CORRECTING MODEL MISSPECIFICATION IN RELATIONSHIP ESTIMATES

Area of Science:

• Genetics
• Bioinformatics
• Population Genetics

Background:

• Large genetic datasets from biobanks and direct-to-consumer testing contain many related individuals.
• Current methods for estimating genetic relationships have limitations, with a practical ceiling around ten degrees (5th cousins).
• Existing relationship estimators are misapplied, leading to biased estimates for distant relatives due to ignoring shared DNA segments identically by descent (IBD).

Purpose of the Study:

• To correct and improve the accuracy of genetic relationship estimation.
• To extend the detectable range of genetic relationships beyond current limitations.
• To enable inference of more distant ancestral connections and their historical context.

Main Methods:

• Derived a corrected likelihood function that conditions on the presence of at least one shared segment of DNA identically by descent (IBD).
• Reformulated relationship inference to account for multiple common ancestors, addressing issues caused by pedigree collapse.
• Applied corrected estimators to genetic data to assess pairwise relationships and their uncertainty.

Main Results:

• The corrected likelihood significantly reduces bias in pairwise relationship estimates for distant relatives.
• The reformulated method extends the detectable relationship range to 200 degrees or more.
• This advancement pushes the time-to-common ancestor inference back to approximately 3,000 years or more.

Conclusions:

• Previous assumptions about the limits of genetic relationship detection are incorrect due to estimator misapplication.
• The developed methods accurately estimate distant genetic relationships, overcoming limitations of existing approaches.
• This research opens new possibilities for understanding deep ancestral history and pre-historical population movements.