Clustering methods for accurate DNA base-calling

Abstract

Routinely extending the useful read-lengths of DNA chromatograms beyond 1 kps by employing intelligent base-calling algorithms will be extremely useful to genomics research because in many cases the entire coding region of a gene could fit into a single long read. By segmenting the chromatograms into base-call "events" to be labeled in terms of the number of bases they represent, base-calling as a pattern classification problem is formulated. An overview of two unsupervised clustering methods that could be used for its solution is presented.