I have both of my (late) parentsβ DNA. That let me skip the generic reference panels and compute my maternal/paternal split directly β placing myself in genetic space between the two people I actually descend from. For the raw genotyping and pharmacogenomics, see Genomics.
Across 280,413 shared polymorphic SNPs, I land essentially at the midpoint of my parents on the principal ancestry axis (PC1) β the ~50/50 maternal/paternal split every child inherits, here measured against my own parents rather than a generic reference panel.
Each dot is one person, placed by the top two principal components of our shared genotypes. Mom and Dad anchor the ends of the ancestry axis (PC1); I sit at its midpoint. The dashed line is that axis; the open ring marks my projected position on it.
| Pair | Allele-sharing | Reads as |
|---|---|---|
| David & Mom | 0.839 | how much I share with my mother |
| David & Dad | 0.840 | how much I share with my father |
| Mom & Dad | 0.742 | how much my parents share with each other |
Standard ancestry tools lean on public reference panels (1000 Genomes, HGDP) that carry only a handful of broad European groups β not enough to separate, say, Italian from Slavic. So instead of a generic panel I used the two best references I could possibly have: my own genotyped parents. Principal-component analysis over the 280,413 SNPs all three of us share places each of us in genetic space; I fall at the midpoint of the line between them.
Allele-sharing (identity-by-state) confirms the geometry is real, not an artifact: I share about the same fraction with each parent (0.839 with Mom, 0.840 with Dad), while my parents share notably less with each other (0.742) β i.e. they come from genuinely distinct populations, and I am the even blend of the two. Pure numpy: no external panels, no proprietary tool.