COMEBin metagenomic binning benchmark dataset — small test set (300 contigs, 94 MB)

This dataset is the small functional-test set of the metagenomic-binning-benchmark project (https://github.com/paulzierep/metagenomic-binning-benchmark), a repeatable benchmark harness for metagenome binners (starting with COMEBin) that records per-run parameters, wall time and peak RAM, and scores bin quality with both CheckM2 and CheckM v1. What this benchmark was used for: the small gate of the optimization workflow. Any candidate COMEBin fix or parameter change is first validated on this set — the small run must complete and produce non-empty bins with quality metrics — before it is benchmarked on the medium 3,000-contig set (comebin_medium) and, only after small + medium validation, on the full-large datasets (demo / CAMI II / CAMI III). Verified results on this set: small_test_v4 produced 3 bins in 122 s, CheckM2 mean completeness / contamination 26.07 % / 2.02 %, CheckM v1 29.08 % / 0.71 %. Contents (6 core files, all checksums in the public manifest): the reduced BATS contig assembly (BATS_SAMN07137077_METAG — the 300 longest contigs, >= 1001 bp), the overlapping paired-end reads and the coverage BAM with index, the COMEBin v1.1.0 result (3 non-empty bins plus training / config logs), and the SHA-256 public manifest. Produced from COMEBin source commit c8f22e4935a66a0fd58b794489a9021671b15437. How to run: clone https://github.com/paulzierep/metagenomic-binning-benchmark, prepare the environment per docs/00-setup.md (micromamba + comebin environment), register the dataset per docs/01-datasets.md, then launch the small run with scripts/run_small_test.sh and evaluate it with scripts/run_eval.sh. Provenance, the release checklist and checksum-verification steps are documented in docs/10-data-preservation.md; the full per-run results table and run history live in docs/11-run-results.md and results/. This record and the benchmark project are maintained by Paul Zierep. License: MIT. DOI: 10.5281/zenodo.22935025.