Computational biology or compbio can be defined as the science of building algorithms and tools to uncover and model biological processes. Some don't make a distinction between Bioinformatics and compbio, but we have separated the definitions on our site.
Some of the major questions include studying evolution of macromolecular molecules DNA and proteins through sequence alignment and reconstructing evolutionary history and taxa relationships through phylogenetics.
The study and prediction of protein structures is another major area of intense research (and is woefully under supported in BioPerl). Many tools for homology modeling and protein threading allow one to compare how protein sequences fold into structures.
The burgeoning field of systems biology strives to integrate gene expression, metabolomics, and proteomics into a synthesis of understanding about how the genetic program works so that one or many cells interact to form and maintain an organism.
Many other things might be classified into computational biology since they use computers and are about biology, but generally have their own field wih a great deal of history. This includes fields like Population Genetics the study of variation within and between populations of a species. This work is supported by the PopGen modules in BioPerl. QTL mapping is used to identify loci in the genome that contribute to a quantitative trait. Association mapping and linkage mapping are useful tools for unraveling the genetic basis of complex traits.
Ecological modeling can also be computationaly intensive work but generally is not considered computational biology, perhaps because of the mainly genetic focus of the field today.