Skip to content
epigenome analysis to rank transcription factors
JavaScript C HTML Python CSS Shell Other
Branch: master
Clone or download
Pull request Compare This branch is 5 commits behind qinqian:master.
Qian Qin
Qian Qin update LICENSE
Latest commit c60207a Dec 19, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
bin upgrade snakemake to 5.7.1, update website Oct 20, 2019
lisa upgrade snakemake to 5.7.1, update website Oct 20, 2019
lisa_docs update cis-elements prediction module, update web interafce Aug 1, 2018
lisa_web update LICENSE Dec 19, 2019
.gitignore add into anaconda packages, improve installation Feb 28, 2019
LICENSE update LICENSE Dec 19, 2019 new lisa init May 10, 2017 add citation Nov 20, 2019
environment.yml add install with environment.yml, improve website Oct 15, 2018


Web version and documentation is hosted at For large scale gene set analysis, we recommend user to install local version.

Preparation of Anaconda environment

wget -c
export PATH="${HOME}/miniconda3/bin:$PATH"

conda create -n lisa python=3.6 && conda config --add channels conda-forge && conda config --add channels bioconda


conda activate lisa
# or for old conda
source activate lisa

conda install -c qinqian lisa

To update, use git clone && cd lisa && python develop.

Get pre-computed datasets from CistromeDB

User can download hg38 or mm10 datasets based on their experiments for human or mouse, the password can be obtained after LISA is published.

wget --user=lisa --password='xxx'

# or

wget --user=lisa --password='xxx'

Then, user need to uncompress the datasets, and update the configuration for lisa.

tar xvfz lisa_v1.0_hg38.tar.gz
lisa_update_conf --folder hg38/ --species hg38

# or

tar xvfz lisa_v1.0_mm10.tar.gz
lisa_update_conf --folder mm10/ --species mm10


Given multiple gene set file gene_set1, gene_set2, gene_set3 et al., each file has one gene (RefSeq id or gene symbol) for each row, user can predict transcriptional regulator ranking using the following commands with random background genes

time lisa model --method="all" --web=True --new_rp_h5=None --new_count_h5=None --species hg38 --epigenome "['DNase', 'H3K27ac']" --cluster=False --covariates=False --random=True --prefix first_run --background=None --stat_background_number=1000 --threads 4 gene_set1 gene_set2 gene_set3 ...

Alternatively, user can generate a fixed background genes based on TAD and promoter activity, and input it to lisa,

lisa_premodel_background_selection --species hg38 --epigenomes="['DNase']" --gene_set=None --prefix=test --random=None --background=dynamic_auto_tad
cut -f 5 -d: test.background_gene.3000 > test.fixed.background_gene

time lisa model --method="all" --web=True --new_rp_h5=None --new_count_h5=None --species hg38 --epigenome "['DNase', 'H3K27ac']" --cluster=False --covariates=False --random=True --prefix first_run --background=test.fixed.background_gene --stat_background_number=1000 --threads 4 gene_set1 gene_set2 gene_set3 ...

User can also input a customized background genes, which should include more than 30 unique RefSeq genes, all input genes are used for modeling and computing statistics, so --stat_background_number is ignored.

time lisa model --method="all" --web=True --new_rp_h5=None --new_count_h5=None --species hg38 --epigenome "['DNase', 'H3K27ac']" --cluster=False --covariates=False --random=True --prefix first_run --background=test.fixed.background_gene --threads 4 gene_set1 gene_set2 gene_set3 ...

Update LISA

git clone
source activate lisa
cd lisa && python develop
lisa_update_conf --folder hg38/ --species hg38
lisa_update_conf --folder mm10/ --species mm10

Remove LISA

conda env remove -n lisa
rm -r mm10/ hg38/


Qin Q, Fan J, Zheng R, Wan C, Mei S, Wu Q. Inferring transcriptional regulators through integrative modeling of public chromatin accessibility and ChIP-seq data. 2019.

Please note that the reference is a preprint hosted at biorxiv.

You can’t perform that action at this time.