Fast and accurate detection of multiple quantitative trait Loci

Area of Science:

• Genetics
• Computational Biology
• Bioinformatics

Background:

• Standard exhaustive search methods for quantitative trait loci (QTL) mapping are computationally intensive and impractical for more than two loci.
• Existing global optimization methods, while faster, may introduce bias due to heuristic termination criteria, compromising accuracy.

Purpose of the Study:

• To introduce a novel computational scheme, PruneDIRECT, for efficient and reliable QTL scans in experimental populations.
• To overcome the limitations of exhaustive search and heuristic optimization in high-dimensional QTL mapping.

Main Methods:

• Developed PruneDIRECT, an optimization scheme incorporating a computable Lipschitz bound on a transformed objective function.
• The method allows for pruning of the search space, avoiding heuristic termination and ensuring equivalence to exhaustive search.

Main Results:

• PruneDIRECT demonstrates significant speedups, typically over 50 times faster than exhaustive search for simultaneous mapping of three QTLs.
• The speedup increases with stronger QTL effects, enabling faster detection of genetic networks.

Conclusions:

• PruneDIRECT offers a well-defined error bound and guarantees accuracy comparable to exhaustive search.
• This method enhances the feasibility of complex QTL analyses, including the detection of candidate expression QTL (eQTL) networks.