Armidale Animal Breeding Summer
Course
Venue:
Dates: in early February each
year
Course Audience: Postgraduate students and other professionals
Armidale Animal Breeding Summer
Course 2012
Professor Daniel Gianola, University Of Wisconsin
Professor Gustavo de los Campos, University of Alabama
at Birmingham
Statistical Methods for Genome-Enabled Selection Materials
Venue:
University of
Dates:
Monday 6 February - Friday 10 February 2012
Content:
In
this course we focus on the problem of predicting complex traits using highly
dimensional pedigrees, molecular markers (e.g., SNPs, sequences) and phenotypic
records. After an overview of different paradigms that have dominated the field
of quantitative genetics in the last century, we will discuss the opportunities
and challenges posed by highly dimensional genomic data from a predictive
perspective, and will introduce alternative statistical learning techniques to
confront these challenges. The toolkit will include parametric (e.g., linear
Bayesian regression models) and some semi-parametric procedures (e.g.,
Reproducing Kernel Hilbert Spaces and Neural Networks). Methods will be
introduced and discussed in the morning lectures and practical applications
using real data and publicly available software will be offered in the
afternoon labs.
Requirements.
This
course is designed for advanced PhD students and postdoctoral fellows with
background in regression methods, statistical distributions, Bayesian Inference
and quantitative genetics, although some review will be provided. Labs will be
based on R (http://www.r-project.org/). Basic exposure to the R environment is
required.
Linear models and
Ordinary Least Squares (OLS: review and algorithms).
Performance of
OLS estimates in ‘large p with small n’ regressions.
Some strategies
to confront the limitations of OLS estimates:
Subset selection
Shrinkage
estimation procedures
Ridge Regression
Penalized
Regression
Computing Ridge
Regression Estimates using various strategies in R
Effect of
Regularization on Goodness of Fit and MSE
The Hat Matrix
and the Linear Model as a “Local Smoother”
Bayesian view of
Ridge Regression
Validation
Methods
Training-Testing
(TRN-TST)
Replicated
TRN-TST
Cross-validation
Choosing optimal
shrinkage
-
Heritability-based rules
-
Bayesian Approach
-
Cross-validation methods
Other Bayesian
Shrinkage Estimation Methods
The Bayesian
Alphabet (BayesA, BayesB,
Bayesian LASSO, Spike-Slab methods)
Methods
Comparison:
Simulated Example
Real Data Example
Reproducing
Kernel Hilbert Spaces (RKHS) methods
Reproducing
Kernel Hilbert Spaces Methods
Choice of Kernel
Automatic Kernel
Selection
Inquiries
Julius van der Werf
Animal Genetics, UNE
phone: 02 6773 2092
fax 02 6773 3922
Material
of previous years:
Armidale Animal Breeding
Summer Course 2011 Materials
Statistical methods and design in
plant breeding and genomics: Ian Mackay
IBD inference in genome association
studies: Elizabeth Thompson
Armidale Animal Breeding Summer Course
2010 Materials
Application of
evolutionary algorithms to solve complex problems in quantitative genetics
and bioinformatics: Brian Kinghorn, Cedric Gondro
Bayesian methods in genome association studies: Dorian Garrick, Rohan
Fernando
Armidale Animal Breeding Summer Course
2009 Materials
Quantitative Genetic Theory
and Analysis- Selection
Theory: Bruce
Walsh
Quantitative
Genetic Models for social interaction and inherited variability: Piter
Bijma
Armidale Animal Breeding Summer Course
2008 Materials
Genomic Selection:
Ben Hayes
Armidale Animal Breeding Summer Course
2007 Materials
Generalized Linear Mixed Models:
Steve Kachman
Armidale Animal Breeding Summer Course
2006 Materials
Gene Expression:
Toni Reverter
Breeding Program Design:
Graser, James, Van der Werf
Armidale Animal Breeding Summer Course
2003 Materials
