KSSD3 is an alignment-free toolkit for fast comparison of genomic sequences. It uses probabilistic sampling of the k-mer space to build compact sketches and compute distances, containment metrics, ANI estimation, and metagenomic composition.

The repository ships with a Makefile and a precompiled copy of XGBoost. To build the kssd3 binary run:

make

The build requires GCC with OpenMP support and zlib. When the build completes, the executable is placed in bin/kssd3.

To make the bundled XGBoost library discoverable, append it to LD_LIBRARY_PATH:

make install_env
source ~/.bashrc

Optionally install the binary into /usr/local/bin with:

sudo make install

kssd3 exposes several subcommands:

• shuffle – generate dimension-reduction shuffle files.
• sketch – create sketches from FASTA/FASTQ sequences.
• dist – compute distances between sketches.
• set – union/intersection/subtraction operations on sketches.
• reverse – recover k-mers from sketches.
• composite – estimate metagenomic composition.
• matrix – compute distance matrices.
• ani – estimate average nucleotide identity (ANI).

Display help for any subcommand with --help or --usage.

# create shuffle file if needed
./bin/kssd3 shuffle --usedefault -o default

# sketch reference and query genomes
./bin/kssd3 sketch -o ref_sketches refs/*.fasta
./bin/kssd3 sketch -o qry_sketches queries/*.fasta

# build inverted index for reference sketches
./bin/kssd3 sketch -i ref_sketches

# estimate ANI
./bin/kssd3 ani -r ref_sketches -q qry_sketches -m 0 > ani.tsv

Command line options for each subcommand are provided by the built-in help system. For license details see LICENSE.txt. The ani_models directory contains example XGBoost models for ANI prediction.