Customer Reviews

Data Analysis Tools for DNA Microarrays

5 star:		(9)
4 star:		(3)
3 star:		(2)
2 star:		(0)
1 star:		(0)

 

 

The targeted audience of this book is biologists who are eager to get an understanding of the analysis tools they use for microarrays. The book does an excellent job addressing this tier of audience.

The book has plenty of examples. Almost all the examples, whether fake or real, are microarray-related. Whenever needed, figures or charts are provided to illustrate ideas. A few chapters that introduce basic statistical concepts provide solved problems and exercises. All these efforts are worthwhile making difficult statistical concepts easy to understand in the context of microarrays and making the book especially valuable for biologists who do not have strong background in statistics.

This book has an emphasis on major statistical aspects of microarray data analysis. There are 17 chapters in this book. About 8 of them are directly related to statistics. Especially, there is one whole chapter devoted to multiple hypothesis testing, one chapter for ANOVA, and one chapter for experimental design. The above subjects are presented in a thorough, yet easy-to-follow style. Statistical issues are often not well addressed in published papers using microarrays. This book on microarray data analysis does an excellent job emphasizing this aspect.

The title of the book indicates "data analysis". However, since this is not a clearly defined term, you should be aware that the book only deals with "the bare minimum" of data analysis. That is routines, such as normalization, transformation, statistical testing, and clustering, that have to be carried out each and every time. Exploratory data visualizing and data mining algorithms are not covered thoroughly in this book. For example, principal component analysis (PCA) is presented as a subsection of a chapter. It does not provide explanations on concepts such as loading factors nor scree test. Series data (e.g. time series) are on two pages only and there is no mention of Fourier transformation. Support vector machine (SVM), which is widely used today as a supervised classification method, is not presented at all.

As I mentioned at the beginning, the targeted audience is biologists. If you are a statistician or a bioinformatician who wants to mathematically explore data analysis algorithms, you should look somewhere else. You may be disappointed that many concepts are not rigorously or accurately defined in this book. For example, the book uses capital letters to denote random variables. But the concept of random variables is not rigorously defined in the book. One of the consequences is the weak definition of mathematical expectation. Another example is the inflation of Type I error rate. On page 220, the author claims that the probability of "drawing the correct conclusion" is 1 - p, where p is the calculated probability of a statistic versus a parameter. However, if the probability of making a correct conclusion excludes the probability of making Type II errors, 1 - p should be stated as the probability of not making Type I errors.

In summary, this is a good book on microarray analysis tools for biologists using microarrays. However, people who are seeking in-depth descriptions of these algorithms should look somewhere else.

I'm more than 2/3 through the book and I've never encountered a topic that I feel could have been better presented. My definition of a Great book is that I can understand and follow it, and this definitely is a Great book! Thanks to the author for writing such readable text. This text has not made it to my bookshelf at work, it stays on my desk.

Being a book worm, as soon as I started working with microarrays I bought a bunch of books on the subject. After six months working with this technique and reading chapters on all the books I've bought I can say with certainty that Draghici's is the best introductory book on microarrays. Other books around are better at describing protocols or explaining the math involved in microarray data analysis but Draghici's book does a very good job at explaining how to analyse microarray data for the biologist (and maybe for other publics but statisticians). Everytime some friend ask me for hints on chapters or books to read for learning (or re-learning) statistics I suggest this book. The first chapters are an excellent review of the basics of statistics necessary for day to day practice. The only complain I have is that the shareware software that comes with the book does not work anymore (it's trial period has already expired and therefore it is not possible to install it even if you get a brand new book). I read this book from cover to cover and I think that, considering how readable it is, anyone could do it.

I have had the book for about a month now and I consult it quite frequently. Great coverage of Microarray Data Anlysis. It manages to be thourough without being dry or using excessive jargon. It's very readable and useful for both novices and experienced readers.

It's main strength lies in the use of excellent examples that show the main pitfalls encountered in analyzing microarray data. It has great coverage of statistics and their potential misuse and misunderstanding when they are applied to gene expression data sets. The experimental design section is especially helpful for researchers that are designing a project.

The graphics are excellent and the book is printed on good quality paper.

The book includes two CD's with demo versions of several commercial software packages.

Overall a great buy.

A much needed book for the biologist interested in using DNA/protein microarrays. Examples are specific for microarrays. The material starts from ground zero and begins
with image analysis. All major methods for analysis are discussed.
Well worth the cost, quality graphics, includes software (have not used as yet).
A must read before discussing experimetnal design with your stats person.

I had to buy this book for a class, but I kept because it has very interesting information about basics of DNA microarrays but also data analysis. It is definitely very useful, I recommend.

I have entered the area of microarray data analysis three years ago, having an engineering/machine learning background which includes good knowledge of statistics. After reading many journal papers about particular algorithms for microarray data analysis, I felt the need to read a book so that I could get the big picture of the field. At the beginning I was skeptical about reading Draghici's book because it was recommended to me as "excellent" by a biologist. I was pretty sure that given my background I will get bored of it quickly. My intuition failed me in this case because after reading it, I found it too as being far from ordinary, and answering my needs as well.

The book is an easy-to-follow introduction to the area of microarray data analysis covering areas from image analysis and preprocessing, to differential expression, clustering, and high level analysis such as ontological analysis. The book is particularly useful in underlying common pitfalls with microarray data. Examples include failing to correct for multiple testing in microarray experiments and the misuse or overuse of the clustering algorithms. Abounding examples and clear illustration are given to support every single aspect treated in the text. In my opinion, graduate level students in biology, bioinformatics and statistics can greatly benefit from the lecture of this book.

Another positive aspect is the fact that, with the exception of one chapter about the available commercial software, this book was written by just one author. This gives a continuity of ideas and a consistency of notations and terms throughout the entire book. This is usually not found in many other books on this topic as they are sometimes just edited collections of chapters written independently by different authors (see for instance the text by Berrar et. al which has about 40 contributors).

A great incentive for me in writing this review was reading an overzealous critique to this book, written by Eric Wu in this webpage. I found some of his comments to be particularly misleading and out of context. For instance he says "the book only deals with the bare minimum of data analysis". Compared with other books in the field, the topics about data analysis covered in the book are not only more numerous but much more thoroughly explained. This book does not expedite the reader to some references but cares about explaining the things. If this book is the "bare minimum" at 500 pages, how is Mr. Wu going to characterize the other well known books in the field such as Knudsen, Simon, Speed, Baldi, etc. which have at most half as many as this book has. Knudsen, for instance, takes the reader from absolute measurements to and including ANOVA in 17 pages. Draghici covers the same topics in 7 chapters or about 250 pages, and that would be without counting the chapters on the basic statistics or image analysis. Another example of biased assessment is when Mr. Wu says "Exploratory data visualizing and data mining algorithms are not covered thoroughly in this book. For example, principal component analysis (PCA) is presented as a subsection of a chapter." The PCA description in the book is more than just fine to me. The book is not supposed to be an encyclopedia of statistics. What the reader needs to know is how PCA can help with the visualization of these multidimensional data sets and not necessarily give all the details about PCA.

A last example I give of superficial judgment in Mr. Wu's view is the so called "inflation of Type I error rate". Mr. Wu says: "... if the probability of making a correct conclusion excludes the probability of making Type II errors, 1 - p should be stated as the probability of not making Type I errors".. In general, this statement would be true. However, the paragraph from the book to which Mr. Wu is referring to actually starts by saying: "When the t statistic for a gene is more extreme than the threshold..." etc. If the observed statistic is more extreme than the threshold, the statistical reasoning requires us to reject the null hypothesis. In this case type II errors (false negatives) CANNOT occur. Hence, in this case, the probability of drawing the correct conclusion is indeed 1-p, exactly as stated in the book.

Overall, I find that the value you get per dollar spent when buying this book is high, and thereby I would strongly recommend it.

Dr. Adi L. Tarca, Windsor (CANADA)

When entering the minefields of microarray data analysis, one has to understand and keep up with state-of-the-art technologies and interdisciplinary literatures. A background in molecular biology is clearly not enough to evaluate the pro and cons of the various statistical methods for selecting truly modulated candidate genes in a given experimental biological system. Choosing between the available analysis software's is not an easy task either. Draghici presents a complete visit of the microarray underworld by initiating the reader to all the facettes of this domain. From the fundamentals of slide production and target hybridization to image processing, statistical analysis, experimental design, data management and biological interpretation, all aspects treated herein are described with pertinent details. Draghici slowly, but successfully, tames the reticent molecular biologist to the arid world of statistics and even entertains the reader with anecdotes and humoristic citations.
Clearly written, with appropriate mathematical examples for each topic, this book even includes exercises at the end of some chapters, for the zealous student sleeping in all of us. It constitutes a very good didactic tool and the included CD's allow a good peek in some of the available image/data analysis software's on the market.
As a core facility manager and eternal student, I strongly recommend Draghici's book to life scientists and students who are struggling with statistical analysis and data mining techniques.

Brigitte Malette, Ph. D.
Project Leader, Microarray Platform
Centre for Structural and Functional Genomics
Concordia University
Montreal

Draghici managed to write a manual on applying microarray (data) with a great feeling for explanation of hard issues. The book is relatively easy to read, very complete and covers most, if not all, analysis techniques that are currently around for microarrays.

Highly recommendable!

I cannot say enough about this book. It is the best, every detail is explained. This author does a very good job not only with the statistics but with the mcroarray as well. Very easy to follow and understand.