A generalization of tSNE and UMAP to single cell multimodal data
Project description
Emerging single-cell genomics technologies profile multiple types of molecules within a single cell. A fundamental step in the analysis of the produced high dimensional data is their visualization using dimensionality reduction techniques such as t-SNE and UMAP. We developed Jvis, a dimensionality reduction framework that naturally generalizes t-SNE and UMAP to the joint visualization of multimodal omics data. Our approach automatically learns a weighting of individual modalities from the data that promotes discriminative features but suppresses random noise. On 8 single-cell multimodal data sets across 5 different technologies, Jvis produced a unified embedding that better agrees with known cell states and RNA and protein velocity landscapes compared to unimodal representations.
The details for the underlying mathematics can be found in insert link here.
Van Hoan Do, Stefan Canzar, A generalization of tSNE and UMAP to single cell multimodal data, xxx.
Installing
Requirements:
Python 3.6 or greater
numpy
scipy
scikit-learn
numba
Install Options
PyPI install, presuming you have numba and sklearn and all its requirements (numpy and scipy) installed:
pip install testJvis-learnIf you have a problem with pip installation then we’d suggest installing the dependencies manually using anaconda followed by pulling umap from pip:
conda install numpy scipy
conda install scikit-learn
conda install numba
pip install testJvis-learnHow to use Jvis
The Jvis package inherits from sklearn TSNE, and UMAP. Therefore, all parameters of tSNE and UMAP are naturally extended for Jvis.
An example of making use of these options:
from Jvis import JUMAP, JTSNE
import numpy as np
# Create a toy example from a random distribution (n_cells = 500)
rna_rand = np.random.rand(500, 100)
adt_rand = np.random.rand(500, 15)
data = {'rna': rna_rand, 'adt': adt_rand} # create a dictionary of modalities.
# Run joint TSNE of the two "random" modalities.
embedding_jtsne = JTSNE(n_components=2).fit_transform(data)
# Run joint UMAP of the two "random" modalities.
embedding_jumap = JUMAP(n_neighbors=20,
min_dist=0.3,
metric='correlation').fit_transform(data)For more realistic examples and Python scripts to reproduce the results in our paper are available at GitHub: xxx
Tunning parameters of t-SNE and UMAP can be found here: https://scikit-learn.org/stable/modules/generated/sklearn.manifold.TSNE.html
License
The JVis package is 3-clause BSD licensed.
Note that the Jvis package is inherited from scikit-learn and UMAP package under 3-clause BSD license.
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for testJvis_learn-0.0.6-py3-none-any.whl
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 | 2c567e2c1d4c26d5bc64649f52b36d49cef16c5c7be9960e4dd4aa9d8f97491e |
|
| MD5 | cbcc7a4eef5bc037064d399048ba1c0e |
|
| BLAKE2b-256 | b09e02c3a9c634f60145834be3e9d83ae97691bfe4f1a7bc59845dc9625c55b6 |
