FASTQ is the standard file format for storing sequencing reads. Each read is stored as four lines: a header, the sequence, a separator, and a quality string. Understanding this format — and what the quality string actually means — is the foundation for everything else we do in bioinformatics QC.

💡 TL;DR A FASTQ file is a FASTA file with quality scores attached. The quality string encodes the probability of error for each base. Higher quality scores mean fewer sequencing errors.

Anatomy of a FASTQ record

Every read in a FASTQ file looks like this:

Each record (sequence) will be stored in exactly 4 lines:

• The header starts with @ and carries the read name and (optionally) metadata. The quality string is always exactly the same length as the sequence — one character per base.
• The sequence line contains the nucleotide sequence of the read.
• The separator line is always a + and may optionally repeat the header.
• The quality string encodes the Phred quality score for each base in the sequence.

Phred quality scores

The quality string looks like gibberish, but each character encodes a Phred quality score (Q) — a number that tells you how likely that base call is wrong.

The formula is:

Q = −10 × log₁₀(P) where P = probability the base is wrong

So a base with a 1-in-100 chance of being wrong has Q = 20. A base with a 1-in-1000 chance has Q = 30. The higher the number, the more confident the sequencer was.

Q30 is the conventional threshold for “good quality” Illumina data. Q20 is often used as a lower bound for trimming.

How scores are encoded

Phred scores are stored as single ASCII characters, not as numbers, to keep the file compact. The encoding is:

ASCII character = Phred score + 33

So Q score 30 → 30 + 33 = 63 → the character ?. Score 40 → 73 → I. Score 0 → 33 → !.

Quality expectations: Illumina vs Nanopore

The two platforms have very different error profiles, and that shapes everything about how you work with their data.

⚠️ A Nanopore dataset with mean Q12 and an Illumina dataset with mean Q12 are very different problems. Illumina Q12 indicates severe degradation or a failed run. Nanopore Q12 was typical for older flow cells and is still usable for many applications (e.g. species-level metagenomic profiling).

How accuracy has improved in Nanopore

Nanopore accuracy has improved substantially over successive chemistries and basecallers, but still not usually as high as Illumina:

Inspecting a FASTQ file

You can peek at a FASTQ file directly, but remember that each record is 4 lines — so “the first 8 lines” gives you two reads:

zcat sample.fastq.gz | head -8

To count reads:

# Each read = 4 lines
zcat sample.fastq.gz | wc -l | awk '{print $1/4}'

To get a quick summary with seqfu:

seqfu stats sample.fastq.gz

• Output includes: number of reads, total bases, mean/min/max length, mean quality.

💡 seqfu handles gzip-compressed files natively — you do not need to decompress first.

Questions

• ❓ Open a FASTQ file and find the quality string for the first read. What is the Q score of the first base? What is the Q score of the last base? Is there a trend?
• ❓ A read has a quality string that starts with IIIIIIII and ends with !!!!!!. What does that tell you about how the sequencer performed across the read?
• ❓ You have a Nanopore dataset with mean Q15 and an Illumina dataset with mean Q30. Which one has more errors per base? By how much?
• ❓ Why can’t you just use the same QC thresholds for Illumina and Nanopore data?