Transcriptome-based distance measures for grouping of germplasm and prediction of hybrid performance in maize

Abstract

Grouping of germplasm and prediction of hybrid performance and heterosis are important applications in hybrid breeding programs. Gene expression analysis is a promising tool to achieve both tasks efficiently. Our objectives were to (1) investigate distance measures based on transcription profiles, (2) compare these with genetic distances based on AFLP markers, and (3) assess the suitability of transcriptome-based distances for grouping of germplasm and prediction of hybrid performance and heterosis in maize. We analyzed transcription profiles from seedlings of the 21 parental maize lines of a 7 x 14 factorial with a 46-k oligonucleotide array. The hybrid performance and heterosis of the 98 hybrids were assessed in field trials. In cluster and principal coordinate analyses for germplasm grouping, the transcriptome-based distances were as powerful as the genetic distances for separating flint from dent inbreds. Cross validation showed that prediction of hybrid performance with transcriptome-based distances using selected markers was more precise than earlier prediction models using DNA markers or general combining ability estimates using field data. Our results suggest that transcriptome-based prediction of hybrid performance and heterosis has a great potential to improve the efficiency of maize hybrid breeding programs.