Good practice for assignment of breeds and populations—a review DOI Creative Commons
Hélène Wilmot, Nicolas Gengler

Frontiers in Animal Science, Journal Year: 2025, Volume and Issue: 6

Published: Feb. 6, 2025

With the purpose to organize methodologies found in (recent) papers focusing on development of genomic breed/population assignment tools, this review proposes highlight good practice for such tools. After an appropriate quality control markers and building a representative reference population, three main steps can be followed develop tool: 1) The selection discriminant markers, 2) model that allows accurate animals their origin, so-called classification step, and, 3) validation developed new evaluate its performances real conditions. first step avoided when mid- or low-density chip is used, depending methodology used assignment. In case SNPs necessary, we advise use one stage define threshold selection. Then, machine learning per se , based selected available markers. To tune model, recommend cross-validation. Finally, animals, not two steps, should (e.g., with balanced accuracy probabilities), also terms computation time.

Language: Английский

Good practice for assignment of breeds and populations—a review DOI Creative Commons
Hélène Wilmot, Nicolas Gengler

Frontiers in Animal Science, Journal Year: 2025, Volume and Issue: 6

Published: Feb. 6, 2025

With the purpose to organize methodologies found in (recent) papers focusing on development of genomic breed/population assignment tools, this review proposes highlight good practice for such tools. After an appropriate quality control markers and building a representative reference population, three main steps can be followed develop tool: 1) The selection discriminant markers, 2) model that allows accurate animals their origin, so-called classification step, and, 3) validation developed new evaluate its performances real conditions. first step avoided when mid- or low-density chip is used, depending methodology used assignment. In case SNPs necessary, we advise use one stage define threshold selection. Then, machine learning per se , based selected available markers. To tune model, recommend cross-validation. Finally, animals, not two steps, should (e.g., with balanced accuracy probabilities), also terms computation time.

Language: Английский

Citations

0