All studies
DNA·November 20, 2026·4 min read

Can I Merge My Old and New 23andMe Tests From Different Chip Versions Into One Raw DNA File

Yes, two 23andMe chip files from one person can be merged by position: you gain markers the first file lacks, fill no-calls, and disagreements become counted no-calls.

Yes. Two 23andMe raw data files from the same person, say a v4 and a v5, can be merged into one, because each chip tests a partly different set of positions. The merger lines the files up by chromosome and position and writes one file with more positions than either original.

First, get the files

Download each kit's raw data separately from your 23andMe account. Each kit gives you its own .zip. Drop both .zip files into the merger as they are: it takes .txt, .csv or .zip, so there is nothing to unzip.

You can also run each file through https://healthosx.com/dna/check first. The checker only reports the chip where the file states it, and a 23andMe download may not name v4 or v5 in its header. So rely on the build and the no-call rate the checker shows, and on which kit you ordered when. Then merge the two files at https://healthosx.com/dna/merge.

The result is one file with more positions than either original. It is not a new test. Nothing is measured again, and no position appears that neither chip looked at.

Why two kits from the same company test different positions

A 23andMe kit reads your DNA on a genotyping chip: a slide carrying a fixed list of probes, one per position. The list is decided when the chip is designed. When the company moves to a new chip version, it drops some positions and adds others, so a v4 file and a v5 file overlap only in part. The overlap count the merger reports is your real number. A v5 file lists roughly 630,000 markers on genome build 37.

That gives two separate ways a second file helps. First, it can hold positions the first chip never had. Second, where the first file shows -- (a no-call, meaning the probe ran but gave no clear answer), the second chip may have read the same position fine.

Keep those two apart. A position missing from a file was never tested. A no-call was tested and failed. Only the second kind can be repaired by the other file, and only if the other chip carries that position.

What the merge does at each position

How a position-by-position merge treats each case
SituationMerged file
Both files call it and agreeThe call, once
Only one file tested the positionCopied from that file
One file has a no-call, the other a callThe call
Both files call it but differA no-call, counted as a disagreement

What the merge page states the merged file contains for each case.

The disagreement row is the one people ask about. If one chip says AG and the other says GG, one of them is wrong, and nothing in either file says which. Writing either letter would be a guess, so the merged file holds a no-call and counts it. Check that count to see how clean the pair was.

The two checks the merger makes

The merger refuses in two cases.

  • Different genome builds. Shared rsIDs have to sit at the same positions in both files. On a different build the same marker has a different address, and the lining-up would pair unrelated positions. Both 23andMe kits normally report build 37, and the merger refuses if the builds differ.
  • Not the same person. Where both files have a call, they must agree on at least 95% of positions. Two different people agree on far fewer, and the merge is refused. A file blending two people would be wrong for both of them.

One limit: this merges chip files only. A whole-genome VCF is a different kind of file, and it is not merged.

What you gain, and what you do not

Gained:

  • Markers only one chip carried, now in a single file.
  • No-calls in one file filled from the other.
  • Counts of overlaps and disagreements.

Not gained:

  • Better accuracy where the two files already agree. Agreement is simply kept.
  • Any position neither chip tested.
  • A way to tell which chip was right at a disagreement. That stays unknown.
A merge adds positions and repairs no-calls. It never turns a guess into a measurement.

Using the merged file, and the mistake to avoid

The common mistake is pasting two text files together in an editor or spreadsheet. That leaves duplicated rsIDs, repeated header lines and, where the chips disagree, two conflicting rows for one position. Duplicates and conflicting rows make the file ambiguous.

Use https://healthosx.com/dna/merge instead. It runs in your browser, so neither file is uploaded, and you choose the layout of the result: 23andMe, AncestryDNA, MyHeritage, FamilyTreeDNA, Living DNA or PLINK.

Will the merged file still be accepted elsewhere? It is no longer an original 23andMe export, and some sites cap positions or treat combined files differently from a single test, so read a site's rules before you upload it there.

The free report at https://healthosx.com/dna reads the merged file like any other.

New to raw DNA files? Start with the full guide: 23andMe Raw Data Interpretation Free in Your Browser

Written by

HealthOSX Research

Keep reading