UNIFAN (Unsupervised Single-cell Functional Annotation) simultaneously clusters and annotates cells with known biological processes (including pathways). For each single cell, UNIFAN first infers gene set activity scores (denoted by r in the figure below) associated with this cell using the input gene sets.
Next, UNIFAN clusters cells by using the learned gene set activity scores (r) and a reduced dimension representation of the expression of genes in the cell. The gene set activity scores are used by an “annotator” to guide the clustering such that cells sharing similar biological processes are more likely to be grouped together. Such design allows the method to focus on the key processes when clustering cells and so can overcome issues related to noise and dropout while simultaneously selecting marker gene sets which can be used to annotate clusters. To allow the selection of marker genes for each cluster, we also add a subset of the most variable genes selected using Seuratv3 (Stuart et al., 2019) as features for the annotator.
- Python >= 3.6
- Python side-packages:
-- pytorch >= 1.9.0
-- numpy >= 1.19.2
-- pandas>=1.1.5
-- scanpy >= 1.7.2
-- leidenalg>=0.8.4
-- tqdm >= 4.61.1
-- scikit-learn>=0.24.2
-- umap-learn>=0.5.1
-- matplotlib >= 3.3.4
-- seaborn >= 0.11.0
It is recommended to use a virtural environment/pacakges manager such as Anaconda. After successfully installing Anaconda/Miniconda, create an environment by the following:
conda create -n myenv python=3.6You can then install and run the package in the virtual environment. Activate the virtural environment by:
conda activate myenvMake sure you have pip installed in your environment. You may check by
conda listIf not installed, then:
conda install pipUNIFAN is built based on Pytorch and supporting both CPU or GPU. Make sure you have Pytorch (>= 1.9.0) installed in your virtual environment. If not, please visist Pytorch and install the appropriate version.
Install by:
pip install git+https://github.com/doraadong/UNIFAN.gitIf you want to upgrade UNIFAN to the latest version, then first uninstall it by:
pip uninstall unifanAnd then just run the pip install command again.
You may import UNIFAN as an package and use it in your code (See Tutorials for details). Or you may train models using the following command-line tool.
Run UNIFAN by (arguments are taken for example):
main.py -i ../example/input/Limb_Muscle.h5ad -o ../example/output -p tabula_muris -t Limb_Muscle -l cell_ontology_class -e ../gene_sets/The usage of this command is listed as follows. Note only the first 5 inputs are required:
usage: main.py [-h] -i INPUT -o OUTPUT -p PROJECT -t TISSUE [-e GENESETSPATH]
[-l LABEL] [-v VARIABLE] [-r PRIOR]
[-f {gene_sets,gene,gene_gene_sets}] [-a ALPHA] [-b BETA]
[-g GAMMA] [-u TAU] [-d DIM] [-s BATCH] [-na NANNO]
[-ns NSCORE] [-nu NAUTO] [-nc NCLUSTER] [-nze NZENCO]
[-nzd NZDECO] [-dze DIMZENCO] [-dzd DIMZDECO] [-nre NRENCO]
[-dre DIMRENCO] [-drd DIMRDECO]
[-n {sigmoid,non-negative,gaussian}] [-m SEED] [-c CUDA]
[-w NWORKERS]
optional arguments:
-h, --help show this help message and exit
-i INPUT, --input INPUT
string, path to the input expression data, default
'../input/data.h5ad'
-o OUTPUT, --output OUTPUT
string, path to the output folder, default
'../output/'
-p PROJECT, --project PROJECT
string, identifier for the project, e.g., tabula_muris
-t TISSUE, --tissue TISSUE
string, tissue where the input data is sampled from
-e GENESETSPATH, --geneSetsPath GENESETSPATH
string, path to the folder where gene sets can be
found, default='../gene_sets/'
-l LABEL, --label LABEL
string, optional, the column / field name of the
ground truth label, if available; used for evaluation
only; default None
-v VARIABLE, --variable VARIABLE
string, optional, the column / field name of the
highly variable genes; default 'highly_variable'
-r PRIOR, --prior PRIOR
string, optional, gene set file names used to learn
the gene set activity scores, use '+' to separate
multiple gene set names, default
c5.go.bp.v7.4.symbols.gmt+c2.cp.v7.4.symbols.gmt+TF-
DNA
-f {gene_sets,gene,gene_gene_sets}, --features {gene_sets,gene,gene_gene_sets}
string, optional, features used for the annotator, any
of 'gene_sets', 'gene' or 'gene_gene_sets', default
'gene_gene_sets'
-a ALPHA, --alpha ALPHA
float, optional, hyperparameter for the L1 term in the
set cover loss, default 1e-2
-b BETA, --beta BETA float, optional, hyperparameter for the set cover term
in the set cover loss, default 1e-5
-g GAMMA, --gamma GAMMA
float, optional, hyperparameter for the exclusive L1
term, default 1e-3
-u TAU, --tau TAU float, optional, hyperparameter for the annotator
loss, default 10
-d DIM, --dim DIM integer, optional, dimension for the low-dimensional
representation, default 32
-s BATCH, --batch BATCH
integer, optional, batch size for training except for
pretraining annotator (fixed at 32), default 128
-na NANNO, --nanno NANNO
integer, optional, number of epochs to pretrain the
annotator, default 50
-ns NSCORE, --nscore NSCORE
integer, optional, number of epochs to train the gene
set activity model, default 70
-nu NAUTO, --nauto NAUTO
integer, optional, number of epochs to pretrain the
annocluster model, default 50
-nc NCLUSTER, --ncluster NCLUSTER
integer, optional, number of epochs to train the
annocluster model, default 25
-nze NZENCO, --nzenco NZENCO
float, optional, number of hidden layers for encoder
of annocluster, default 3
-nzd NZDECO, --nzdeco NZDECO
float, optional, number of hidden layers for decoder
of annocluster, default 2
-dze DIMZENCO, --dimzenco DIMZENCO
integer, optional, number of nodes for hidden layers
for encoder of annocluster, default 128
-dzd DIMZDECO, --dimzdeco DIMZDECO
integer, optional, number of nodes for hidden layers
for decoder of annocluster, default 128
-nre NRENCO, --nrenco NRENCO
integer, optional, number of hidden layers for the
encoder of gene set activity scores model, default 5
-dre DIMRENCO, --dimrenco DIMRENCO
integer, optional, number of nodes for hidden layers
for encoder of gene set activity scores model, default
128
-drd DIMRDECO, --dimrdeco DIMRDECO
integer, optional, number of nodes for hidden layers
for decoder of gene set activity scores model, default
128
-n {sigmoid,non-negative,gaussian}, --network {sigmoid,non-negative,gaussian}
string, optional, the encoder for the gene set
activity model, any of 'sigmoid', 'non-negative' or
'gaussian', default 'non-negative'
-m SEED, --seed SEED integer, optional, random seed for the initialization,
default 0
-c CUDA, --cuda CUDA boolean, optional, if use GPU for neural network
training, default False
-w NWORKERS, --nworkers NWORKERS
integer, optional, number of workers for dataloader,
default 8
Github rendering disables some functionalities of Jupyter notebooks. We recommend using nbviewer to view the following tutorials.
In UNIFAN training tutorial, we illustrate how to run UNIFAN step-by-step on the example data: Limb_Muscle from Tabula Muris.
You may download the gene sets in gene_sets. As default, we use the GO terms for biological processes (c5.go.bp.v7.4.symbols.gmt), canonical pathways (c2.cp.v7.4.symbols.gmt) and the TF-DNA interacitons data (Mouse_TF_targets.txt).
UNIFAN takes AnnData files as input. See AnnData for details. To prepare the example data (Limb_Muscle in Tabula Muris), first download the Tabula Muris senis data. Then run the Python script getExample.py to preprocess the count data using the following command:
python getExample.py -p ./facs.h5ad -i ../example/input -t Limb_Muscle
The usage of this command is listed as follows:
usage: getExample.py [-h] -p PATH -i FOLDER -t TISSUE [-k TOPK]
optional arguments:
-h, --help show this help message and exit
-p PATH, --path PATH string, path to the downloaded data, default
'./facs.h5ad'
-i FOLDER, --folder FOLDER
string, path to the folder to save the data, default
'../example/input'
-t TISSUE, --tissue TISSUE
string, specify the output tissue; if using the
default None, then all tissues will be outputted and
saved separately in the folder; default None
-k TOPK, --topk TOPK integer, optional, number of most variable genes,
default 2000
We also provide Data preprocessing showing how we preprocessed the other datasets we used in the manuscript.
In cluster annotating tutorial, we illustrate how to use the coefficients learned by UNIFAN to annotate clusters. Particularly, we show how to select representing gene sets / genes for each cluster, evaluate if selected genes are likely marker genes and visualize the annotations.
- 10-11-2022:
-- Add tutorial on preprocessing the datasets used in the manuscript
Check our paper at Genome Research. Link to preprint.
The software is an implementation of the method UNIFAN, jointly developed by Dongshunyi "Dora" Li and Ziv Bar-Joseph from System Biology Group @ Carnegie Mellon University and Jun Ding from McGill University.
- dongshul at andrew.cmu.edu
This project is licensed under the MIT License - see the LICENSE file for details