This article is part of the supplement: EGASP '05: ENCODE Genome Annotation Assessment Project
AUGUSTUS at EGASP: using EST, protein and genomic alignments for improved gene prediction in the human genome
Institut für Mikrobiologie und Genetik, Universität Göttingen, Goldschmidtstraße, 37077 Göttingen, Germany
Genome Biology 2006, 7(Suppl 1):S11 doi:10.1186/gb-2006-7-s1-s11Published: 7 August 2006
A large number of gene prediction programs for the human genome exist. These annotation tools use a variety of methods and data sources. In the recent ENCODE genome annotation assessment project (EGASP), some of the most commonly used and recently developed gene-prediction programs were systematically evaluated and compared on test data from the human genome. AUGUSTUS was among the tools that were tested in this project.
AUGUSTUS can be used as an ab initio program, that is, as a program that uses only one single genomic sequence as input information. In addition, it is able to combine information from the genomic sequence under study with external hints from various sources of information. For EGASP, we used genomic sequence alignments as well as alignments to expressed sequence tags (ESTs) and protein sequences as additional sources of information. Within the category of ab initio programs AUGUSTUS predicted significantly more genes correctly than any other ab initio program. At the same time it predicted the smallest number of false positive genes and the smallest number of false positive exons among all ab initio programs. The accuracy of AUGUSTUS could be further improved when additional extrinsic data, such as alignments to EST, protein and/or genomic sequences, was taken into account.
AUGUSTUS turned out to be the most accurate ab initio gene finder among the tested tools. Moreover it is very flexible because it can take information from several sources simultaneously into consideration.