leomorelli / scGET

1. scGET

The goal of single cell Genome and Epigenome by Transposases sequencing (scGET-seq) is to discriminate between accessible and compacted chromatin regions within each single cell. The discrimination of chromatin accessibility relies on two different transposases: transposase-5 binds to the accessible chromatin (tn5) and transposase-H, a chimeric form of tn5 (tnH), which recognizes the compacted chromatin.

scGET architecture is built using Snakemake: a workflow management system, which guarantees the possibility to parallelize independent jobs. scGET workflow is described by the image below: starting from sequenced FASTQ files, scGET will generate an AnnData object where tn5 matrix and tnh matrix are present as two different layers.

2. Installing scGET

First, scGET repository must be cloned:

git clone --recursive https://github.com/leomorelli/scGET.git

Before getting your hands dirty with scGET analyses, it is necessary to create a suitable conda environment. However, some packages cannot be installed, using conda. Therefore, we have designed a 4-step process, allowing an easy and quick generation of the scget environment.

1. The conda environment can be automatically generated, thanks to the scget.yaml file:

conda env create -f scget.yaml
conda activate scget

2. TagDust package must be installed, after the activation of the scget environment. First, the package must be downloaded and compiled; second, from the tagdust directory,the binary tagdust file can be copied in the scget environment:

wget https://sourceforge.net/projects/tagdust/files/tagdust-2.33.tar.gz
tar -zxvf tagdust-2.33.tar.gz 
cd tagdust-2.33
./configure 
make
make check
cp ./src/tagdust $CONDA_PREFIX/bin

3. Similarly, also samtools must be installed:
   • git repositories of samtools and htslib must be cloned
   • htslib must be compiled and installed
   • samtools must be compiled and installed

git clone https://github.com/samtools/samtools.git
git clone https://github.com/samtools/htslib.git

cd htslib
autoreconf -i
git submodule update --init --recursive
./configure --prefix=$CONDA_PREFIX
make 
make install

cd samtools
autoheader
autoconf -Wno-syntax
./configure --prefix=$CONDA_PREFIX --without-curses
make
make install

4. scatACC repository should be automatically retrieved within the current repository, otherwise it must be cloned from github:

git clone https://github.com/dawe/scatACC.git

In order to perform the analysis through slurm, it may be useful to check if screen package has already been installed:

screen --version

Output (example):

Screen version 4.08.00 (GNU) 05-Feb-20

If screen has not been installed yet, it could be easily installed via sudo:

sudo apt update
sudo apt install screen

3. Slurm set up

Although scGET can be used locally, it is optimized to work on a cluster, managed by Slurm workload manager.

• Inside ${HOME}/.config, a series of nested directories should be created, such that you obtain the following path ${HOME}/.config/snakemake/slurm. Inside the slurm folder, a config.yaml file can be generated:

mkdir -p ${HOME}/.config/snakemake/slurm
cd ${HOME}/.config/snakemake/slurm
vi config.yaml

• After that, the config.yaml file must be compiled as explained below (remember to update the queue name specified by the -p option and your mail-user):

jobs: 38
cluster: "sbatch --mem={resources.mem_mb} -c {resources.cpus} --job-name {rule}.smk -o {OUTPUT_PATH}/logs_slurm/{rule}_%j.o -e {OUTPUT_PATH}/logs_slurm/{rule}_%j.e --mail-type=FAIL --mail-user=[e-mail]"
default-resources: [cpus=1, mem_mb=5000]
resources: [cpus=40, mem_mb=60000]
restart-times: 3
use-conda: true

4. Configuration

The path for scatACC directory (should be within the current directory), together with the path for the genome and the bed_file must be clarified in the config.yaml present in the scGET folder.

EXAMPLE:

Let's assume that the scGET directory is located in our home directory (${HOME}/scGET); scatACC directory is then in a directory ${HOME}/scGET/scatACC); on the other hand, the genome file (hg38.fa), lays in the "references" directory (${HOME}/references/hg.38), together with the bed_file (${HOME}/references/hg385kbin.bed):

• First, you should open the config.yaml file, in the scGET directory:

cd ${HOME}/scGET
vi config.yaml

Output:

sample: ''

reads: [1,2,3]

barcodes: {'tn5':['CGTACTAG','TCCTGAGC','TCATGAGC','CCTGAGAT'],'tnh':['TAAGGCGA','GCTACGCT','AGGCTCCG','CTGCGCAT']}

genome: ${HOME}/genome.fa

bed_file: ${HOME}/genome.bed

threads: 8

cell_number: 5000

scatacc_path: '${HOME}/scGET/scatACC'

input_path: ''

input_list: ''

output_path: ''

• After that, we must modify the field scatacc_path, specifying our actual scatACC path, the field genome, clarifying the genome path with the genome file name and the field bed_file with the path for the bed file:

Output:

sample: ''

reads: [1,2,3]

barcodes: {'tn5':['CGTACTAG','TCCTGAGC','TCATGAGC','CCTGAGAT'],'tnh':['TAAGGCGA','GCTACGCT','AGGCTCCG','CTGCGCAT']}

genome: ${HOME}/references/hg38.fa

bed_file: ${HOME}/references/hg385kbin.bed

threads: 8

cell_number: 5000

scatacc_path: '${HOME}/scGET/scatACC'

input_path: ''

input_list: ''

output_path: ''

N.B. the REFERENCE GENOME must be INDEXED before the analysis

If the genome has not been indexed yet, you can make up for this in three steps:

• Activate the scget conda environment
• Open the directory where the reference genome is stored
• Index the genome, using samtools library

conda activate scget
cd ${HOME}/references
bwa index hg38.fa

5. Input file

Two inputs are mandatory to start the scGET analisys:

• The path for fastq input files
• A .txt file, listing names of the files ready to be analyzed

EXAMPLE:

Let's assume that fastq files are stored in ${HOME}/files/samples directory: ${HOME}/files/samples represents the input path; while names of files inside ${HOME}/files/samples directory represent the content of the .txt file, we must create.

ls ${HOME}/files/samples

Output:

sample_S1_L001_R1_001.fastq.gz

sample_S1_L001_R2_001.fastq.gz

sample_S1_L001_R3_001.fastq.gz

sample_S1_L002_R1_001.fastq.gz

sample_S1_L002_R2_001.fastq.gz

sample_S1_L002_R3_001.fastq.gz

From the output above, it easy to understand which read number corresponds to each file (R1, R2 and R3). The .txt file, must be built as follow:

• Each line corresponding to a file name
• Next to the file name, the read number should be clarified
• Finally, the sample name must be indicated next to the read number. This step allows the simultanous analysis of different samples. -> file.fq.gz | read_n° | sample_name

EXAMPLE:

vi input_info.txt

After that, it must be modified as explained below:

sample_S1_L001_R1_001.fastq.gz 1 S1

sample_S1_L001_R2_001.fastq.gz 2 S1

sample_S1_L001_R3_001.fastq.gz 3 S1

sample_S1_L002_R1_001.fastq.gz 1 S1

sample_S1_L002_R2_001.fastq.gz 2 S1

sample_S1_L002_R3_001.fastq.gz 3 S1

sample_S2_L001_R1_001.fastq.gz 1 S2

sample_S2_L001_R2_001.fastq.gz 2 S2

sample_S2_L001_R3_001.fastq.gz 3 S2

sample_S2_L002_R1_001.fastq.gz 1 S2

sample_S2_L002_R2_001.fastq.gz 2 S2

sample_S2_L002_R3_001.fastq.gz 3 S2

6. Run

Now it's time to start the analysis! It is important to remember that the scGET analysis must be performed from the scGET directory or from a directory in which the Snakefile, the config.yaml and the scripts files are copied. Therefore, before starting the workflow, you should reach the scGET directory and activate the scGET environment.

cd ${HOME}/scGET
conda activate scget

In order to start with scGET analysis, you must run the following command, specifying the input_path, the output_path and the input_list generated above:

snakemake --cores 8 --config input_path=/home/files/experiment_test output_path=/home/results input_list=input_file.txt --profile slurm

7. Output

Once scGET analysis is finished results files as well as log files are generated and stored in the output directory:

• Results files are stored in a directory named after the sample name
• Log files are stored in the logs_slurm directory, located in the directory, indicated by the output_path

The location of results directory is indicated by the parameter output_path.

N.B.

If you need to dig more into scGET settings, you can find more info about scGET usage in the advanced.md file.