fastp Preprocessing for RNA-Seq FASTQ Files

NotchBio uses fastp to remove adapters and low-quality read tails before Salmon or STAR quantification. The preprocessing step creates cleaned FASTQ files and a versioned report for every sample, so you can compare the reads before and after filtering without overwriting the raw uploads.

When Should You Run fastp on RNA-Seq Reads?

Run preprocessing after reviewing the raw quality-control report. Adapter contamination, low-quality 3′ tails, or poly-G artifacts are good reasons to trim. A FastQC warning by itself is not always a reason: RNA-seq libraries can show expected sequence-content bias near the beginning of reads.

Start with the recommended preset unless the library protocol or raw QC report gives you a specific reason to change it. Aggressive trimming can shorten usable reads and reduce mapping rates, so the goal is to remove a demonstrated problem—not to make every FastQC module green.

How to Run fastp in NotchBio

1. Open the project and go to Run.
2. Complete or select the raw Quality control version you want to review.
3. Open Preprocessing.
4. Choose the samples and a fastp preset.
5. Review the effective parameters, then start the run.
6. Keep the page open to watch per-sample progress, or return later and select the saved version.

Every parameter set creates a separate pipeline version. A later Salmon or STAR run records which preprocessing version it consumed.

fastp Parameters and Defaults

Setting	Recommended default	What it controls	Change it when
Qualified quality (Phred)	20	Bases below this score are treated as low quality	Your protocol requires a stricter or more permissive quality floor
Minimum read length	36 nt	Reads shorter than this after trimming are discarded	You have short-read libraries or an aligner-specific minimum
Cut front	Off	Enables sliding-window trimming from the 5′ end	Raw QC shows a sustained low-quality 5′ region
Cut tail	Off	Enables sliding-window trimming from the 3′ end	Raw QC shows a sustained low-quality 3′ tail
Window size	4 nt	Number of bases evaluated by the sliding window	You need a protocol-specific trimming rule
Window mean quality	20	Quality threshold used by the sliding window	You enable front or tail cutting
Trim poly-G	Off	Removes poly-G tails commonly associated with two-color chemistry	NextSeq/NovaSeq data shows a poly-G artifact
Trim poly-X	Off	Removes long homopolymer tails	QC or the library protocol shows a real poly-X artifact
Custom R1/R2 adapters	Auto/blank	Overrides automatic adapter detection	The library kit specifies adapters that auto-detection misses

The relaxed preset lowers the qualified-quality threshold to 15 and minimum length to 30. The stricter preset enables front and tail cutting, raises the window mean quality to 24, and enables poly-G trimming. Use these presets as deliberate alternatives, not automatic upgrades.

How to Read the fastp Report

Compare these metrics across all samples rather than judging one sample in isolation:

• Reads before and after filtering: a large loss may indicate aggressive settings, poor input quality, or adapter-dominated libraries.
• Pass rate: unexpected outliers are more important than a universal target. Replicates prepared together should usually have similar retention.
• Q30 rate after preprocessing: this should improve or remain high, but a high Q30 value does not compensate for losing most reads.
• Adapter-trimmed reads: confirms whether adapter contamination was present and how much fastp removed.
• Read length after trimming: very short retained reads can map ambiguously even if their base quality is good.

After preprocessing, continue to Salmon or STAR and compare mapping rates. A trimming run that improves base quality but sharply reduces mapping is not necessarily better.

Common fastp Problems and Fixes

Too Many Reads Were Discarded

Return to the raw QC report and identify which rule is responsible. Try the recommended or relaxed preset, disable front/tail cutting, or lower the minimum length only when the downstream quantifier can use the shorter reads. Create a new version instead of replacing the previous run so the results remain comparable.

Poly-G Tails Remain in NextSeq or NovaSeq Data

Enable Trim poly-G and rerun preprocessing. Confirm the change in the after-filtering report and then check whether the Salmon/STAR mapping rate improves.

Paired-End Samples Lose Different Numbers of R1 and R2 Reads

fastp handles paired reads together. If one mate fails filtering, the pair may not remain usable as a paired observation. Check the original file pairing and read counts before loosening quality thresholds.

The Run Finishes but Mapping Gets Worse

Compare the new preprocessing version with the prior version. Over-trimming, a wrong custom adapter, or overly short reads can reduce unique mapping. Use the version with the most defensible balance of retained reads, quality, and downstream mapping—not simply the greenest report.

What to Record for Reproducibility

Record the selected fastp version, preset, quality threshold, minimum length, front/tail settings, window values, poly-G/poly-X settings, and any custom adapter sequences. NotchBio stores these parameters with the run so collaborators can trace the cleaned FASTQ files back to the exact configuration.

For a command-line explanation of the same workflow, see FASTQ Quality Control: FastQC, fastp, and MultiQC Workflow. For help deciding whether a warning needs trimming, read Understanding Your QC Report: FastQC and MultiQC.