ABSTRACT
Nothing in biology makes sense except in the light of evolution. —T. Dobzhansky (1) This chapter describes concepts and algorithms of the Dali (distance matrix alignment)
method. Specific problems addressed are structural alignment and domain decomposition, culminating in the automated detection of remote homology. In each case, the computational problem conceptually boils down to determining the optimal size and number of clusters by optimizing a cost function, a general approach with applications in many partition problems and related to ideas of information theory. Our key thesis is that the economy of description-the objective of the clustering-reflects an underlying economy of conception in molecular evolution and thereby yields biologically relevant results.
