taf-diamond

TAFFISH wrapper for DIAMOND, a fast protein sequence aligner for BLAST-like protein search, translated DNA-to-protein search, DIAMOND database creation, DAA archive handling, and protein clustering.

This repository packages upstream DIAMOND release 2.2.5 as the TAFFISH package 2.2.5-r1.

Installation

Install from the public TAFFISH Hub index:

taf update
taf install diamond

Install the exact release:

taf install diamond 2.2.5-r1

For local testing before the app is published to the public index:

taf install --from .

Usage

Show TAFFISH app help:

taf-diamond --help

Show the TAFFISH package version:

taf-diamond --version

Show upstream DIAMOND version:

taf-diamond diamond version

Show upstream command help:

taf-diamond diamond help
taf-diamond diamond blastp
taf-diamond diamond makedb

Create a DIAMOND protein database:

taf-diamond diamond makedb --in proteins.faa -d proteins --threads 8

Search protein queries against a DIAMOND database:

taf-diamond diamond blastp \
  -d proteins \
  -q query.faa \
  -o blastp.tsv \
  --threads 8 \
  --outfmt 6 qseqid sseqid pident length evalue bitscore

Search translated nucleotide queries against a protein database:

taf-diamond diamond blastx \
  -d proteins \
  -q transcripts.fa \
  -o blastx.tsv \
  --threads 8 \
  --outfmt 6 qseqid sseqid pident length evalue bitscore

Write a DIAMOND alignment archive and convert it later:

taf-diamond diamond blastp -d proteins -q query.faa -a matches --threads 8
taf-diamond diamond view -a matches.daa -o matches.tsv

For portable, simple workflows, direct tabular output with --outfmt 6 is the recommended default. This TAFFISH package checks a tiny DAA archive plus diamond view conversion in Docker validation and smoke.

Inspect or extract sequences from a DIAMOND database:

taf-diamond diamond dbinfo -d proteins.dmnd
taf-diamond diamond getseq -d proteins.dmnd -o proteins.from-db.faa

Cluster protein sequences:

taf-diamond diamond linclust -d proteins.faa -o clusters.tsv --threads 8 --approx-id 90
taf-diamond diamond linclust -d proteins.faa -o clusters.tsv --reps reps.faa --threads 8 --approx-id 90
taf-diamond diamond cluster -d proteins.faa -o clusters.tsv --threads 8

Run database self-alignment without a separate query file:

taf-diamond diamond blastp \
  -d proteins \
  --self \
  -o self.tsv \
  --threads 8 \
  --outfmt 6 qseqid sseqid pident length evalue bitscore

DIAMOND 2.2.4 introduced --single-step for controlled orchestration of distributed clustering. Upstream requires it to be used together with --parallel-tmpdir; it is an advanced multi-node control rather than a local one-command clustering shortcut.

DIAMOND 2.2.5 adds optional reseeding controls for search workflows:

taf-diamond diamond blastp \
  -d proteins \
  -q query.faa \
  -o blastp-reseek.tsv \
  --threads 8 \
  --reseek-diags \
  --reseek-xdrop 20 \
  --word-threshold 10 \
  --double-hit-window 40 \
  --outfmt 6 qseqid sseqid pident length evalue bitscore

These controls are advanced tuning options. Keep upstream defaults unless the analysis has a documented reason to change reseeding behavior.

Because this is a command-mode TAFFISH tool, the first non-option argument is the in-container command. DIAMOND's actual executable is named diamond, so the clearest form is:

taf-diamond diamond blastp ...
taf-diamond diamond blastx ...
taf-diamond diamond makedb ...
taf-diamond diamond linclust ...

Do not use taf-diamond blastp ... as the normal form. In command mode, TAFFISH will interpret blastp as an executable inside the container, not as a subcommand of diamond.

The -- separator is mainly useful for option-leading arguments to the default diamond command:

taf-diamond -- --help
taf-diamond -- --version

For DIAMOND commands such as blastp, makedb, view, or linclust, use the explicit taf-diamond diamond <command> ... form.

This README lists common usage patterns, not the full upstream manual. The TAFFISH wrapper calls upstream diamond directly, so official DIAMOND commands and options are available as upstream implements them. Use taf-diamond diamond help, taf-diamond diamond <command>, or the upstream manual for the complete option list.

Package

name: diamond
command: taf-diamond
version: 2.2.5-r1
kind: tool
image: ghcr.io/taffish/diamond:2.2.5-r1
upstream release: v2.2.5
runtime version: diamond version 2.2.5
help build banner: diamond v2.2.5.185

Container

The container image is built from docker/Dockerfile. It starts from debian:12-slim, downloads the official upstream v2.2.5 diamond-linux64 binary archive from GitHub, verifies it with a pinned sha256 checksum, and keeps the upstream binary plus documentation/license files extracted from the matching source tarball.

The runtime image includes these user-facing commands and runtime tools:

diamond
sh

The official Linux binary used here is dynamically linked against the standard glibc runtime libraries present in Debian. It does not require Python, R, Perl, or external bioinformatics helper executables for the core workflows tested here.

The official diamond-linux64 binary is used as released by upstream. Docker validation and smoke for this release also exercise a minimal --compress zstd search path.

DIAMOND 2.2.5 improves clustering and sequence-loading performance, fixes additional clustering memory growth and non-determinism, and fixes inter-toolchain inconsistency in sketch-based search. It also fixes invalid line breaks in qtitle output when the query input is FASTQ.

The linear-time clustering stages now use banded extension by default. This is an upstream behavior change that can affect performance and clustering output. Use --ext full when a workflow intentionally needs the previous full-matrix extension behavior. The release adds --reseek-diags, --reseek-xdrop, --word-threshold, and --double-hit-window, and removes --linclust-banded-ext and --lin-combo.

The preceding 2.2.4 release added --self and --single-step, improved clustering and FASTA loading, and included ARM64 source-build fixes. This wrapper preserves those upstream behaviors. Smoke runs a tiny self-alignment, checks the distributed --single-step option surface, and exercises the new reseek controls without fabricating a multi-node environment.

The preceding 2.2.3 release changed clustering defaults to BLAST-like compositional matrix adjustment and conservative SEG masking and added options including --comp-based-stats 6, --relative-entropy-tolerance, --taxdump, --masking seg-all, --reps, --symmetrize-evalue, and --fpu-compat.

This package intentionally provides the normal upstream DIAMOND executable and does not add a custom TAFFISH entrypoint. The upstream command surface includes listed workflows such as makedb, blastp, blastx, view, merge-daa, getseq, dbinfo, cluster, linclust, and other upstream-provided commands.

The image is the official upstream CPU linux64 build. DIAMOND does not require GPU runtime options for the workflows packaged here. This release is native linux/amd64 only because upstream does not publish a matching official Linux arm64 binary for v2.2.5. Upstream source supports ARM64, but the release asset packaged here remains an x86-64 ELF binary. The TAFFISH wrapper requests --platform linux/amd64 for Docker and Podman, so Apple Silicon and other arm64 hosts can still use the package through Docker/Podman amd64 emulation. Do not interpret that as native arm64 support.

The official binary does not enable the upstream EXTRA development command set, so development-only commands such as blastn, random-seqs, or smith-waterman are outside this TAFFISH package. If an upstream maintenance or development command is listed but unavailable in the official binary, this package preserves that upstream behavior rather than wrapping around it.

The image is built and validated for:

linux/amd64

The TAFFISH metadata declares a Docker smoke check:

exist: diamond, sh
test:  diamond version reports 2.2.5 and help reports build 2.2.5.185
test:  top-level help plus blastp, linclust, --self, --single-step, --reps, reseek controls, and view option surfaces are available
test:  makedb, dbinfo, blastp, reseek execution, zstd output, DAA writing, and diamond view conversion work
test:  tiny FASTA and FASTQ blastx searches produce tabular hits, with qtitle kept on one line
test:  a tiny linclust clustering run produces a cluster table and representatives FASTA
test:  a tiny database self-alignment produces a tabular hit

During TAFFISH Hub indexing, this smoke metadata verifies that the published image exposes the expected command surface, reports the pinned upstream version, includes the runtime pieces used by common workflows, and can run representative local database, search, DAA conversion, translated search, and clustering tasks. It does not download remote databases or exhaustively validate every DIAMOND option.

Each smoke command is self-contained because the public index runs every [smoke].test entry in a fresh temporary container. No smoke entry depends on files created by a previous entry.

Upstream

source:  https://github.com/bbuchfink/diamond
release: https://github.com/bbuchfink/diamond/releases/tag/v2.2.5
license: GPL-3.0-or-later

The TAFFISH app wrapper, Dockerfile, and documentation in this repository are released under Apache-2.0. The bundled upstream DIAMOND software is distributed under the upstream GPL-3.0-or-later license; keep that distinction in mind when redistributing derived images.

Please cite DIAMOND when using this package in scientific work:

Buchfink B, Reuter K, Drost HG. Sensitive protein alignments at tree-of-life
scale using DIAMOND. Nature Methods. 2021. doi:10.1038/s41592-021-01101-x.
PMID: 33828273.

For DIAMOND DeepClust workflows, upstream also lists:

Buchfink B, Xie C, Huson DH. Ultrafast and sensitive protein clustering using
DIAMOND DeepClust. Nature Methods. 2026. doi:10.1038/s41592-026-03030-z.
PMID: 41876643.