Quality Reports - Usage Guide

Overview

Quality reports are the first step in any NGS analysis. FastQC (or the faster drop-in falco) generates per-file reports of Phred quality, per-base composition, GC, duplication, overrepresented sequences, and adapter content; MultiQC aggregates them into one cross-sample summary. The skill teaches reading the PLOTS against the assay rather than trusting FastQC's pass/warn/fail, which are calibrated to whole-genome DNA and false-fail on RNA-seq, amplicon, bisulfite, small-RNA, single-cell, and long-read data.

Prerequisites

# Conda (recommended)
conda install -c bioconda fastqc multiqc falco seqkit nanoplot

# pip (MultiQC, NanoPlot)
pip install multiqc nanoplot

Quick Start

Tell your AI agent what you want to do: - "Run FastQC on all my FASTQ files and aggregate with MultiQC" - "Check the quality of my sequencing data before trimming" - "My FastQC shows adapter contamination, what should I do?" - "Is this duplication level a problem for RNA-seq?" - "QC my Nanopore reads"

Example Prompts

Initial QC

"Run FastQC on all FASTQ files in my data directory and build a MultiQC summary"

"Check the quality of my raw reads before trimming and flag any outlier samples"

Aggregating and comparing

"Combine all my FastQC reports into one MultiQC report with custom sample names"

"Compare QC before and after trimming in a single report"

Interpreting results

"My FastQC fails per-base sequence content in the first 12 bases, is that a problem?"

"I see a 3' rise in G content on NovaSeq data, what is that?"

"Explain the duplication plot in my MultiQC report for an RNA-seq library"

What the Agent Will Do

Run FastQC (or falco) per file with appropriate threading, or NanoPlot for long reads
Aggregate the per-file reports with MultiQC and read the General Statistics table for outliers
Interpret each flagged module against the assay (distinguish expected false-fails from real defects)
Route remediation to the right skill (adapter-trimming, quality-filtering, fastp-workflow, contamination-screening)

Tips

Treat a FastQC FAIL as a hypothesis about a WGS library; for any other assay, first ask whether the module is expected to deviate for that chemistry.
On 2-color instruments (NextSeq/NovaSeq), high-quality poly-G tails appear as a 3' G-content rise, NOT as low quality; remove them with a chemistry-aware poly-G trim, not a quality cutoff.
Blocky/quantized quality boxes on NovaSeq are binned quality scores (expected), not bad data.
High duplication is normal for RNA-seq, amplicon, and ChIP/ATAC; it is a prompt to assess library complexity (preseq), not an automatic reason to remove duplicates.
The "Overrepresented sequences" module prints the offending sequence so it can be BLASTed to identify the source.
MultiQC scrapes logs rather than re-analyzing data; check multiqc_sources.txt if two samples merge into one row.
Run FastQC before and after trimming and overlay them in MultiQC to verify improvement.

Resources

read-qc/adapter-trimming - Remove read-through adapter flagged by the adapter-content panel read-qc/quality-filtering - Drop low-quality reads and trim ends read-qc/fastp-workflow - All-in-one QC + trim, including 2-color poly-G read-qc/contamination-screening - Resolve bimodal-GC or unexpected overrepresented sequences read-qc/rnaseq-qc - Transcriptome QC on the aligned BAM sequence-io/sequence-statistics - Programmatic per-file sequence summaries