The problem. Raw sequencing reads carry adapter contamination, low-quality tails, and other artefacts. Historically you chained several tools — one to trim adapters, one to filter by quality, one (FastQC) to report — each re-reading the whole file. It works, but it’s slow and fiddly, and it’s the very first thing standing between raw data and analysis.

The idea. fastp does adapter trimming, quality filtering, read pruning, and quality reporting in a single multi-threaded pass over the FASTQ, with sensible auto-detection (including adapters) and a built-in before/after QC report. One fast tool replaces a small pipeline of them.

Why it matters. This is the literal first step of both aml_rnaseq_nf and variant_calling_nf — garbage in, garbage out, so read cleaning gates everything downstream. Reading the paper is a reminder that pipeline quality starts before alignment: a bad adapter-trim quietly corrupts every count and every variant call after it. fastp’s popularity is the same story as MultiQC — the right tool removing friction from a step you must not skip.

Verdict. Foundational to practical workflows and a near-universal default; its speed is the point at scale. Not glamorous, but it’s the hygiene step everything else depends on. Read it for what good preprocessing actually checks for.