How big will my sequencing files be? FASTQ vs BAM

Before a run or a big download it helps to know roughly how much disk and transfer you are committing to. File size falls out of read count and read length, with compression the main variable.

FASTQ: two bytes per base, plus overhead

An uncompressed FASTQ stores one sequence character and one quality character per base — two bytes — plus a fixed overhead per read for the header, the+ line and newlines (roughly 50 bytes). So raw size ≈ reads × (2 × read length + ~50). Gzip typically shrinks FASTQ to about a quarter of that.

One million single-end 150 bp reads: 1,000,000 × (2 × 150 + 50) ≈334 MB raw, or about 83 MB gzipped. Scale up to a full 400 M read-pair run at 150 bp paired-end and you are looking at roughly 261 GB raw / 65 GB gzipped — count both mates.

BAM: packed and compressed, but quality dominates

A BAM packs each base into 4 bits (half a byte) and BGZF-compresses the whole file, so the sequence shrinks dramatically. But it still keeps one quality value per base, and quality strings compress poorly — which is why BAM is not as tiny as the 4-bit packing alone suggests. The same one million 150 bp reads come to roughly 150 MB as a BAM: smaller than raw FASTQ, larger than you might guess.

Treat these as ballparks

The FASTQ figures are fairly reliable; the BAM estimate is rougher. Real BAM size depends on read-name length, auxiliary tags, mapping, duplicate marking and how compressible the data is. Use the numbers to size a disk with margin, not to predict bytes exactly.