Format: ORAL

Sara Laura arancic 1, Nikolina Pleic 2,3, Kreimir Križanovic 4, Ivan Radosavljevic 1

1 Department of Biology, Faculty of Science, University of Zagreb, Zagreb, Croatia 2 Department of Medical Biology, School of Medicine, University of Split, Split, Croatia 3 University of Applied Sciences ASPIRA, Split, Croatia 4 Department of Electronic Systems and Information Processing, Faculty of Electrical Engineering and Computing, University of Zagreb, Zagreb, Croatia

Chouardia litardierei(Hyacinthaceae) is a bulbous, perennial species of the western parts of the Balkan Peninsula. It is characterized by a pronounced ecological plasticity since its populations are distributed across contrasting habitat types: karst poljes, wet meadows near the seashore, and dry mountain slopes. Thus,C. litardiereipresents a valuable model for addressing one of the central questions of evolutionary biology: what genomic mechanisms underlie local adaptation and ecological divergence? Here we present the results of a genome-wide association study (GWAS) that aimed at elucidating the genetic architecture of some complex traits in studied species. For this purpose, we set the common garden experiment with 214 individuals from nine selected populations, three representing each assumed ecotype. We performed comprehensive morphometric and phenological analyses and genotyped all individuals using a genotyping-by-sequencing approach. Finally, we combined all the obtained results in a comprehensive study to characterize the genotype-phenotype relations of selected traits. For the analysis, we used 24,660 SNPs, five reproduction-related morphological, and five phenological traits. All traits were considered polygenic, and GWAS analyses were performed assuming an additive genetic model. For each association analysis, we considered two different statistical approaches: the frequentist single-locus models and Bayesian multi-locus models. Within the single-locus models, for each trait, we fitted a standard linear mixed model (LMM) in GEMMA, as well as a Poisson linear mixed model in GMMAT for traits that represented count data, to correctly model the traits distribution. Within the multi-locus models, we fitted a Bayesian sparse linear mixed model (BSLMM) and a latent Dirichlet Process Regression (DPR) model for each trait. Results were visualized using Manhattan and QQ plots. Multiple candidate loci were detected for different traits, and the results are being discussed.