Alternative additives associated in the feeding of laying hens: performance, biometrics, bone traits, and economic evaluation-an unsupervised machine learning approach.

Given the current bans on the use of some growth promoting antibiotics in poultry nutrition, the need to use alternative additives which could replace traditional promoters in diets has arisen. The objective of this study was to evaluate the effect of alternative additives, associated or not, in replacing the antibiotic growth promoter in the diets of laying hens on performance, egg quality, biometry, bone characteristics, and economic viability. A total of 378 birds at 97 weeks of age, weighing 1691 ± 80g with an average production of 79.96 ± 4.9%, were randomly distributed and submitted to different diets: negative control - NC (no additive); positive control - PC, conventional growth promoter (Enramycin); associated organic acids (OA); symbiotic (S); Essential oil (EO); OA + S; and S+EO. The diet did not influence (P > 0.05) performance, egg quality, biometry, and bone traits. However, the use of alternative additives and their associations with the exception of S+OA, provided better economic indices when compared to NC and CP. The first component showed a negative relationship between feed conversion per mass and dozen eggs with gut length, Seedor index, egg production, and egg mass; the second component showed a positive relationship between yolk, pancreas, proventriculus, and gizzard; and, finally, the third component showed that feed consumption has a negative relationship with bone strength and deformity. The first two canonical functions were significant and discriminated 100% of the differences between the diets. Moreover, 50% of the birds were correctly classified in their group of origin, in which the positive control group (83.3%) and OA+S presented the highest rates of correct responses (66.7%). Bone deformity and bowel length were the only two variables with discriminatory power. Natural growth promoters alone or in association do not harm performance, egg quality, digestive organs biometry or bird bone characteristics, in addition to promoting greater economic return. Thus, they can be considered possible substitutes for traditional antibiotics. Finally, unsupervised machine learning methods are useful statistical techniques to study the relationship of variables and point out the main biomarkers of poultry production.