Preprint proposes cross-selection method balancing multiple traits via progeny distribution prediction in plant breeding

A preprint posted to bioRxiv presents a statistical framework for cross-potential selection in plant breeding programmes where improvement in one target trait risks undesirable changes in other genetically correlated traits. The authors term the approach Cross Potential Selection for Multiple Traits, and it centres on explicitly modelling the predicted progeny distribution of each candidate crossing pair — rather than simply the expected mean — using estimated genotypic values and genetic covariances across traits.

In many breeding scenarios, a single trait can be optimised relatively straightforwardly through genomic selection, but improving a target trait while keeping multiple essential traits within acceptable bounds is complicated by genetic correlations, which can cause indirect responses that move one trait in an undesirable direction when another is selected. The proposed framework aims to identify crossing combinations whose predicted offspring distributions have acceptable probabilities of meeting all specified trait requirements simultaneously.

The method has potential relevance to crop improvement contexts — for example, maintaining disease resistance or nutritional quality while selecting for yield — and the authors demonstrate its application with simulated and empirical examples. This is a preprint and has not yet been peer-reviewed.