Customer Reviews

Genetic Analysis of Complex Traits Using SAS

5 star:		(0)
4 star:		(1)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

Regardless of how one feels about SAS as a programming language, it is readily apparent that it is very popular in areas such as financial and biological modeling. This book gives an introduction to how it is used in genetic analysis, and even though each chapter is written by a different author, the book can be useful to those (such as this reviewer) who are not experts in genetics but who may be called upon to apply their mathematical and statistical knowledge to problems in genetics (but using SAS instead of some other programming language to do so). Although the book assumes a thorough knowledge of genetics, it can still be read profitably by anyone who has a background in SAS and some knowledge of genetics. Being an interpreted language, SAS performance can be a problem with many applications, and its value in science is questionable for projects that require heavy computational power. For medium-sized projects though it can be helpful, even though its semantics can be hard to get used to for those who have programmed in more object-oriented environments.

SAS has been used widely to perform statistical studies in genetics using "classical" tools such as multivariate analysis and maximum likelihood, but there is one chapter in this book where Bayesian inference techniques are used for genetic analysis. In addition, and this makes the discussion in the chapter even more valuable, is that the estimation of the posterior distribution is done using Markov chain Monte Carlo (MCMC) techniques. The first genetics problem on which this is done regards two-point linkage analysis where Bayesian inference is used to estimate the recombination rate in a backcross between two completely homozygous lines for each of two loci. Even though this problem has an analytical solution, the authors use a simple Monte Carlo simulation to estimate the posterior mean and variance of the recombination rate to motivate how SAS can be used in this case.It should be pointed out here that the authors use SAS PROC Capability in their code and not all readers have this in their SAS implementation, but it can be replaced by PROC Univariate with no problems. This problem is generalized to the case of where there are three linked marker loci, with Bayesian inference and MCMC (via the Metropolis-Hastings algorithm) used to estimate the loci order and the recombination rates between the markers. The authors give the actual SAS code to implement this analysis, which is very readable (in spite of the ancient and annoying "goto" statements that are used within it). MCMC techniques are essential though in more general problems where analytical solutions are not possible. This is the case for a general genetic map construction that the authors discuss but do not give the explicit SAS code for (but it can be found on the Website that is associated with the book). They discuss briefly the pitfalls in doing MCMC for this case, and give a few alternatives. Bayesian inference is then applied to QTL analysis for the simple case of a single QTL model for backcrossing.