Data Mining as a Tool to Infer Chicken Carcass and Meat Cut Quality from Autochthonous Genotypes.

Abstract

Simple SummaryThe present study is a meta-analysis of ninety-one research documents dealing with carcass quality characterization in autochthonous chicken genotypes. Documents were published between 2002 and 2021. Data mining methods were used to determine which variables should be considered or otherwise discarded from comprehensive carcass quality differential models to improve the study’s efficiency and accuracy. Even if the impact on carcass quality of certain variables such as chicken sex, meat firmness, chewiness, L* meat 72 h post-mortem, a* meat 72 h post-mortem, b* meat 72 h post-mortem, and pH 72 h post-mortem could be presumed, these should not be considered if strongly related variables are simultaneously considered too, to prevent redundancy problems. In contrast, carcass/cut weight, pH, carcass yield, slaughter age, protein, cold weight, and L* meat must be regarded strictly due to their high potential to explain differences and correctly classify carcass cuts across chicken genotypes. The standardization of characterization methods of minority populations (with limited censuses and lacking population structure, but well-adapted to alternative systems) enhances the possibility of success of the implementation of sustainable conservation strategies through the dissemination of knowledge on local breeds and the competitivization of their distinctive products within specific market niches.The present research aims to develop a carcass quality characterization methodology for minority chicken populations. The clustering patterns described across local chicken genotypes by the meat cuts from the carcass were evaluated via a comprehensive meta-analysis of ninety-one research documents published over the last 20 years. These documents characterized the meat quality of native chicken breeds. After the evaluation of their contents, thirty-nine variables were identified. Variables were sorted into eight clusters as follows; weight-related traits, water-holding capacity, colour-related traits, histological properties, texture-related traits, pH, content of flavour-related nucleotides, and gross nutrients. Multicollinearity analyses (VIF ≤ 5) were run to discard redundancies. Chicken sex, firmness, chewiness, L* meat 72 h post-mortem, a* meat 72 h post-mortem, b* meat 72 h post-mortem, and pH 72 h post-mortem were deemed redundant and discarded from the study. Data-mining chi-squared automatic interaction detection (CHAID)-based algorithms were used to develop a decision-tree-validated tool. Certain variables such as carcass/cut weight, pH, carcass yield, slaughter age, protein, cold weight, and L* meat reported a high explanatory potential. These outcomes act as a reference guide to be followed when designing studies of carcass quality-related traits in local native breeds and market commercialization strategies.