https://mbhall88.github.io/tags/clair3/Recent content in Clair3 on Microbes made me do itHugoen-auWed, 22 Apr 2026 14:16:33 +1000https://mbhall88.github.io/post/minimap2-lrhq-preset-testing/Wed, 22 Apr 2026 14:16:33 +1000https://mbhall88.github.io/post/minimap2-lrhq-preset-testing/<h2 id="evaluating-minimap2s-lrhq-preset-for-bacterial-nanopore-variant-calling">Evaluating minimap2&rsquo;s <code>lr:hq</code> preset for bacterial nanopore variant calling</h2> <h2 id="introduction">Introduction</h2> <p>Oxford Nanopore Technologies (ONT) sequencing accuracy has improved dramatically in recent years. With basecalling models like <a href="https://github.com/nanoporetech/dorado">Dorado</a> v5.2.0 super-accuracy (<code>sup</code>), error rates are consistently hovering around the 1% mark. To match this shift in raw read quality, <a href="https://github.com/lh3/minimap2/"><code>minimap2</code></a>[<a href="#references" class="citation-link">1</a>] introduced the <code>lr:hq</code> preset in <a href="https://github.com/lh3/minimap2/releases/tag/v2.27">version 2.27 (March 2024)</a>, which is calibrated for long reads with an error rate of &lt;1%.</p> <p>This introduction was driven by internal benchmarking from ONT developers (see <a href="https://github.com/lh3/minimap2/issues/1127">minimap2 issue #1127</a>) who found that <code>-x map-ont -k19 -w19 -U50,500</code> maximised both speed and downstream accuracy for high-quality reads. As such, the <code>lr:hq</code> preset was added to mirror those options.</p>