Interest in optional tools for faster processing and methylation QC? #643

Would you welcome contributions adding any of these optional tools? I'm involved in their development and happy to help:

• chelae: faster adapter trimming with better trim-point accuracy in published simulations, including sensitivity to short adapter remnants through paired-read overlap. It produces fastp-compatible reports for MultiQC.
• methylsieve: conversion-failure filtering for BS/EM-seq, plus learned M-bias masking and control metrics for BS/EM-seq and TAPS. Its bounded-prefix learning could help with the M-bias discussion in Add bwa-mem3 (TAPS and bisulfite/EM-seq) and rastair version 2 #638.
• mako: faster SAM/BAM sorting using fgumi's sorting engine. In our large-WGS benchmark, coordinate sorting was approximately 2.5x faster than samtools sort, with validated equivalent output.
• dupblaster: substantially faster duplicate marking than Picard in published WGS benchmarks, with directional methylation support that keeps original top/bottom strands distinct. It processes aligner output before sorting.
• riker: faster sequencing QC, including capture metrics as an alternative to Picard CollectHsMetrics, with multiple collectors sharing one BAM pass. Published benchmarks show 12-38x speedups over corresponding Picard tools.

I'd start with the bwa-mem3 path, keep defaults unchanged, and validate performance and downstream methylation results on converted libraries rather than assume the published gains transfer directly. Which would be useful? We could split accepted items into focused PRs.

Authored by AI, reviewed by a Human.