Key message Transposon insertion polymorphisms can improve prediction of complex agronomic traits in rice compared to using SNPs only, especially when accessions to be predicted are less related to the training set. Abstract Transposon insertion polymorphisms (TIPs) are significant sources of genetic variation. Previous work has shown that TIPs can improve detection of causative loci on agronomic traits in rice. Here, we quantify the fraction of variance explained by single nucleotide polymorphisms (SNPs) compared to TIPs, and we explore whether TIPs can improve prediction of traits when compared to using only SNPs. We used eleven traits of agronomic relevance from by five different rice population groups (Aus, Indica, Aromatic, Japonica, and Admixed), 738 accessions in total. We assess prediction by applying data split validation in two scenarios. In the within-population scenario, we predicted performance of improved Indica varieties using the rest of Indica accessions. In the across population scenario, we predicted all Aromatic and Admixed accessions using the rest of populations. In each scenario, Bayes C and a Bayesian reproducible kernel Hilbert space regression were compared. We find that TIPs can explain an important fraction of total genetic variance and that they also improve genomic prediction. In the across population prediction scenario, TIPs outperformed SNPs in nine out of the eleven traits analyzed. In some traits like leaf senescence or grain width, using TIPs increased predictive correlation by 30–50%. Our results evidence, for the first time, that TIPs genotyping can improve prediction on complex agronomic traits in rice, especially when accessions to be predicted are less related to training accessions.

Transposable Elements Polymorphisms (TIPs) are significant sources of genetic variation. Previous work has shown that TIPs can improve detection of causative loci on agronomic traits in rice. Here, we quantify the fraction of variance explained by Single Nucleotide Polymorphisms (SNPs) compared to TIPs, and we explore whether TIPs can improve prediction of phenotypes when compared to using only SNPs. We used eleven traits of agronomic relevance from by five different rice population groups (Aus, Indica, Aromatic, Japonica and Admixed), 738 varieties in total. We assess prediction by applying data split validation in two scenarios. In the within population scenario, we predicted performance of improved Indica varieties using the rest of Indica and additional samples. In the across population scenario, we predicted all Aromatic and Admixed samples using the rest of populations. In each scenario, Bayes C and a Bayesian reproducible kernel Hilbert space regression were compared. We find that TIPs can explain an important fraction of total genetic variance, often more than the fraction explained by SNPs, and that they also improve genomic prediction, especially in the across population prediction scenario, where TIPs outperformed SNPs in nine out of the eleven traits analyzed. In some phenotypes like leaf senescence or grain width, using TIPs increased predictive correlation by 40%. Our results evidence, for the first time, that TIPs genotyping can improve prediction on complex agronomic traits in rice, especially when samples to be predicted are less related to training samples.

As a consequence of the process of domestication, wild and domestic individuals are adapted to very different environmental conditions. Although the phenotypic consequences of domestication are observable, the genetic causes are not evident in many cases. Artificial selection could be modifying the selection coefficients of new and standing variation in the population under domestication. Here, we aim to detect a genome-wide signal of domestication under a model of polygenic adaptation. We use forward simulations to investigate the 1D and 2D site frequency spectra (SFS) of mutations in two populations (Wild and Domestic) with divergent histories (demographic and selective) following a domestication split. We simulate ten different scenarios, varying the strength of selection upon beneficial mutations and the proportion of mutations whose selection coefficients change after domestication. First, we describe that in domesticated populations selection at linked sites needs to be invoked to explain the SFS of neutral mutations and that the mode of linked selection affecting the neutral SFS depends on the duration of the domestication bottleneck. Second, we find that some aspects of the full distribution of fitness effects (DFE), such as the shape and strength of the deleterious DFE, are accurately estimated in both populations when using only the 1D-SFS. However, the detection of significant differences in the beneficial DFE between populations remains challenging in most, but not all simulated scenarios when only the 1D-SFS is used. Third, when considering the 2D-SFS and a new joint DFE model, we are able to detect more subtle differences in the full DFE that are hidden in the 1D-SFS analysis. In conclusion, our work highlights the strengths and limitations of detecting a polygenic signal of domestication under a variety of domestication scenarios and genetic architectures.