Disentangling clustering configuration intricacies for divergently selected chicken breeds

Divergently selected chicken breeds are of great interest not only from an economic point of view, but also in terms of sustaining diversity of the global poultry gene pool. In this regard, it is essential to evaluate the classification (clustering) of varied chicken breeds using methods and models based on phenotypic and genotypic breed differences. It is also important to implement new mathematical indicators and approaches. Accordingly, we set the objectives to test and improve clustering algorithms and models to discriminate between various chicken breeds. A representative portion of the global chicken gene pool including 39 different breeds was examined in terms of an integral performance index, i.e., specific egg mass yield relative to body weight of females. The generated dataset was evaluated within the traditional, phenotypic and genotypic classification/clustering models using the k-means method, inflection points clustering, and admixture analysis. The latter embraced SNP genotype datasets including a specific one focused on the performance-associated NCAPG-LCORL locus. The k-means and inflection points analyses showed certain discrepancies between the tested models/submodels and flaws in the produced cluster configurations. On the other hand, 11 core breeds were identified that were shared between the examined models and demonstrated more adequate clustering and admixture patterns. These findings will lay the foundation for future research to improve methods for clustering as well as genome- and phenome-wide association/mediation analyses.