KinPlot

A raw DNA file from your whole-genome BAM or CRAM

Choose the BAM or CRAM file of your whole-genome test, MyHeritage's included, and get raw DNA text files in the layout the testing companies use, plus your Y-DNA and mtDNA read deep enough for your terminal haplogroup. The file is read on this device and never uploaded.

Choose your BAM or CRAM file

A BAM or a CRAM is the aligned-reads file a sequencing company provides: a BAM is usually 30 to 100 GB, a CRAM 10 to 30 GB. No index file (.bai, .crai) and no reference genome are needed. Nothing is sent to us. The page reads it from your disk a piece at a time; only the result, two text files of about 15 and 40 MB, is made, and it stays in this tab until you save it.

What to expect

  • MyHeritage's whole-genome test. Its CRAM reads each position about twice, too few to call chromosomes 1 to 22 or X reliably, so the page does not call them from it. MyHeritage's own DNA data file fills those in by its own statistical method. What the CRAM does give is the Y-DNA and mtDNA, which that file leaves out, and the page adds them to it, after checking that both files are the same person's.
  • You pay after you have seen it work. Converting and the summary are free. If the page cannot convert your file it says why, and nothing is charged. The files themselves are made once the order is paid.
  • This is a first version. It reads BAM and CRAM files.
  • It takes a while. The whole file has to be read: roughly half an hour to an hour for a 50 to 90 GB BAM and about 20 minutes for a 16 GB CRAM, on a fast disk; longer on an external or slow one. Keep the tab open; a phone will not manage it.
  • A CRAM is read without its reference genome. A CRAM stores each read only where it differs from the reference genome, which is normally needed to read it back. This page carries the reference's own base at each marker instead, so nothing has to be downloaded or supplied. It works for a file made against the usual GRCh37 or GRCh38 reference, which the page checks in the file's header.
  • Which markers. Two files come out of one reading. One has the markers of the 23andMe v5 chip that carry an rs number, on chromosomes 1 to 22, X, Y and mtDNA. The other adds about a million positions that ancestry tools read, among them every marker this site's own pages use.
  • X, Y and mtDNA. Whether the file is a man's is told from the reads on the Y chromosome. A man's X and Y are written with one letter, as testing companies write them; a woman's file has no call on the Y. mtDNA is written with one letter. In a GRCh37 file whose mtDNA follows the older numbering (UCSC hg19's chrM), mtDNA is left as no call.
  • Insertions and deletions are not called. A chip reports some markers as an insertion or a deletion. Those are written as no call here: this page reads single letters only.
  • Either genome build. A file aligned to GRCh37 (hg19) or GRCh38 (hg38) is read at its own positions; the result is written at GRCh37 positions, as testing companies' files are.
  • How a genotype is called. From reads placed with a mapping quality of 20 or more and bases read with a quality of 20 or more, leaving out duplicates. A marker needs 8 such reads. A second allele in a quarter of the reads or more is written as two alleles; in 15% or fewer, as one allele twice; in between, and wherever the reads show an allele the marker does not have, as no call (--).
  • Stretches left out of some GRCh38 files. The GRCh38 reference comes with alternate versions of some stretches (the HLA region of chromosome 6 is the largest). Where a file was aligned to a reference that has them, without the step that handles them, the reads of one of your two alleles can go to the alternate version and the marker would read as one allele twice. The page tells such a file by its reads, and writes the markers there, about 2.6% of them, as no call. On one man's file this removed more than half of the markers where his chip showed two alleles and the reads one, and cost about half a percent of the calls that were right.
  • What is refused. A file of several people, a file cut short, reads not yet aligned to a genome, a reference genome other than those two, and a file whose reads cover the markers fewer than 10 times on average, such as a low-coverage test or a file of one chromosome.