AUGUSTUS: ab initio prediction of alternative transcripts

Abstract

AUGUSTUS is a software tool for gene prediction in eukaryotes based on a Generalized Hidden Markov Model, a probabilistic model of a sequence and its gene structure. Like most existing gene finders, the first version of AUGUSTUS returned one transcript per predicted gene and ignored the phenomenon of alternative splicing. Herein, we present a WWW server for an extended version of AUGUSTUS that is able to predict multiple splice variants. To our knowledge, this is the first ab initio gene finder that can predict multiple transcripts. In addition, we offer a motif searching facility, where user-defined regular expressions can be searched against putative proteins encoded by the predicted genes. The AUGUSTUS web interface and the downloadable open-source stand-alone program are freely available from .

Draft genome sequence of pigeonpea (Cajanus cajan), an orphan legume crop of resource-poor farmers

“…For de novo prediction, Augustus 43 , GENSCAN 44 and GlimmerHMM 45 were used to predict genes with parameters trained on A. thaliana. We merged three de novo predictions into a unigene set.…”

“…De novo gene models that were supported by two or more de novo methods were retained. For overlapping gene models, the longest one was selected and finally, we got de novo-based gene models (43,647).…”

Simultaneous gene finding in multiple genomes

“…Both candidate exons and introns are obtained within AUGUSTUS by random sampling of gene structures from the posterior distribution defined by a semi-Markov conditional random field. The sampling of gene structures in AUGUSTUS has previously been introduced to identify alternative transcripts (Stanke et al, 2006). In general, sampling yields just the most likely splicing variants, which do not sufficiently represent the space of all possible gene structures.…”

Predicting Gene Structures from Multiple RT-PCR Tests

“…To obtain the splicing graph, we have used Augustus gene finder (Stanke et al, 2006) capable of sampling random gene structures from the posterior probability distribution defined by the underlying generalized HMM. For each locus, we have created the splicing graph based on 1000 samples.…”

“…These exons can be joined to potential transcript based on compatibility of their reading frames. Gene finder Augustus (Stanke et al, 2006) uses an HMM path sampling algorithm to output multiple predictions that may correspond to alternative transcripts of the same gene. These methods can also be generalized to take into account additional information, such as ESTs or protein similarity (Stanke et al, 2008).…”