The raw files behind your genome sequence: FASTQ, BAM, CRAM, and VCF, explained

Most buyers of a consumer whole genome sequencing kit never click the raw data download button. They get the app, browse the health reports, and move on. That is a reasonable use of the product. But the raw files are what makes whole genome sequencing structurally different from a chip-based ancestry test, and knowing what they contain matters the moment your provider goes quiet -- shuts down, gets acquired, or stops updating its reporting software. Nebula Genomics discontinued its consumer service in early 2025, per ProPhase Labs, its parent company; customers who had not downloaded their files beforehand found themselves navigating uncertainty about whether their data would remain accessible under the relaunched DNA Complete brand. Downloading your data is a form of insurance that costs nothing beyond storage space.

There are four file formats in a standard whole genome sequencing pipeline, and they form a one-way chain. The sequencer produces FASTQ. Alignment software converts FASTQ into BAM or CRAM. Variant-calling software converts BAM into VCF. Each step is a transformation: later formats are smaller and more immediately useful, but earlier formats preserve more of the original measurement. If you want to reprocess your data with a better pipeline five years from now, you go back to an earlier file.

FASTQ is the raw output from the sequencer. It contains every read the instrument recorded -- roughly 900 million paired reads for a 30x human genome, each 150 bases long in both directions -- plus a quality score for each base position. For a 30x human genome, FASTQ files run approximately 100 to 150 gigabytes. According to genomics course materials published by Auburn University, FASTQ is the universal starting point for next-generation sequencing workflows; no consumer analysis app accepts FASTQ directly, since processing it requires a bioinformatics pipeline. BAM -- Binary Alignment Map -- is the result of running that pipeline: the FASTQ reads are mapped to a reference human genome (GRCh38 is the current standard), and the output records each read's chromosomal position. A BAM file for a 30x genome is typically 40 to 80 gigabytes. BAM is the format accepted by most platforms that offer reanalysis of data from other providers: Sequencing.com's app marketplace accepts VCF and BAM uploads from customers who sequenced elsewhere, per a 2026 review of the service published by KnowYourDNA.

CRAM is a compressed version of BAM -- the same alignment data encoded more efficiently, typically 10 to 15 gigabytes rather than 40 to 80. CRAM is mathematically lossless when the correct reference genome file is available for decompression, but that decompression step adds friction for most users, and support for CRAM in consumer analysis software still lags behind BAM. MyHeritage began offering CRAM downloads for kits processed with its whole genome sequencing technology, per a press release on BusinessWire dated October 14, 2025 -- among the first consumer genealogy platforms to provide access to full WGS files at no additional charge. VCF -- Variant Call Format -- is the smallest and most immediately useful format. Rather than containing reads, it lists the positions in your genome where your sequence differs from the reference -- roughly 4 to 5 million such positions for a typical person -- along with quality scores and population-frequency annotations. A VCF for a 30x genome runs 400 to 700 megabytes: small enough to store on a thumb drive and upload to an analysis platform in minutes. Population genomics documentation published by CD Genomics describes VCF as "the starting point for subsequent expert analysis," and at the consumer level, it is what most third-party analysis tools, from trait reports to pharmacogenomics panels, actually read.

The practical guidance is short. Download the VCF from any provider that offers it, and keep a copy somewhere you control: your own external drive, a private cloud folder. That file is small, portable, and the standard input for almost every consumer analysis tool available today. If storage is not a concern, download the BAM as well; it gives you the option to redo variant calling with a newer pipeline if one becomes available. Skip the FASTQ unless you know you will need it or are comfortable managing files above 100 gigabytes. Most consumer use cases -- uploading to a third-party reporting service, running a pharmacogenomics panel, comparing variants with a family member -- work entirely from the VCF. The one thing to avoid is leaving your data exclusively on a provider's servers indefinitely. Providers have business continuity plans; whether those plans include your raw files is a different question, and it is the question that matters when the plan changes.