Identity-by-descent matrix decomposition using latent ancestral allele models

Area of Science:

• Population Genetics
• Quantitative Genetics
• Bioinformatics

Background:

• Accurate modeling of genetic relatedness is crucial for genetic linkage and association studies.
• Relatedness is typically inferred from marker or pedigree data, often represented by identity-by-descent matrices.
• Existing methods lack explicit representation of ancestral allele origins and their counts.

Purpose of the Study:

• To develop a method for explicitly representing ancestral allele origins and their number.
• To decompose the genetic relatedness matrix (Q) using a latent class model.
• To provide an efficient algorithm for calculating the assignment probabilities matrix (P).

Main Methods:

• Decomposition of the identity-by-descent matrix (Q) using a latent class model with K classes (latent ancestral alleles).
• Construction of an n x K matrix (P) representing individual assignment probabilities to latent classes.
• Approximation of Q by PP(T) under non-negativity and row-sum constraints on P, solved via an efficient algorithm.

Main Results:

• An efficient algorithm was developed to calculate the assignment probabilities matrix (P).
• The latent class model successfully decomposed locus-specific identity-by-descent matrices.
• The model accurately recovered known ancestry for maize inbreds, demonstrating its utility.

Conclusions:

• The proposed latent class model provides an explicit and interpretable representation of genetic relatedness.
• This method enhances the analysis of genetic structure by revealing ancestral allele origins.
• The model shows significant potential for improving genetic linkage and association studies, particularly in plant breeding.