Genome-Wide Association Mapping for Intelligence in Military Working Dogs: Development of Advanced Classification Algorithm for Genome-Wide Single Nucleotide Polymorphism (SNP) Data Analysis

This project collected data to genetically map superior intelligence in the military working dog. A behavioral testing regimen was developed by canine cognitive expert Dr Karen Overall (UPENN) which enabled quantitative intelligence testing of individual dogs and blood samples were taken, and genome-wide SNP typing completed by means of the Affymetrix Canine SNP (single nucleotide polymorphism) Array v2. In order to identify SNP markers for mapping of small-effect-sized genes that contribute to highly complex polygenic traits, it is necessary to develop a more robust computational method for the analysis of SNP profile data. To accomplish this, we are undertaking two parallel efforts, Biologically Guided Feature Selection and Computational Based Feature Synthesis and Classification. As a proof-of-concept, we conducted a classification analysis focused on a subset of tested canines consisting of German Shepherds, Labrador Retrievers, and Belgian Malinois. Using this new classification technique, samples from the three breeds clustered into the correct breed with an accuracy ranging from 89 - 100 %. Classification accuracy was not significantly affected by data process methods (including data cleanup methods) or SNP annotation quality, thus suggesting that this algorithm is highly robust. With further refinement and optimization, this technique could be used to classify complex phenotypes in an unsupervised manner and allow identification of associated SNP markers. Prepared in cooperation with Penn Med Translation Research Lab., Philadelphia, PA, Ohio Univ., School of Electrical Engineering and Computer Science