Skip to content

TranslationalBioinformaticsUnit/LIBRA

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Latest version Downloads Documentation DOI

LIBRA - Machine Translation between paired
Single-Cell Multi-Omics Data

This repository contains the LIBRA code and online data used for Single-cell multi-omics integration and prediction analysis employed on LIBRA manuscript. Libra metrics are also available for quantifying outputs quality as well as novel PPJI preservation measurement. Seurat code employed to analyze LIBRA input omics as well as for clustering and visualization pipelines are providen.

The Python package sc-Libra, has been developed with the aim of extending and summarizing the developer code used on the paper to a user-friendly version and is freely available in the PyPI repository. Read online package documentation for detailled description and guidelines.

Summary

LIBRA is a deep learning model that is designed for Single-cell multi-omics integration and prediction. LIBRA performs this by using an unbalance Autoencoder which learns a shared low-dimensional embedding from both experiment omics, combining each sample's uniqueness for generating a enriched representation of integrated data respect to the original experiment independent data. This tool has been first developed in R code, a code snapshot is providen for R users. Next, adaptative LIBRA (aLIBRA) tool has been develop for paralellize training of LIBRA models using a grid structure for selecting optimal hyperparameters in a automatic way excluding the requirement of doing this by users saving considerable time. Snapshot code is providen in Python code for conceptual understanding.

As a result from these raw developer-codes provided, sc-Libra package is provided as a built-in resource to perform the pipeline propossed.

For further details, please refer to the online manuscript currently at biorxiv repository (will be updated asap).

Installation

To run sc-Libra pipeline the following settings are required:

  • Install Python >=3.7.0.
  • Install R >=3.5.2.
  • Install sc-libra python package: $ pip install sc_libra

For stepwise guide follow the online documentation.

Datasets

Find Neurips provided dataset for LIBRA testing at figsahre repository to be downloaded here.

Following datasets consist only on the sparse versions without cell/feature identity, go to corresponding autor references for original datasets.

LIBRA name GSE link Modalities Technology Genomic ref used Download sparse matrix
DataSet1 GSE126074 scRNAseq + scATACseq SNARE-seq Mus_musculus.GRCm38 Ver: 3.0.0 RNA and ATAC
DataSet2 GSE128639 scRNAseq + scADT CITE-seq Homo_sapiens.GRCh38 Ver: 3.0.0 RNA and ADT
DataSet3 GSE130399 scRNAseq + scATACseq Paired-seq Mus_musculus.GRCm38 Ver: 3.0.0 RNA and ATAC
DataSet4 GSE140203 scRNAseq + scATACseq SHARE-seq Mus_musculus.GRCm38 Ver: 3.0.0 RNA and ATAC
DataSet5 10X Genomics scRNAseq + scATACseq 10X multiome Homo_sapiens.GRCh38 Ver: 3.0.0 RNA and ATAC
DataSet6 GSE194122 scRNAseq + scATACseq 10X multiome Homo_sapiens.GRCh38 Ver: 3.0.0 RNA and ATAC
DataSet7 GSE194122 scRNAseq + scADT CITE-seq Homo_sapiens.GRCh38 Ver: 3.0.0 RNA and ADT
DataSet8 GSE109262 scRNAseq + scATACseq scNMT-seq Mus_musculus.GRCm38 Ver: 3.0.0 RNA and ATAC

Usage

  • Easiest way of running LIBRA analysis is though sc-Libra python package.
  • Package documentation is online available using "Read the Docs" platform.

Material of interest

LIBRA benchmarking comparison:

For validating LIBRA performance we compared it against other:

Further details are provided at supplementary material added at LIBRA manuscript.

LIBRA visual workflow:

workflow.png

About

Machine Translation between paired Single Cell Multi Omics Data

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published