How pedigree depth changes a COI result: analysis of 30 recorded dog pairings
Research note · Pedigree methods
Abstract
Pedigree inbreeding coefficients depend on the ancestors present in the record. We examined that dependence in a published 237-dog research pedigree, comparing the expected offspring coefficient of inbreeding (COI) for each of its 30 distinct recorded sire–dam pairings using the full available record and successively shorter ancestral views. Five pairings had nonzero COI in the full record. With only one generation of each parent’s ancestry, two of those five read as zero, and four had a lower coefficient than the full-record calculation. This is a sensitivity analysis of one pedigree, not an estimate of genetic diversity across a breed.
Why test pedigree depth?
A zero pedigree COI can mean that the recorded ancestors do not overlap. It can also mean that the records stop before an overlap becomes visible. Studies comparing short and deep dog pedigrees have found that the depth selected for analysis can materially change the calculated inbreeding coefficient (Dreger et al., 2016). We wanted a small, fully inspectable case in which the record itself could be shortened and the calculation repeated.
Data and methods
The source is the pedigree and genotype dataset accompanying Campbell et al. (2016), a study of recombination in Labrador Retriever–Greyhound crosses. Its PLINK pedigree file contains 237 sample IDs: 207 with two recorded parents, five with one, and 25 with neither. We used only the sample and parent IDs. We did not analyze the SNP calls, infer disease status, or use dates or clinical outcomes.
Our unit of analysis was a distinct recorded sire–dam pair. The 207 dogs with two recorded parents represent 30 such pairings; repeated offspring of the same pair count once in the main comparison. We ordered parents before offspring and built a tabular relationship matrix. For an animal i and an earlier animal j, A[i,j] = (A[sire(i),j] + A[dam(i),j]) / 2; the diagonal is A[i,i] = 1 + A[sire(i),dam(i)] / 2. Each pair’s expected offspring pedigree COI is A[sire,dam] / 2. Unknown ancestors contribute zero and are treated as unrelated founders. As an internal check, each recorded offspring’s inbreeding coefficient from the matrix matched the coefficient calculated from its two parents.
We then followed parent links to one, two, three, four, or five generations above each pair’s sire and dam. At each boundary, we treated earlier ancestors as unrecorded and recomputed the matrix. “Full record” means all ancestry present in this file; it does not mean that the dogs’ biological ancestry is complete.
Results
In this table, “apparent zero” means a pairing that reads zero at the stated depth but is nonzero in the full available record. “Below full” counts pairings with a smaller coefficient than their full-record value.
| Depth from each parent | Pairs with COI > 0 | Apparent zeros | Below full |
|---|---|---|---|
| 1 generation | 3 of 30 | 2 | 4 |
| 2 generations | 5 of 30 | 0 | 2 |
| 3 generations | 5 of 30 | 0 | 2 |
| 4 generations | 5 of 30 | 0 | 1 |
| 5 generations | 5 of 30 | 0 | 0 |
| Full available record | 5 of 30 | — | — |
The two one-generation zeros occur in pairings whose full-record COI is 3.125%. Those two pairings account for 14 recorded offspring, but they remain two pairings, not 14 independent observations. A more complex pair, PFZ19E09 × PFZ18D01, moved from 25.00% with one generation to 36.33% with the full available record.
| Recorded sire × dam | Offspring in file | 1-generation COI | 2-generation COI | Full-record COI |
|---|---|---|---|---|
| PFZ19E09 × PFZ18D01 | 1 | 25.00% | 35.94% | 36.33% |
| PFZ19E09 × PFZ19G09 | 3 | 12.50% | 15.63% | 16.41% |
| PFZ23A05 × PFZ23A07 | 9 | 0.00% | 3.13% | 3.13% |
| PFZ23D07 × PFZ23A06 | 5 | 0.00% | 3.13% | 3.13% |
| PFZ23E07 × PFZ26A06 | 4 | 12.50% | 12.50% | 12.50% |
Values in the table are rounded for reading; the 30-pair results file contains the unrounded calculated values. In a local AnimalTrace run using the published parent links, the pairing simulator returned 12.5% for PFZ23E07 × PFZ26A06 and identified their shared recorded ancestor, PFZ26F05. It also returned 0.0% within the recorded pedigree for a separate pair with an untraced sire, while flagging that missing ancestry. The product check illustrates why the number and its evidence gap need to travel together.
Interpretation and limits
The result is about record sensitivity. It does not show that any particular dog had more or less genomic inbreeding, nor that one pairing was healthy or advisable. The source is a single research colony with its own mating structure. Its 25 founder records and five one-parent records leave ancestry beyond the file unknown. A zero in the full file therefore remains a statement about this recorded pedigree, not proof of biological unrelatedness.
We report counts rather than a significance test because this is a deterministic recalculation of the available pairings, with no population sampling design or independent breed cohort. The largest values belong to specific source IDs and should not be treated as breed-wide rates. No dates are available for a temporal trend analysis.
For AnimalTrace, the practical lesson is measurable: a newly linked parent can change a pairing’s pedigree calculation, and a previously returned zero may have depended on a shallow record. Search and saved updates can bring the changed animal back for review; Match should be rerun with the current parent links and should show when ancestry remains incomplete. Our worked AnimalTrace example follows that workflow with a family subset.
Reproducibility and sources
- Source study: Campbell et al., 2016, A Pedigree-Based Map of Recombination in the Domestic Dog Genome.
- Source data: Authors’ public repository,
new_pedigree_dogs.faminsidedog_genotype_data.tar.gz. - Derived output: All 30 recorded pairings and coefficients at each depth.
- Analysis method: Download the standard-library Python script. It reads the
.famfile, checks parent references and cycles, calculates the relationship matrix, verifies each recorded offspring against its parents, and writes the derived table.
This is an AnimalTrace research note using a public dataset. It has not undergone external peer review.