Raw DNA file check

Consumer testing companies each export raw data in their own text format, with different column layouts, different SNP naming and very different SNP counts. This diagnostic parses the file you give it, reports the company, chip and genome build the file names itself with, counts the SNPs and no-calls on each chromosome, the share of SNPs on chromosomes 1 to 22 that carry a call and how many of those are heterozygous, and states which of our reports and free readings can actually read it. The check itself orders nothing and keeps nothing: the file is deleted once the report renders.

The SNP count is the number that decides what a file can support, and it is the one nobody is told. The density of SNPs on chromosomes 1 to 22 governs how precisely an admixture model can be fitted; the Y-chromosome and mitochondrial SNPs govern whether a paternal or maternal haplogroup can be read at all. Files from the same company differ by chip generation, and some products carry no Y SNPs whatsoever, which is why a file can work perfectly for one analysis and be unusable for another.

Seeing the per-chromosome breakdown before you buy anything is the point. A file thin in autosomal SNPs will still produce an admixture model, but with wider standard errors, and no amount of analyst effort recovers coverage the file never had. What this check cannot do is judge the data's quality beyond what the file states: it counts what is present, the calls and the no-calls, but cannot detect a mislabelled sample, a mixed-up kit, or genotyping error in individual calls. The share called is measured on chromosomes 1 to 22, because a woman's file lists Y-chromosome SNPs it cannot call. The heterozygous share counts the SNPs whose two letters differ; it moves with the chip's design and with ancestry, and it is a count, not a finding about health or family. Runs need a free, email-verified account.

Suppose the check reports: 23andMe v5 detected, ~630,000 SNPs on chromosomes 1 to 22 spread evenly across them with 98.5% called, ~3,600 Y-chromosome SNPs, ~4,000 mitochondrial SNPs, and the verdicts: Ancient Ancestry yes, Ancient Matches yes, Ancestry Map yes, paternal line yes, maternal line yes, DNA Traits yes. That is the healthy-file shape: a current-generation chip with dense autosomes and enough uniparental SNPs for haplogroup calls. The number that matters most downstream is the autosomal count, because standard errors in formal models are set by how many of those SNPs survive the merge with the ancient panel; coverage is the ceiling no analyst effort raises.

Now the common deviations and what each means. A file with no called Y SNPs is not broken (a woman's kit looks exactly like that, and several products ship none), but the paternal line verdict will honestly read no, whatever the customer expected, and the maternal line still reads. A file whose SNP count lands far below its vendor's usual range is typically an older chip generation or a truncated download; re-exporting from the vendor fixes the second, nothing fixes the first. And a whole-genome VCF reads differently by design: variant counts rather than fixed positions, with the verdicts reflecting the conversion our uploaders perform.

The habit this page exists to create: run the diagnostic before paying anyone, us included. Every refusal our order flow makes on file grounds is visible here first, for free, which is cheaper for everybody than a refund conversation.

The usual alternative is no diagnostic at all: buyers assume a '23andMe file' is one fixed thing, when marker sets differ across that company's five chip generations by hundreds of thousands of positions. The second alternative is the analyzer-and-converter ecosystem (DNA Kit Studio and its kin, or DNA Genics' file tools), which inspects and rewrites raw files ably but answers a different question. Those tools report what a file contains; this page reports the same counts (the SNPs, the share called, the heterozygous share) and then answers the question a buyer actually has: which specific analyses can this file support, stated per product against the panels those products really merge with.

What no checker anywhere can promise: that the DNA in the file is the person on the label, that the vendor's calls are error-free, or that a file good for one analysis is good for all. The honest scope is exactly what is printed (format, counts, per-analysis verdicts), and the file itself is deleted once the report renders.