Exhausted model selection for multitrait mapping QTL: application to barley (Hordeum vulgare L.) dataset

Abstract

In this study, a QTL mapping method combining multiple traits is proposed. The key feature of the new method is that it tests all possible traits and chooses those affected traits in the model for a specific QTL, whereas these “working” traits may be different for different QTL. The BIC model selection criteria was used to choose the best model and to ascertain the “working” traits; then the “working” traits were combined together to detect QTL affecting these traits, by which it can boost QTL signals and increase statistical power. The new method has been applied to Barley dataset, which contains eight traits. The results showed that sometimes a QTL might affect more than one trait but not always affected all traits; the QTL detection power with new method was increased and 4 more QTLs were detected compared with single-trait method; generally, − logP values were much higher than those by single-trait analysis, but they were highly correlated, confirmed the unbiaseness of the new multitrait method. These results suggest that the developed multitrait method is effective for QTL mapping. The computational program is written in Fortran language for convenience to use, which is available on request.