An effective filter for IBD detection in large data sets.

Area of Science:

• Genomics
• Computational Biology
• Population Genetics

Background:

• Identity by descent (IBD) inference identifies shared genomic segments due to common ancestry.
• Accurate IBD detection is crucial for genetic studies, including disease gene mapping and population history research.
• Current IBD inference methods prioritize accuracy for short, ancient segments but suffer from high computational costs, limiting scalability to large cohorts.

Purpose of the Study:

• To introduce SpeeDB, a novel method for rapid IBD detection in large, unphased genotype datasets.
• To overcome the computational limitations of existing IBD inference techniques for large-scale genomic studies.

Main Methods:

• SpeeDB employs an efficient opposite-homozygous filter to rapidly exclude non-IBD genomic segments.
• The filter is applied to a database of individuals relative to a target individual.
• Remaining segments are then processed by standard IBD detection algorithms.

Main Results:

• SpeeDB filtered out 99.5% of genomic regions while retaining 99% of true IBD segments in simulated 4 cM regions.
• When applied to simulated fourth cousins, SpeeDB achieved a 10,000x speedup in IBD detection compared to unfiltered methods.
• The SpeeDB-filtered approach maintained 99% of true IBD segments in the final output.

Conclusions:

• SpeeDB offers a computationally efficient solution for IBD detection in large cohorts.
• The method significantly reduces running time without substantial loss of true IBD segment detection.
• SpeeDB facilitates large-scale genomic studies previously hindered by computational demands.