Skip to content

Streaming alignment with bwa

sracha can stream FASTQ directly to an aligner via -Z, eliminating intermediate files and reducing disk I/O.

Prerequisites

Install bwa, samtools, and ucsc-twobittofa:

1
2
pixi add bwa samtools ucsc-twobittofa
# or: conda install -c bioconda bwa samtools ucsc-twobittofa

Quick start: align to chr22

Stream chr22 from the hs1 (T2T-CHM13v2.0) 2bit file, gzip it, and index:

1
2
3
twoBitToFa -seq=chr22 https://hgdownload.soe.ucsc.edu/goldenPath/hs1/bigZips/hs1.2bit stdout \
  | gzip > chr22.fa.gz
bwa index chr22.fa.gz

Stream directly from NCBI to a sorted BAM — no intermediate files:

1
2
3
4
sracha get SRR28588231 --split interleaved -Z \
  | bwa mem -p -t 8 chr22.fa.gz - \
  | samtools sort -@ 4 -o SRR28588231.chr22.bam
samtools index SRR28588231.chr22.bam

sracha downloads the SRA file to a hidden temp file, streams interleaved FASTQ to stdout, then auto-deletes the temp file. No .sra or .fastq.gz files are left on disk.

Verify:

1
samtools flagstat SRR28588231.chr22.bam

Full human genome workflow

Download and index hs1:

1
2
wget https://hgdownload.soe.ucsc.edu/goldenPath/hs1/bigZips/hs1.fa.gz
bwa index hs1.fa.gz    # ~1 hour, ~5 GB RAM

Stream alignment:

1
2
3
4
sracha get SRR28588231 --split interleaved -Z \
  | bwa mem -p -t 8 hs1.fa.gz - \
  | samtools sort -@ 4 -o SRR28588231.bam
samtools index SRR28588231.bam

Two-step workflow

If you already have an SRA file on disk, use fastq -Z:

1
2
3
sracha fastq SRR28588231.sra --split interleaved -Z \
  | bwa mem -p -t 8 hs1.fa.gz - \
  | samtools sort -@ 4 -o SRR28588231.bam

How it works

  • -Z streams uncompressed FASTQ to stdout; pair it with --split interleaved so paired reads come out as a single interleaved stream
  • bwa mem -p reads interleaved paired-end FASTQ from stdin
  • samtools sort reads SAM from stdin and writes a coordinate-sorted BAM
  • Backpressure flows naturally through the Unix pipe

Performance tips

  • Split threads between sracha and bwa (e.g., -t 4 for sracha, -t 8 for bwa on 12 cores)
  • Streaming avoids writing intermediate FASTQ files (saves disk I/O and space)
  • For maximum throughput, use --format sralite if quality scores aren't critical