Inside your 23andMe raw data file: format, columns, and the weird bits
You requested your raw data from 23andMe, unzipped the download, and got a text file with a name like genome_Firstname_Lastname_v5_Full_20260824.txt. Open it and you are looking at a wall of tab-separated lines. Here is what every part of it means, including the bits that confuse everyone.
The header
The file starts with comment lines marked #. They tell you three useful things: when the file was generated, that the data is aligned to reference genome build 37 (GRCh37, which matters if you ever compare position numbers against newer build 38 resources), and that genotypes are reported on the plus strand of that reference. Hold that last thought.
The four columns
Every data line has the same shape:
rsid chromosome position genotype
rs4680 22 19951271 AG
- rsid: the variant's universal name, the one you can look up anywhere.
- chromosome: 1 through 22, then X, Y, and MT (mitochondrial).
- position: where on that chromosome, in build 37 coordinates.
- genotype: your two letters, or one letter for positions where you carry a single copy (male X and Y, mitochondrial DNA).
A v5 file holds roughly 600,000 to 640,000 lines. That sounds enormous and is still well under 0.1 percent of your genome: the chip tests a curated set of positions, not everything.
The weird bits, decoded
-- entries. Double dashes mean the chip could not make a confident call at that position for your sample. Every file has thousands of these. It is normal, and it means "not measured", never "negative".
i-numbered markers. Scattered among the rsids you will find ids like i3000001. These are 23andMe's internal names for markers that did not have a public rsid when the chip was designed. Most third-party tools skip them.
Single-letter genotypes. Men see one letter on X and Y positions because they carry one copy. Everyone sees single letters on mitochondrial positions. Some other providers pad these to doubled letters instead, which is one of several quirks parsers must handle.
Your letters vs the internet's letters. The file reports plus-strand letters, while articles and papers often name the same variant by the other strand's letters. The classic: MTHFR C677T lives in your file as G or A, with no C or T in sight. Before you google a genotype and scare yourself, read how accurate raw data actually is and why chatbots fumble exactly this.
What the file is good for
The ancestry report you already saw used a slice of these positions. The rest hold your caffeine metabolism, chronotype lean, lactose tolerance, muscle fiber type, and a few hundred other replicated wellness findings that 23andMe's own reports never surface. That second life is exactly what Helisoma reads: drop the file in, your browser parses it locally (it is never uploaded), and every finding comes back with your genotype, its honest effect size, and the study behind it. Free preview, $49 once for everything.
And a housekeeping tip regardless of what you do next: the file is your genome in plain text, so store it like a password, not like a holiday photo.