RNome Minimal Modification Analysis Pipeline

Nextflow pipeline used for aligning nanopore modBAM files using minimap2 (preserving modification tags) and summarising modifications using modkit.

Data processing notes:

The output (raw) modBAMs from MinKNOW were first merged and sorted using samtools (1.19.1):

 samtools merge -  data/bam_pass/*.bam -@ 16 | samtools sort -@ 16 -o PBG54229_pass_2aeee96b_4a0fe7c1_merged.bam

Overview

The pipeline:

1. Converts modBAMs to fastq files preserving RNA modification tags using samtools and aligns to a reference (genome or transcriptome) using minimap2
2. Runs 'pileup' (summarises) RNA modifications on merged BAMs with modkit & outputs as bedRmod format. Specifically m6A, m5C, inosine & pseudouridine mods.

The pipeline is written to run on an HPC cluster, or AWS Server/EC2 instance with all data locally available on the server and requires Nextflow (≥22.04) and Docker.

Running the pipeline

Uses a samplesheet to specify inputs (modbam paths, sample_id and reference against which to align the data). Use unique sample_id values to avoid filename collisions. The same sample can be aligned to different references by using different row entries.

Example samplesheet.csv:

sample_id,reference_type,reference,bam
sample1_gm,genome,/path/to/GRCh38.fa.gz,/path/to/sample1.bam
sample1_tx,transcriptome,/path/to/gencode.fa.gz,/path/to/sample1.bam

Run:

nextflow run main.nf \
    -profile aws_ec2,docker \
    --samplesheet samplesheet.csv \
    --outdir results/

Includes a convenience script to run the nextflow command on an EC2 instance or an HPC cluster using slurm scheduler & singularity.

bash run_ec2.sh

sbatch run_slurm.sh

Output Structure

results/
├── alignment/              # Individual aligned BAMs
│   ├── sample1_gm.bam
│   └── sample1_tx.bam
├── bedRmod/               # Modification calls
│   ├── sample1_gm_bedRmod.bed.gz
│   ├── sample1_tx_bedRmod.bed.gz
├── reference/             # Processed references
└── pipeline_info/         # Execution reports