Genome and Transcriptome Based Prediction and Regulator Inference of Molecular and Whole-Plant Phenotypes

University Of Florida, Gainesville FL

Investigators

Abstract

Right now, plant breeders have to wait to determine if their seedlings will mature into useful plants or not. This leads to delays and increase in costs of development of new varieties that are more productive, disease resistant and tolerant to environmental stresses. In the case of forestry, this wait can be decades-long, thus hampering tree improvement for the timber, paper and bioenergy industries. Development of new cultivars could be accelerated if there was a way to tell what useful traits a plant would have based on its DNA alone. The ability to use DNA to predict the traits of a grown plant, from its development to its productivity, remains one of the recognized grand challenges of plant biology. Even more relevant and incomplete is the understanding of the chain of events that lead from DNA to trait. A complete understanding of a plant's biological system will allow the identification of the key regulatory points that impact its traits of economic importance, while also allowing their prediction. In this project novel approaches will be developed to predict the development and the growth properties of plants, and their immediate regulators. These approaches will integrate information beyond DNA to include gene expression data, and create a more complete description and prediction of trait regulation. Tools developed here will be disseminated to the community by the release of analysis software packages, and workshops will be conducted to enhance teaching of quantitative genetics and statistical methods in predictive models to the scientific community. The project incorporates a training component focused on high-school teachers and students. High-school teachers will join the investigator laboratories to learn about real-world applications of genomics and develop curriculum with support of the University of Florida Center for Precollegiate Education and Training (CPET). Students from underrepresented groups will be trained in science through CPET's Student Science Training Program. The proposed research will also provide ample opportunities to train graduate students at the intersection between quantitative and statistical genetics, and genomics. Genomic prediction of complex traits is a critical tool in animal and plant breeding. However, the ability to fully explain a phenotypic trait is still limited, in part because several dimensions of information are either not available or not incorporated to prediction models. In this project, genome sequence information will be used initially to predict gene expression, and gene expression of the ensemble of genes will be applied to predict whole-plant phenotypes. As a last step, genome sequence and transcriptome data will be integrated for prediction of phenotypes and their regulators. Thus, a model for prediction that incorporates multiple layers of information will serve as a tool for prediction of plant molecular properties (e.g. gene expression), as well as growth and developmental traits. This project will make use of a large Populus deltoides association population for which abundant genome, transcriptome and phenotypic resources are already available or will be developed here. Prediction tools will also be applicable to other species for which similar type of resources are available. The research methods to be applied to each aim will consist of four components: (1) detection of large-effect factors (e.g., individual or groups of variants with large regulatory effects) using single-factor regression methods, (2) development of prediction models that combine whole-genome information from large and small-effect factors, (3) biological validation of findings using genetic modification, and (4) statistical validation using an independent population. Genotypic and gene expression data will be publically available in NCBI's dbSNP and the sequence read archive/gene expression omnibus databases (www.ncbi.nlm.nih.gov). Analysis packages will be deposited in the R Archive (http://cran.r-project.org/), and project details and plant resource access procedures will be described in the project web site (http://sfrc.ufl.edu/forestgenomics/Kirstlab/).