad

The input file must be an anndata object saved in h5ad format (see anndata - Annotated data), with at least one .obsm field of dimension 2 or more.

Using Seurat objects with Yomix

You can use Seurat objects by converting them to .h5ad format in R:

Load required libraries:

library(rhdf5)
library(dplyr)
library(patchwork)
library(SeuratDisk)
library(Seurat)
library(SeuratData)

Load the object:

my_file <- readRDS("path.rds")

If it is a SingleCellExperiment object, convert to Seurat:

if (inherits(my_file, "SingleCellExperiment")) {
  my_file <- as.Seurat(my_file)
}

Save as H5Seurat:

SaveH5Seurat(my_file, filename = "filename.h5seurat")

Convert to .h5ad:

Convert("filename.h5seurat", dest = "h5ad", output.path = "/.h5ad")

List of contributors

Nicolas Perrin-Gilbert

Joshua Waterfall

Pierre Fumeron

Nisma Amjad

Jason Z. Kim

Erkan Narmanli

Christopher R. Myers

James P. Sethna

Jérôme Contant

Thomas Fuks

Julien Vibert

Silvia Tulli

Philippe Martin