A nextflow workflow for detection of microbial reads from cancer whole-genome-sequenced bams.

Based on the Ge et al. 2025 manuscript, modified to work from oncoanalyser bam files.

We start from an oncoanalyser aligned bam, extract all unmapped reads, realign to CHM13 with bowtie, pull unmapped reads once more and run kraken-uniq to classify reads against the microbial2023 database.

Start by downloading the reference genomes and databases this pipeline will use.

Download and prepare bowtie indices for CHM13v2.0 genome.

wget https://s3-us-west-2.amazonaws.com/human-pangenomics/T2T/CHM13/assemblies/analysis_set/chm13v2.0.fa.gz

gzip -dc chm13v2.0.fa.gz | bgzip -c > chm13v2.0.fa.bgz

bowtie2-build chm13v2.0.fa.bgz

Download microbial2023 krakenuniq database

wget "https://genome-idx.s3.amazonaws.com/kraken/uniq/krakendb-2023-08-08-MICROBIAL/kuniq_microbialdb_minus_kdb.20230808.tgz"
wget "https://genome-idx.s3.amazonaws.com/kraken/uniq/krakendb-2023-08-08-MICROBIAL/database.kdb"

Untar the database & move database.kdb file into database folder

tar -xvf kuniq_microbialdb_minus_kdb.20230808.tgz

mv database.kdb kuniq_microbialdb_minus_kdb.20230808

nextflow run selkamand/micrite-screen \
  --bam sample.bam \
  --decoys contigs.txt \
  --ref /path/to/bowtie2/index/prefix/chm13v2.0.fa \
  --kraken_db /path/to/krakenuniq_db \
  --threads 8 \
  --threads_kraken 4 \
  --bowtie2_preset sensitive \
  --preload_size 4G \
  --outdir results

Because there are so many options, its much easier to specify the paramaters using a yaml file. For example within the root directory of this repo you could run

nextflow run . -profile docker -params-file params.yaml

All files are written to --outdir.

Pre-alignment unmapped reads

• <sample>.pre_R1.fastq.gz
• <sample>.pre_R2.fastq.gz Unmapped read pairs extracted from the original BAM.

CHM13 re-alignment

• <sample>.sorted.bam
• <sample>.sorted.bam.bai Coordinate-sorted BAM and index after re-alignment to CHM13.

Post-alignment unmapped reads

• <sample>.post_R1.fastq.gz
• <sample>.post_R2.fastq.gz Read pairs still unmapped after CHM13 alignment (input to KrakenUniq).

Taxonomic classification

• <sample>.krakenuniq.report.txt KrakenUniq taxonomic classification report for final unmapped reads.

From inside this directory.

Build local version for OSX

docker buildx build --platform linux/arm64 --load --tag selkamandcci/micrite-screen:0.0.2 .

Build final version to push to dockerhub

docker buildx build --push --platform linux/amd64,linux/arm64 --tag selkamandcci/micrite-screen:0.0.2 .

This repo includes some small files for testing. If you git clone the repo you should be able to run the following.

You may need to change -profile from singularity to docker if running on MacOS

nextflow run . -profile singularity \
  --ref testfiles/data/ref/ref.fa \
  --kraken_db testfiles/data/dbs/minuku/ \
  --bam testfiles/data/aligned.no_decoy.bam \ 
  --decoys testfiles/data/decoys.txt \
  --outdir outdir 

We also supply a test profile to allow testing with:

nextflow run . -profile test

Or using nf-test

nf-test test