reset-bio 1.3.0

Reference genome Selection Tool

ReSeT — Reference genome Selection Tool

ReSeT selects a minimal, representative subset of sequences from a pre-clustered collection. Given a pairwise distance matrix and cluster labels, it minimises an objective that penalises (1) the cost of selecting each sequence, (2) within-cluster coverage gaps, and (3) cross-cluster similarity between selected sequences. The result is a compact reference set that covers every taxon while avoiding redundancy.

Installation

pip install reset-bio

Quick start

import numpy as np
from reset import Solution

# Distance matrix (n × n) and cluster assignments (integer, length n)
D = np.array([[0.0, 0.2, 0.8], [0.2, 0.0, 0.7], [0.8, 0.7, 0.0]])
clusters = np.array([0, 0, 1])

sol = Solution(D, clusters, selection_cost=0.1)
sol.local_search(max_runtime=10)
print(np.flatnonzero(sol.selection))   # indices of selected sequences

Command-line interface

reset \
  --clusters     clusters.csv \
  --distances    distances.tsv \
  --distance_format mash \
  --selection_cost 1e-1 \
  --scale 1e-5 \
  --output selected.txt

Supported distance formats: mash, sourmash_cosine, sourmash_jaccard, generic.

Distance formats

Format	Description
mash	Mash lower-triangular output (mash triangle)
sourmash_cosine	Sourmash CSV (column 12 = cosine distance)
sourmash_jaccard	Sourmash CSV (column 6 = Jaccard distance)
generic	Any delimited (i, j, distance) file

Multiprocessing

Use Solution_shm for parallel local search:

from reset import Solution_shm
sol = Solution_shm(D, clusters, selection_cost=0.1)
sol.local_search(num_processes=4, max_runtime=60)

Dashboard

An interactive Streamlit dashboard is included. Install with dashboard extras and launch:

pip install "reset-bio[dashboard]"
streamlit run $(python -c "import reset.dashboard.app as a; import os; print(os.path.abspath(a.__file__))")

The dashboard defaults to the bundled toy example — no configuration needed.

License

MIT — see LICENSE.