How to Convert 23andMe Raw Data to VCF Format Free

Yes, you can convert a 23andMe raw data file to VCF format free. A browser-based converter at https://healthosx.com/dna/convert/23andme-to-vcf writes a GRCh37 VCF at known SNP sites, never uploads the file, and never guesses the reference letters a 23andMe download does not contain.

A chip file gives your two letters per marker. A VCF also needs the reference letter at that position. The converter takes that reference letter from a 2.17 million-site list, and rows outside the list are left out and counted, never guessed.

How do I convert my 23andMe raw data to VCF? Drop the raw .txt or the whole .zip from 23andMe on the 23andMe-to-VCF converter page. The file is read in the browser. There is no account and no upload queue.

The output is a one-sample VCF on GRCh37, the same build 23andMe reports. Each row carries the ID from your 23andMe file, a reference allele looked up from a known-site list, and your called alleles. A no-call stays a no-call. Positions the chip did not test are not filled in.

What the file is missing before conversion Open a 23andMe raw file and you see comment lines starting with #, then rows like:

rs1800562 6 26093141 AG

That row says: rsID, chromosome, position, your genotype. It does not say what the reference allele is at chr6:26093141. A VCF row cannot be built from those four fields alone, because REF is not in the file.

The converter here does not guess. It matches each row to a list of 2.17 million known SNV sites and takes the GRCh37 reference letter for that exact position. If the row is not in the list, it is left out and counted.

What the output VCF looks like A VCF begins with metadata lines starting with ## and a header line starting with #CHROM. For a converted 23andMe file, a row looks like:

6 26093141 rs1800562 G A . . . GT 0/1

The REF is the reference letter, not your letter. The ALT is what your file carried. If your 23andMe row was --, the no-call is kept as a no-call instead of being written as a fake genotype.

Because the output is GRCh37, it works in tools that expect the same build. If a tool asks for GRCh38, this output is not that. The converter does not lift positions to another build; it keeps the coordinates 23andMe reported.

What about rows off the known-site list? You may lose some rows when converting a chip file to VCF. That is not a bug. The raw 23andMe file tells you which positions the chip read and what your letters are, but it does not carry the reference sequence. Without a reference letter, a correct VCF row cannot be written. Leaving the row out, and telling you how many were left out, is the behavior to want.

The page leaves out rows that are not on the known-site list and reports how many were left out before you download the VCF. If you absolutely need every row including off-list positions, a local tool such as bcftools with a GRCh37 reference fasta is the alternative. The in-browser converter handles rows on the known-site list without an install and without uploading the file.

What people get wrong about this conversion The biggest mistake is treating a converted 23andMe VCF as a whole-genome VCF. It is not. It is a VCF at the positions the 23andMe chip tested, roughly a few hundred thousand common SNPs, not millions of variants across exomes and genomes. It will not contain rare variants, indels, or structural variants.

Another mistake is expecting every 23andMe ID to be an rsID. Some rows start with i instead of rs. Those internal probe names are normal. The converter keeps the ID that is in your file. Tools that match by rsID may ignore those rows, and that is a property of the raw file, not the conversion.

After conversion, what can I actually do with it? A converted GRCh37 VCF can be read by haplogroup tools, some annotation tools, and any pipeline that expects a one-sample VCF but only looks at common chip positions. It also gives you a local VCF so you do not have to keep sending a 23andMe text file around. But conversion does not make any genotype more meaningful. The same letters in the 23andMe file are the same letters in the VCF.

If you already have the raw file, you may not need to convert it for every tool. Some tools read the 23andMe file directly. When a specific tool demands VCF, that is the case for conversion.

Open the free 23andMe to VCF converter at https://healthosx.com/dna/convert/23andme-to-vcf and run your own file. The conversion stays in your browser, and the page counts rows it had to leave out before you download the VCF.