Core Hunter 3: Flexible Core Subset Selection

Many methods and measures have been proposed to sample and evaluate core collections that allow effective utilization of large genetic resources. Core Hunter is a multi-purpose core subset selection tool that uses local search algorithms to optimize one or more criteria, depending on the purpose of the core. Distance based evaluations are often used because they are easy to interpret and can be computed for both genetic marker data and phenotypic traits. Core Hunter 2 (CH2) maximized average and minimum distance but it was later suggested to maximize average entry-to-nearest-entry (E-NE) distance to obtain diverse cores, or to minimize average accession-to-nearest-entry (A-NE) distance to maximally represent all accessions from the entire collection. We propose Core Hunter 3 (CH3) in which these improved measures have been incorporated. It is verified whether the complex mixed replica (MixRep) search used by CH2 is still needed when maximizing the new E-NE criterion instead of minimum distance, and results of CH3 and CH2 are compared. We also evaluate the ability of CH3 to simultaneously maximize diversity and either representativeness or allelic richness and compare the results with those of the GDOpt and SimEli methods.

A simple stochastic hill climber is already able to find high E-NE values, but the diversity of the core is slightly further increased, and variability across independent samples further reduced, by using the parallel tempering algorithm. A more complex algorithm like MixRep is not needed to optimize this measure. CH3 yields higher E-NE values than CH2 while still ensuring a high minimum distance, and is faster for large datasets. In addition, CH3 can sample equally representative cores (low A-NE value) as GDOpt, which was specifically designed for this purpose, and is able to construct cores that are simultaneously more diverse and either are more representative or have higher allelic richness than those obtained by SimEli.

Core Hunter 3 is a flexible, fast and broadly applicable core subset selection tool that samples multi-purpose cores based on genetic marker data or phenotypic traits. It combines and outperforms the strengths of other methods and can easily be extended with new evaluation measures without the need to alter the underlying optimization algorithms. Core Hunter 3 is freely available on http://www.corehunter.org.