Implementation of a Comparative Gene Finder

Abstract

In this chapter we exemplify the implementation of a gene finder by describing the software SLAM in a little more detail. SLAM is a cross-species gene finder particularly adapted to eukaryotes, and works by simultaneously aligning and annotating two homologous sequences. The basic framework of SLAM is a generalized pair hidden Markov model, which is a seamless merging of pair hidden Markov models typically used for pairwise alignments, and generalized hidden Markov models that have been successfully implemented in several single species gene finders. We begin by detailing the structure of the program and continue by going into some of the algorithmic details of the algorithms. We finish by describing various measures used to assess the accuracy of gene finding softwares. The main purpose of such measures is not only to detect problems with the algorithms during development, but to be able to benchmark the software against other methods.