Bioinformatics File Formats

Every pipeline is a chain of tools, and each tool reads one format and writes another. A sequencer writes FASTQ, an aligner turns it into BAM, a variant caller turns that into VCF, and a quantification tool turns it into a count matrix. The same pattern holds off the sequencer. A cytometer writes FCS, a qPCR machine writes a vendor export or RDML, a mass spectrometer writes a vendor raw file that becomes mzML, and an Olink or Luminex run arrives as a plate reader export.

If you do not know the format, you cannot tell whether a file is correct or corrupted, and you cannot debug a pipeline that fails silently.

The formats at a glance

The table below is the map of this section. File sizes vary with sequencing depth, panel size and organism.

Format	What it stores	Typical size	Written by	Read by
FASTQ	Raw sequencing reads	1-50 GB	Sequencer (Illumina, ONT)	Aligners (STAR, BWA)
SAM/BAM	Aligned reads	2-100 GB	Aligners	Variant callers, featureCounts
CRAM	Aligned reads, reference compressed	0.5-30 GB	samtools	Same as BAM
VCF/BCF	Variants (SNPs, indels)	10 MB-10 GB	GATK, bcftools	Annotation and filtering tools
BED	Genomic intervals	KB-MB	Peak callers, scripts	bedtools, IGV
GFF/GTF	Gene annotations	50-500 MB	Ensembl, GENCODE	STAR, featureCounts
H5AD/MTX	Single-cell count matrices	50 MB-10 GB	Cell Ranger, Scanpy	Scanpy, Seurat
FCS	Per-event cytometry measurements	1-500 MB	Cytometers	flowCore, FlowJo
RDML, vendor exports	qPCR amplification and Cq	0.1-50 MB	qPCR instruments	RDML, vendor software
CEL, IDAT	Microarray intensities	5-50 MB	Array scanners	affy, oligo, limma
mzML, mzXML	Mass spectra	0.1-10 GB	Vendor converters	mzR, Spectra
NPX export	Olink protein concentrations	1-100 MB	Olink software	OlinkAnalyze
xPONENT export	Luminex bead fluorescence	1-50 MB	xPONENT, Bio-Plex	drLumi, base R
POD5, BAM	Long-read signal and reads	10-500 GB	ONT sequencers	pod5, dorado, samtools

You do not need to memorise this table. Each step in a workflow has specific input and output formats, and the pages below cover each one.

The pages

The first seven pages cover the sequencing formats. The rest cover the assay formats that arrive from bench instruments.

Page	What it covers
FASTQ: raw reads	the four-line record, Phred scores, paired-end files, gzip
SAM, BAM and CRAM	the 11 alignment fields, FLAG, MAPQ, CIGAR, sorting and indexing
VCF: variants	the variant columns, INFO and FORMAT fields, genotypes, gVCF
BED, GFF and GTF	intervals and gene models, and the 0-based versus 1-based trap
Single-cell formats	the 10x matrix directory, H5AD, Loom and Seurat RDS
Inspecting files	the command line checks to run before any analysis
FCS: flow cytometry	the binary layout, the keyword segment, reading with flowCore
qPCR files	RDML and the Bio-Rad, QuantStudio and LightCycler exports
Microarray files	CEL and IDAT binaries, and the GEO series matrix text format
Mass spectrometry	mzML and mzXML, the vendor raw formats, reading with mzR
Olink NPX	the NPX export columns, LOD and QC flags, reading with OlinkAnalyze
Luminex	the xPONENT export, bead counts and median fluorescence
Long-read sequencing	POD5 signal, modified-base BAM tags, PacBio HiFi

How the code runs here

The sequencing pages use command line tools such as samtools and bcftools, and their shell examples are illustrative. The assay pages carry R examples, and every R example on them runs in a pinned Podman container in the companion code repository under guides/file-formats/. Every printed output on those pages comes from that run. The readers for these formats live in R and Bioconductor, so the section is R only. Python is named in the prose where it is the standard tool.