How to Convert Raw DNA From GRCh38 to GRCh37 for Free
Convert a GRCh38 raw DNA file to GRCh37 for free in your browser. The file never uploads, letters flip where strands cross, and unplaceable rows drop.
Convert a GRCh38 raw DNA file to GRCh37 for free by opening the browser-based page at https://healthosx.com/dna/convert/grch38-to-grch37, dropping your file, and downloading the GRCh37 copy. The file never leaves your device. The converter uses Ensembl's GRCh38-to-GRCh37 chain. It flips letters where the mapping crosses strands. It drops positions that cannot be placed rather than guessing.
How do I convert my raw DNA file from GRCh38 to GRCh37 for free?
Do it in four steps. Before you start, confirm your file is on GRCh38. Check the header in a text editor or use https://healthosx.com/dna/check. If your file is already GRCh37, you do not need this conversion; leave it unchanged.
- Open https://healthosx.com/dna/convert/grch38-to-grch37.
- Drop a raw file. It reads .txt, .csv, .zip, .vcf and .vcf.gz exports from 23andMe, AncestryDNA, MyHeritage, FamilyTreeDNA, Living DNA, WeGene and whole-genome VCFs.
- Choose the output layout you need: 23andMe, AncestryDNA, MyHeritage, FamilyTreeDNA, Living DNA or PLINK.
- Download the converted file. It was read and written by your browser, not uploaded to a server.
This page is for files on GRCh38; confirm your build first with https://healthosx.com/dna/check. Most 23andMe and AncestryDNA files are already on GRCh37, so leave those alone.
GRCh37 and GRCh38 are two address systems for the same genome
GRCh37 is also called build 37 or hg19. GRCh38 is also called build 38 or hg38. The human genome did not change between builds. The map changed, so a variant gets a different position number in each build.
Most consumer genotyping chips report positions on GRCh37 because their probes were designed against that assembly. Many whole-genome sequencing providers now deliver VCFs on GRCh38. A tool that matches variants by position has to know which address system the file uses. If a tool expects GRCh37 and your file says GRCh38, it will reject the file or read the wrong location.
| GRCh37 (hg19) | GRCh38 (hg38) | |
|---|---|---|
| Also called | build 37, hg19 | build 38, hg38 |
| Typical source | consumer chips | newer sequencing VCFs |
| Position matching | tools expect GRCh37 | tools expect GRCh38 |
| What changes | coordinates only | coordinates only |
Scroll the table sideways for the rest →
Two builds, one genome. Coordinates change between builds; the underlying DNA does not.
What the converter actually does: lift, flip, drop
This is not subtracting a fixed number from every coordinate. The conversion uses Ensembl's GRCh38-to-GRCh37 chain, which records how chromosome segments moved between the two assemblies.
- Lift: each chromosome and position on GRCh38 is mapped to its one-to-one location on GRCh37.
- Flip: if Ensembl's chain says that segment is on the opposite strand in GRCh37, the genotype letters are reverse-complemented. A becomes T, C becomes G.
- Drop: if a position has no one-to-one GRCh37 location, the row is left out and counted. The converter will not guess.
GRCh38 to GRCh37 conversion changes coordinates, not genotypes. Letters flip only when the strand crosses; unplaceable rows are dropped rather than guessed.
Genotype calls are preserved. Letters are reverse-complemented only where the chain crosses strands. A no-call stays a no-call. The converter does not call new variants. It also does not fill missing rsIDs from a guess, and it will not rescue rows the chip did not read.
What real data looks like before and after
Take rs429358, the APOE variant. On GRCh38 it sits at chromosome 19, position 44,908,684. On GRCh37 it sits at chromosome 19, position 45,411,941. The genotype is still whatever your file says, C or T. Only the address changes.
In a converted file, the column layout stays the same. The position column moves to GRCh37. The genotype letters are reverse-complemented where the chain crosses strands, so a T on GRCh38 can become an A on GRCh37 because the same base pair is now described from the other strand.
Do I need to convert if my file already says GRCh37?
No. If the header says build 37, hg19 or GRCh37, you do not need this conversion; leave it unchanged.
The common mistake is converting the wrong direction. A 23andMe v5 file lists roughly 630,000 markers on genome build 37. That file is already what most consumer tools want. VCFs from newer whole-genome sequencing are the files that usually need the move from GRCh38 to GRCh37.
Open the raw file in a text editor, not Excel. The first comment lines usually state the build. Spreadsheets can change chromosome names or drop leading zeros, so a text editor is safer for checking.
What if I have a VCF with no rsIDs?
For the build conversion itself, rsIDs are not required. The converter can lift GRCh38 positions to GRCh37 using the chain. If you also need a chip-style layout, it matches rsIDs to positions using a list of 2.17 million known SNV sites where the position and reference letter both match. Rows off that list are left out and counted, never guessed.
That means a small share of positions may not make it into a 23andMe-style output. The page tells you how many were left out. It is better to lose a row than to write a wrong rsID.
Open the converter at https://healthosx.com/dna/convert/grch38-to-grch37 when your file is on GRCh38 and the tool you want to use expects GRCh37. The full report at https://healthosx.com/dna reads the same raw file for ancestry, traits, medication genes and both family lines.
New to raw DNA files? Start with the full guide: How to Download & Analyze Your 23andMe Raw Data Free
Written by
HealthOS Research