DeepG4ToolsComparison: A snakemake pipeline to run and compare G4 DNA prediction tools with DeepG4

The predictions for differents tissues and cancer with DeepG4 is available here

The code to generate the precision/recall curve is available here.

Overview

It’s based on Snakemake to manage the workflow and Docker to isolate the application and run it with the appropriate tool versions.

Installation

Clone the repository :

git clone https://github.com/morphos30/DeepG4ToolsComparison.git
cd DeepG4ToolsComparison

Install the docker image and run it :

docker build . -t morphos30/g4docker -f Dockerfile/Dockerfile
docker run -it -v $(pwd):/DeepG4ToolsComparison morphos30/g4docker /bin/bash

Where $(pwd) is the working directory of DeepG4ToolsComparison on your computer.

Launch the pipeline :

cd /DeepG4ToolsComparison
snakemake --use-conda -j 30

You have to set the option --use-conda in order to install and run each tool in its proper environment.

Workflow specifications

Input

• DNA sequences into bed format, split into positive set and negative set, written into the bed directory.

Note : if you want add a new dataset, edit the Snakefile file and add the bed files in the dictionnary EXPERIMENTS, without the .bed extension. Example :

TestSet_Peaks_BG4_G4seq_HaCaT_GSE76688_hg19_201b_Ctrl_gkmSVM_0.8_42_Ctrl_gkmSVM.bed TestSet_Peaks_BG4_G4seq_HaCaT_GSE76688_hg19_201b_Ctrl_gkmSVM_0.8_42.bed

EXPERIMENTS = {
  "TestSet_Peaks_BG4_G4seq_HaCaT_GSE76688_hg19_201b_Ctrl_gkmSVM_0.8_42_Ctrl_gkmSVM":{"CTRL":"TestSet_Peaks_BG4_G4seq_HaCaT_GSE76688_hg19_201b_Ctrl_gkmSVM_0.8_42_Ctrl_gkmSVM","EXP":"TestSet_Peaks_BG4_G4seq_HaCaT_GSE76688_hg19_201b_Ctrl_gkmSVM_0.8_42"}
}

Where CTRL is the negative set and EXP is the positive set.

• DNA Accessibility (ATAC-seq/DNAse-seq/MNase-seq) in bigwig format or directly the averaged value for each sequence in a one-column tsv file.

ATACFILE = {
    "TestSet_Peaks_BG4_G4seq_HaCaT_GSE76688_hg19_201b_Ctrl_gkmSVM":["ATAC_entinostat_mean.bw"]
}

or one-column tsv file in fasta/{Experiment_name}/{Experiment_name}_atac_merged.tsv. Example :

fasta/TestSet_Peaks_BG4_G4seq_HaCaT_GSE76688_hg19_201b_Ctrl_gkmSVM/TestSet_Peaks_BG4_G4seq_HaCaT_GSE76688_hg19_201b_Ctrl_gkmSVM_atac_merged.tsv

head TestSet_Peaks_BG4_G4seq_HaCaT_GSE76688_hg19_201b_Ctrl_gkmSVM_atac_merged.tsv 
0.01628741641898675
0.028752257447422012
0.028878783223623482
0.055516399884055316
0.02825982069785745
0.03582923041809851
0.023904436394151577
0.07724288611280039
0.01740800116454673
0.05779605688479145

Rulegraph :

Workflow output for each tools :

Outputs	Tools	Methods
ATACDeepG4_ATACnormBG	ATACDeepG4	DeepG4 using accessibily (DeepG4 in paper)
ATACDeepG4_classictuningOH5	ATACDeepG4	DeepG4 without accessibility (DeepG4* in paper)
penguinn_retrained	penguinn	penguinn using custom model trained on BG4G4seq dataset
penguinn	penguinn	penguinn using default model
G4detector_retrained	G4detector	G4detector using custom model trained on BG4G4seq dataset
G4detector	G4detector	G4detector using default model
quadron_retrained	quadron	quadron using custom model trained on BG4G4seq dataset
quadron_score	quadron	quadron using default model