IntegrateALL is a machine learning based pipeline for multilevel-data extraction and subtype classification in B-cell precursor ALL. It provides the user with an interactive report, containing subtype prediction, virtual karyotype, fusions, quality control, single nucleotide variants using a snakemake workflow management system.

• Machine Learning Classification: Automated subtype prediction for B-cell precursor ALL using ALLCatchR and automated karyotype prediction using KaryALL
• Comprehensive Analysis: Fusion detection, CNV analysis, variant calling, and quality control
• Interactive Reporting: HTML reports with visualizations and predictions
• Two-Workflow Architecture: Separate setup and analysis workflows for improved stability
• Automatic Classification: Classification according to WHO-HAEM5 and ICC classification

IntegrateALL uses a two-workflow architecture with multiple Snakefiles:

1. Setup Workflow (setup.smk): One-time installation of reference data and tools (~21GB)
2. Analysis Workflows:
   • Snakefile: Standard analysis workflow for local/single-node execution
   • Snakefile.cluster: Optimized workflow for cluster environments with job-specific resource allocation

• Linux operating system (tested on Ubuntu/CentOS)
• Snakemake >= 7.3
• Conda/Mamba for dependency management
• 50GB free disk space (21GB for references + analysis space)
• Minimum 50GB RAM for STAR alignment

# Download and install Miniconda
cd ~/software
wget https://repo.anaconda.com/miniconda/Miniconda3-py311_23.11.0-2-Linux-x86_64.sh
sh Miniconda3-py311_23.11.0-2-Linux-x86_64.sh

# Install mamba for faster dependency resolution
conda install mamba=1.5.8 -n base -c conda-forge

# Clone the repository
git clone https://github.com/NadineWolgast/IntegrateALL.git
cd IntegrateALL

# Or download and extract the zip file
wget https://github.com/NadineWolgast/IntegrateALL/archive/refs/heads/main.zip
unzip main.zip && cd IntegrateALL-main

# Create and activate the conda environment
conda activate base
mamba env create --name integrateall --file environment.yaml
conda activate integrateall

Edit the config.yaml file to match your system:

absolute_path: /absolute/path/to/IntegrateALL   # No trailing slash!
star_mem: 50000    # Minimum 50GB RAM for STAR
threads: 4         # Adjust to your CPU cores

⚠️ Important: Use absolute paths without trailing slashes to avoid errors.

Install all required reference data and tools:

snakemake --snakefile setup.smk --cores 4 --use-conda --conda-frontend conda

⚠️ Important: Always use --use-conda to ensure correct tool versions are used for reference data creation.

This setup downloads (~21GB total):

• Reference genome and annotations (~16GB)
• STAR genome index (~1GB)
• RNAseqCNV reference data (~50MB)
• FusionCatcher database (~4.4GB)
• R packages (ALLCatchR, RNAseqCNV)
• Arriba draw_fusions tool (~10MB)

⏱️ Time: 1-3 hours (depending on internet speed)

Setup only runs once - subsequent executions skip existing components.

1. Prepare sample sheet: Edit samples.csv with your FASTQ file paths:

sample_id,left,right
Sample_01,/path/to/sample_01_R1.fastq.gz,/path/to/sample_01_R2.fastq.gz
Sample_02,/path/to/sample_02_R1.fastq.gz,/path/to/sample_02_R2.fastq.gz

2. Test with provided data (optional):

sample_id,left,right
Test,/path/to/IntegrateALL/data/samples/sub1_new.fq.gz,/path/to/IntegrateALL/data/samples/sub2_new.fq.gz

Choose the appropriate Snakefile for your environment:

• Snakefile: Standard workflow for local execution or single-node processing
• Snakefile.cluster: Standard cluster workflow that always runs both Arriba and FusionCatcher for comprehensive fusion detection
• Snakefile.conditional: Optimized cluster workflow with Arriba-first approach that only runs FusionCatcher if initial classification fails, significantly reducing runtime for samples with clear driver fusions

1. Dry run (preview jobs):

snakemake -n --use-conda --conda-frontend conda

2. Full analysis:

snakemake --cores 20 --use-conda --conda-frontend conda

Note: The --conda-frontend conda flag prevents libmamba-related errors during conda environment creation.

3. Single sample:

snakemake --cores 4 --use-conda --conda-frontend conda STAR_output/YOUR_SAMPLE_ID/Aligned.sortedByCoord.out.bam

Option 1: Standard cluster workflow (comprehensive fusion detection)

snakemake -s Snakefile.cluster --cores 20 --use-conda --conda-frontend conda

Option 2: Optimized cluster workflow (Arriba-first approach)

snakemake -s Snakefile.conditional --cores 20 --use-conda --conda-frontend conda

Recommendation: Use Snakefile.conditional for faster processing when you expect clear driver fusions. Use Snakefile.cluster for comprehensive analysis when you need maximum sensitivity or are analyzing challenging samples.

Option 3: SBATCH submission script

First, configure the submission script for your cluster:

# Edit submit_snakemake_job.sh to match your cluster configuration
cp submit_snakemake_job.sh my_submit_script.sh
# Adjust partition name, memory, CPU count, and conda paths

Then submit the job:

sbatch my_submit_script.sh

Option 4: Simple SLURM:

srun -c 20 --mem 100G snakemake --cores 20 --use-conda --conda-frontend conda

Option 5: SLURM Executor:

snakemake --slurm --default-resources mem_mb=5000 threads=4 slurm_partition=YOUR_PARTITION --jobs 200 --use-conda --conda-frontend conda --keep-going

• interactive_output/{SAMPLE_ID}/output_report_{SAMPLE_ID}.html

• Alignments: STAR_output/{SAMPLE_ID}/
• Fusions: fusions/{SAMPLE_ID}.pdf, fusioncatcher_output/{SAMPLE_ID}/
• Fusion Intersects: data/fusion_intersect/{SAMPLE_ID}.csv
• Variants: Variants_RNA_Seq_Reads/{SAMPLE_ID}/
• CNV Analysis: RNAseqCNV_output/{SAMPLE_ID}/
• Classification: allcatch_output/{SAMPLE_ID}/predictions.tsv
• Karyotype Prediction: karyotype_prediction/{SAMPLE_ID}.csv
• Quality Control: qc/fastqc/{SAMPLE_ID}/, qc/multiqc/{SAMPLE_ID}/
• Individual Classification: Final_classification/{SAMPLE_ID}_output_report.csv
• Aggregated Summary: Final_classification/Aggregated_output_curation.csv

• TPM: data/tpm/{SAMPLE_ID}.tsv
• CPM: data/cpm/{SAMPLE_ID}.tsv
• Counts: data/counts/{SAMPLE_ID}.tsv
• Raw counts: STAR_output/{SAMPLE_ID}/ReadsPerGene.out.tab
• Combined counts: data/combined_counts/ensemble_counts.tsv, data/combined_counts/gene_counts.tsv

Run only specific analysis modules by editing the rule all section in Snakefile:

rule all:
    input:
        "check_samples.txt",
        expand("fusioncatcher_output/{sample_id}/final-list_candidate-fusion-genes.txt", 
               sample_id=list(samples.keys())),
        # Uncomment desired outputs:
        # expand("STAR_output/{sample_id}/Aligned.sortedByCoord.out.bam", sample_id=list(samples.keys())),
        # expand("allcatch_output/{sample_id}/predictions.tsv", sample_id=samples.keys()),
        # expand("interactive_output/{sample}/output_report_{sample}.html", sample=list(samples.keys()))

# Deactivate environment
conda deactivate

# Reactivate for analysis
conda activate integrateall

1. "Snakefile not found": Ensure you're in the IntegrateALL directory
2. Memory errors: Increase star_mem in config.yaml (minimum 50000)
3. Permission denied: Check file paths and permissions in samples.csv
4. Environment conflicts: Remove and recreate the conda environment

For issues and bug reports, please use the GitHub issue tracker.

If you use IntegrateALL in your research, please cite: https://doi.org/10.1101/2025.09.25.673987