Genome scanning of behavioral selection in a canine olfactory detection breeding cohort

Abstract

Research on working dogs is growing rapidly due to increasing global demand. Here we report genome scanning of the risk of puppies being eliminated for behavioral reasons prior to entering the training phase of the US Transportation Security Administration’s (TSA) canine olfactory detection breeding and training program through 2013. Elimination of dogs for behavioral rather than medical reasons was based on evaluations at three, six, nine and twelve months after birth. Throughout that period, the fostered dogs underwent standardized behavioral tests at TSA facilities, and, for a subset of tests, dogs were tested in four different environments. Using methods developed for family studies, we performed a case-control genome wide association study (GWAS) of elimination due to behavioral observation and testing results in a cohort of 528 Labrador Retrievers (2002–2013). We accounted for relatedness by including the pedigree as a covariate and maximized power by including individuals with phenotype, but not genotype, data (approximately half of this cohort). We determined genome wide significance based on Bonferroni adjustment of two quasi-likelihood score tests optimized for either small or nearly-fully penetrant effect sizes. Six loci were significant and five suggestive, with approximately equal numbers of loci for the two tests and frequencies of loci with single versus multiple mapped markers. Several loci implicate a single gene, including CHD2, NRG3 and PDE1A which have strong relevance to behavior in humans and other species. We briefly discuss how expanded studies of canine breeding programs could advance understanding of learning and performance in the mammalian life course. Although human interactions and other environmental conditions will remain critical, our findings suggest genomic breeding selection could help improve working dog populations.

Machine learning prediction and classification of behavioral selection in a canine olfactory detection program

“…We previously reported genetic mapping of elimination for behavioral reasons in the same cohort 17 . There we referenced the behavioral selection related to this cohort and the general population of “hunting line” Labrador Retrievers.…”

“…While there have been exploratory and prospective studies of new testing schemes in detection and assistance working dogs, they have not been deployed widely yet 16 . However, there are large working dog datasets of training, performance and health data from federal and private institutions that have not been thoroughly analyzed yet 15 , 17 . It thus remains possible that existing standardized datasets, which continue to be collected and are already large and thus ideal for ML, could be the most efficient and productive route to improve the understanding of behavioral traits required for working dogs.…”

“…We recently genetically mapped the trait of pre-training elimination for behavioral reasons in the same TSA cohort 17 . The actual reason for elimination in that and the present work is not clearly defined, except that behavioral and medical elimination were distinguished.…”