What your Nucleus downloads contain, and what to do with them

The five downloads

Nucleus's FAQ says you can "download your whole genome as VCF, CRAM, SV, CNV, or FASTQ file types" from the account settings page, and separately as a file "formatted for upload to sites like Promethease or GEDmatch". Its health test is "whole-genome sequencing to read your DNA at an average of 30X coverage". Here is what each download is, and which ones matter here.

• The VCF is the list of positions where you differ from the reference, usually well under a gigabyte. It is the file for the $39 full report: clinical findings, carrier status, the wider ClinVar screen across 1,363 more carrier genes, the pharmacogenes a VCF can carry, traits, ancestry and the polygenic scores.
• The CRAM is the aligned reads in compressed form, the equivalent of the BAM other providers ship. It is the file for the $49 Deep Read panel. A CRAM is written against one exact reference assembly and can only be decoded with that assembly; the file check reads which one yours needs and says whether we hold it, before you pay.
• The SV and CNV files are Nucleus's own structural-variant and copy-number calls. We do not read them. Deep Read measures copy number directly from the reads for the genes where it matters, CYP2D6 above all.
• The FASTQ is the unaligned raw reads. We do not use it. Keep it; it is the only file from which everything else can be rebuilt.
• The Promethease/GEDmatch export is a chip-style genotype list derived from the genome. It is the least of the five. Upload the VCF instead; it carries everything the export does and a great deal more.

Nucleus does not state on its pages which reference build its files are on. Nothing is asked of you: the file check reads the build from the header, and a GRCh37 file is lifted to GRCh38 before analysis. A GRCh38 VCF is usually ready in about a quarter of an hour; a large GRCh37 file can take a couple of hours.

What a second reading adds

Nucleus's reports are one interpretation, with a strong emphasis on polygenic risk. This one is built differently, and the difference is the product. The panel is fixed and named: every gene, condition and trait examined is on one page before you pay. Every finding is attributed to the source it was restated from, and the version of each source, ClinVar, ACMG SF and CPIC among them, is printed on the report. And the report keeps what it found apart from what it could examine: a variant-only VCF cannot tell "you are clear" from "nobody looked", so a clean result from it is reported as absent, never as examined and clear.

The CRAM changes that. Deep Read counts the bases covered at 10x or better across every region of the curated panel, so each gene comes back examined, partly examined, or not measured; the wider ClinVar screen's positions are read from the VCF. The same reads resolve CYP2D6 copy number, call two recurrent deletions, type HLA-DQ, and drive the 29-gene pharmacogenomic panel to CPIC star-allele resolution, ending in a one-page hand-off you can take to a prescriber. Where the reads are too thin for a gene, the panel declines that gene rather than calling it. We carry 17 disease polygenic scores from the PGS Catalog, with 9 emerging-research scores kept apart from them, and we will not race anyone on the count.

Who Nucleus is

Nucleus Genomics is a New York company founded in 2021. Its launch announcement in March 2024 put the 30x kit at $399 with a $39-a-year membership for new features, and named a CLIA-certified, CAP-accredited laboratory in North Carolina; its health test page lists the kit at $499 today. Its terms say that if you close your account "we will not retain any information or data" beyond what the law requires, so download your files before you do. As with every provider on these pages, the copies you hold are the ones you control.

How to get your files out of Nucleus

1. Sign in and open Account settings, or choose Downloads from the account menu, as Nucleus's FAQ describes.
2. Download the VCF first. It is the small file and the one the full report needs.
3. Download the CRAM and its index (.crai) if you want the Deep Read panel. It is large; give it a wired connection.
4. Upload them here as delivered. Do not rename, unpack or convert them.

Nucleus's products and download process are described from its public pages on 3 September 2026 and may change. We do not speak for Nucleus.

Your files do not leave the EU

Processing and storage are in Finland, under GDPR Article 9 special-category protection, and nothing crosses an EU border. The VCF you upload is deleted after analysis by default; the reads are deleted the moment the Deep Read run finishes, whether it succeeded or failed. With your consent, and only if you want re-analysis later, we keep a small normalised variant file that you can delete from your dashboard at any time. Security & your data.

Common questions

Which Nucleus download do I use?

The VCF for the full report, and the CRAM as well if you want the Deep Read pharmacogenomic panel. The SV and CNV files are Nucleus's structural-variant calls and we do not read them; the FASTQ is the unaligned raw reads and we do not use it either. The file check below names any of these without uploading a byte.

Nucleus gives me a CRAM, not a BAM. Does that work?

Yes. A CRAM stores your reads as differences from the exact reference assembly it was compressed against, so it can only be decoded with that same assembly. The file check reads which one yours needs and tells you before any money changes hands; if it is one we do not hold, we say so rather than decode it wrongly.

Which reference build are Nucleus files on?

Nucleus does not say on its pages. It does not matter for the upload: the file check reads the build from the header, and a GRCh37 file is lifted to GRCh38 before analysis, which takes longer.

What does this add to what Nucleus already told me?

A second reading by a different service, of a fixed panel you can read by name before you pay, against pinned versions of ClinVar, ACMG and CPIC that are printed on the report. From the CRAM, what only reads can answer: CYP2D6 copy number, a per-gene record of how much was covered, two recurrent deletions and typed HLA-DQ, with the 29-gene pharmacogenomic panel called from the same reads and a one-page hand-off for a prescriber.

Do you keep my genome?

The uploaded VCF is processed and stored in Finland and deleted after analysis by default. Reads for the Deep Read panel are deleted the moment the run finishes. Nothing leaves the EU, and there is no Aimosti genome vault to be sold on.

Sequenced elsewhere? Dante Labs files · Nebula and DNA Complete files · Sequencing.com files · What your reads add, whoever sequenced you