The problem of ascertainment for linkage analysis

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Ascertainment corrections in genetic linkage analysis are typically deemed unnecessary when individuals are selected based on trait phenotype alone.
• Standard linkage analysis methods condition on trait data, aligning with the ascertainment assumption-free (AAF) method.

Purpose of the Study:

• To investigate the necessity of ascertainment corrections under proband-dependent (PD) sampling in genetic linkage analysis.
• To determine if standard linkage methods yield biased estimates when pedigree structure depends on proband selection.

Main Methods:

• Theoretical analysis of likelihood calculations in genetic linkage analysis.
• Examination of the impact of proband-dependent sampling on parameter estimation.

Main Results:

• Conditioning on all trait data is not a valid ascertainment assumption-free (AAF) method under proband-dependent (PD) sampling, except for single ascertainment.
• PD sampling leads to asymptotically biased estimates of genetic parameters, including recombination fraction (R) and trait distribution parameters (theta).
• Lod scores and mod scores yield biased estimates of R and theta when data are collected under PD sampling.

Conclusions:

• The assumption of no necessary ascertainment correction is invalid for proband-dependent (PD) sampling in genetic linkage analysis.
• Bias in genetic parameter estimates is likely prevalent in most linkage datasets due to widespread PD sampling.
• The complexity of PD sampling may render the formulation of a correct likelihood conditional on observed pedigree structure intractable.