Seqsum

Robust individual and aggregate checksums for nucleotide sequences. Accepts data from either standard input or fast[a|q][.gz|.zst|.xz|.bz2] files. Generates individual checksums for each sequence, and an aggregate checksum-of-checksums for a collection of sequences. Warnings are shown for duplicate sequences and within-collection checksum collisions at the chosen bit depth. Sequences are uppercased prior to hashing with xxHash (xxh3_128) and may optionally be normalised (with -n) to use only the characters ACGTN-. Read IDs and base quality scores do not inform the checksum. Outputs TSV or JSON to stdout.

A typical use case is for determining whether reordered, renamed or otherwise bit-inexact fasta/fastq files have equivalent sequence composition. Another use is generating the shortest possible collision-free identifiers for sequence collections.

Install (Python 3.10+)

# Latest release
pip install seqsum

# Latest commit
git clone https://github.com/bede/seqsum
pip install ./seqsum

# Development
git clone https://github.com/bede/seqsum.git
cd seqsum
pip install --editable '.[dev]'
pytest

CLI usage

# Fasta with one record
$ seqsum nt MN908947.fasta
MN908947.3	ca5e95436b957f93

# Fasta with two records
$ seqsum nt MN908947-BA_2_86_1.fasta
MN908947.3	ca5e95436b957f93
BA.2.86.1		d5f014ee6745cb77
aggregate	837cfd6836b9a406

# Fastq (gzipped) with 1m records, redirected to file, with progress bar
$ seqsum nt illumina.r12.fastq.gz --progress > checksums.tsv
Processed 1000000 records (317kit/s)
INFO: Found duplicate sequences

# Fasta via stdin
% cat MN908947.fasta | seqsum nt -
MN908947.3	ca5e95436b957f93

$ seqsum nt -h                       
usage: seqsum nt [-h] [-n] [-s] [-b BITS] [-i] [-a] [-j] [-p] input

Robust individual and aggregate checksums for nucleotide sequences. Accepts input
from either stdin or fast[a|q][.gz|.zst|.xz|.bz2] files. Generates individual
checksums for each sequence, and an aggregate checksum-of-checksums for a
collection of sequences. Warnings are shown for duplicate sequences and
within-collection checksum collisions at the chosen bit depth. Sequences are
uppercased prior to hashing with xxHash and may optionally be normalised to use only
the characters ACGTN-. Read IDs and base quality scores do not inform the checksum

positional arguments:
  input                 path to fasta/q file (or - for stdin)

options:
  -h, --help            show this help message and exit
  -n, --normalise       replace U with T and characters other than ACGT- with N
                        (default: False)
  -s, --strict          raise error for characters other than ABCDGHKMNRSTVWY-
                        (default: False)
  -b BITS, --bits BITS  displayed checksum length
                        (default: 64)
  -i, --individual      output only individual checksums
                        (default: False)
  -a, --aggregate       output only aggregate checksum
                        (default: False)
  -j, --json            output JSON
                        (default: False)
  -p, --progress        show progress and speed
                        (default: False)

Python usage

from seqsum import lib

fasta_contents = ">read1\nACGT"
checksums, aggregate_checksum = lib.sum_nt(fasta_contents)
print(checksums)

from pathlib import Path
from seqsum import lib

checksums, aggregate_checksum = lib.sum_nt(Path("tests/data/MN908947-BA_2_86_1.fasta"))
print(checksums, aggregate_checksum)