algorithms

This is THE book on string algorithms; covers all the normal exact match algs (Z, BM, KMP) and then goes on to discuss suffix trees in great depth (but with great clarity!). The second half of the book deals with inexact matching mostly using dynamic-programming-based algs. Some of the stuff generalizes nicely to non-string DP algs as well. Worth the investment just for increasing "algorithmic maturity", not to mention Gusfield's gift for clear exposition makes it a pleasant read.

If you haven't read this book, you don't know biological string matching. The book's focus is clearly on string algorithms, but the author gives good biological significance to the problems that each technique solves. I came away from this book understanding the algorithms, but also knowing why the algorithms were valuable.

No, there isn't any real source code here. That should not be a problem - this book aims above the cut&paste programmer. The book in meant for readers who can not only understand the algorithms, but apply them to unique solutions in unique ways.

String matching is far too broad a topic for any one book to cover. The study can include formal language theory, Gibbs sampling and other non-deterministic optimizations, and probability-based techniques like Markov models. The author chose a well bounded region of that huge territory, and covers the region expertly. The reader will soon realize, though, that algorithms from this book work well as pieces of larger computations. The book's chosen limits certainly do not limit its applicability.

By the way, don't let the biological orientation put you off. DNA analysis is just one place where string-matching problems occur. The author motivates algorithms with problems in biology, but the techniques are applicable by anyone that analyzes strings.

If you like definition-theorem-proof-example and exercise books, Gusfield's book is the definitive text for string algorithms. The algorithms are abstracted from their biological applications, and the book would make sense without reading a single page of the biological motivations. Gusfield aims his book at readers who are fluent in basic algorithms and data structures (at the level of Cormen, Leisersohn and Rivest's excellent text). The exercises are wonderfully illustrative, being neither trivial nor impossible.

All of the major exact string algorithms are covered, including Knuth-Morris-Pratt, Boyer-Moore, Aho-Corasick and the focus of the book, suffix trees for the much harder probem of finding all repeated substrings of a given string in linear time. In addition to exact string matching, there are extensive discussions of inexact matching. Even the discussions of widely known topics like dynamic programming for edit distance are insightful; for instance, we find how to easily cut space requirements from quadratic to linear. There is also a short chapter on semi-numerical matching methods, which are also of use in information retrieval applications. Inexact matching is extended to the threshold all-against-all problem, which finds all substrings of a string that match up to a given edit distance threshold. The theoretical development concludes with the much more difficult problem of aligning multiple sequences with ultrametric trees, with applications to phylogenetic alignment for evolutionary trees (an approach that has also been applied to the evolution of natural languages).

Note that there is no discussion of statistical string matching. For that, Durbin, Eddy, Krogh and Mitchison's "Biological Sequence Analysis: Probabilistic Models of Proteins and Nucleic Acides" is a good choice, or for those more interested in language than biology, Manning and Schuetze's "Statistical Natural Language Processing". There is also no information on more structured string matching models such as context-free grammars, as are commonly used to analyze RNA folding or natural language syntax. Luckily, Durbin et al. and Manning and Schuetze also provide excellent coverage of these higher-order models in their books.

This book is not about efficient implementation. If you need to build these algorithms, you'll also need to know how to write efficient code and tune it for your needs. This is an algorithms book, pure and simple.

As a computer scientist, I found the discussions of computational biology to be more enlightening than in other textbooks on similar topics such as Durbin et al., because Gusfield does not assume the reader has any background in cellular biology. Instead, he provides his own clear and gentle introductions illustrated with algorithms, applications, open problems and extensive references. Like most Cambridge University Press books, this one is beautifully typeset and edited.