Files in this item



application/pdfWEN-DISSERTATION-2017.pdf (2MB)
(no description provided)PDF


Title:Evaluation of soybean cyst nematode (SCN) resistance in perennial glycine species and genome-wide association mapping and genomic prediction study for SCN resistance in common bean and prediction of the short distance movement of soybean rust urediniospores through machine learning
Author(s):Wen, Liwei
Director of Research:Hartman, Glen L
Doctoral Committee Chair(s):Hartman, Glen L
Doctoral Committee Member(s):Brown, Patrick J; Lambert, Kris N; Domier, Leslie L
Department / Program:Crop Sciences
Discipline:Crop Sciences
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Perennial Glycine Species
Soybean Cyst Nematode
Disease Resistance
Genome Wide Association Mapping
Genomic Prediction
Machine learning
Soybean Rust
Abstract:Since agriculture started, there have been numerous occasions when plant diseases of crops had severe impact on human activities. From the famine caused by potato late blight (Phytophthora infestans) in Ireland in 1846, to the dramatic economic loss caused by downy mildew of grapes (Plasmopara viticola) in the Mediterranean in 1865, to the loss of the valuable banana cultivar ‘Gros Michel’ caused by Fusarium oxysporum Schlect. f. sp. cubense, plant diseases have caused significant historical and economic importance. The goal of plant disease management is to reduce the economic and aesthetic damage caused by plant diseases, and the focus of my thesis centers around studying diseases and their pathogen in an effort to supplement long-term effective management strategies for important diseases of soybean. Soybean cyst nematode (SCN; Heterodera glycines; HG) is a widely occurring and damaging pathogen with a wide host range. SCN is the leading cause of soybean yield loss in the US and it will likely become a major yield-limiting threat to common bean (Phaseolus vulgaris L.), another highly susceptible host of SCN. Developing resistant cultivars is the most cost-effective method for managing this disease. In the first chapter of my thesis, I focused on identifying additional sources of resistance to SCN in perennial Glycine species which can be potentially used for improving resistance of soybean to SCN. 13 perennial Glycine species of 282 PIs were inoculated with HG types 0, 2, and 1.2.3 first, and then 36 PIs out of this set were further evaluated by inoculating with HG type, a population that overcomes all the resistance genes in soybean. The Glycine species evaluated contains many PIs that are highly resistant to SCN with 10 species classified as immune or highly resistance to three HG types, indicating a much broader resistance in these PIs. With additional work on hybridizing the perennial Glycine species and soybean along with techniques of gene cloning and gene transfer, many of the genes in the perennial Glycine species could be used to develop additional soybean genotypes with SCN resistance. In the second chapter of my thesis, genome-wide association study (GWAS) was used to detect SNPs significantly associated with SCN resistance in the core collection of P. vulgaris and to make genomic predictions (GPs) of SCN resistance to two HG types. GWAS identified SNPs that are significantly associated with resistance to two HG types, and GP for resistance to two SCN HG types achieved high prediction accuracy. The findings in this chapter demonstrated GWAS and GP as valuable tools for developing new resistant common bean varieties with SCN resistance in the future. Epidemiology studies concerning the environmental and biological factors affecting disease entry, establishment and development are also extremely important for the successful management of diseases. The third chapter of my thesis focuses on developing mathematical models to predict the disease epidemic of soybean rust (Phakopsora packyrhizi), another devastating fungal disease of soybean with rapid establishment and development in the fields, using environmental and biological variables. Four machine learning models, including Absolute Shrinkage and Selection Operator (LASSO) method, zero-inflated Poisson/regular Poisson regression model, random forest, and neural network were built and compare to describe deposition of urediniospores collected in passive and active traps. The high prediction accuracy of some of the models demonstrated the applicability of machine learning in disease risk assessment, and the finding of this project is potentially helpful in guiding farmers to make proper and in-time disease management decisions.
Issue Date:2017-07-03
Rights Information:Copyright 2017 Liwei Wen
Date Available in IDEALS:2017-09-29
Date Deposited:2017-08

This item appears in the following Collection(s)

Item Statistics