AUGUSTUS is a software tool for gene prediction in eukaryotes, initially developed as an ab initio method based on a Generalized Hidden Markov Model (GHMM). The original version of AUGUSTUS predicted only one transcript per gene and did not account for alternative splicing. However, an extended version of AUGUSTUS has been developed to predict multiple splice variants, making it the first ab initio gene finder capable of predicting multiple transcripts. This new version also includes a motif searching facility where user-defined regular expressions can be searched against predicted proteins. The AUGUSTUS web interface and downloadable open-source program are freely available at http://augustus.gobics.de. The introduction highlights the limitations of existing gene prediction tools and the importance of handling alternative splicing, which affects 40–60% of human genes. The new version of AUGUSTUS addresses this by allowing users to control the number of predicted splice variants per gene, balancing sensitivity and specificity. The materials and methods section explains the sampling and posterior probability estimation techniques used to predict alternative transcripts. The web server description details the input and output formats, including the ability to search for regular expressions to analyze putative protein products. The results section demonstrates the accuracy of AUGUSTUS on a large test dataset from the EGASP workshop, showing that the sensitivity increases with the number of predicted transcripts, while specificity decreases. The gene-level sensitivity improves significantly when using the 'many transcripts' option, making it particularly useful for finding at least one correct splice form per gene. The acknowledgments section credits the development of the multiple transcript prediction option to the Tobacco Genome Initiative and acknowledges funding sources.AUGUSTUS is a software tool for gene prediction in eukaryotes, initially developed as an ab initio method based on a Generalized Hidden Markov Model (GHMM). The original version of AUGUSTUS predicted only one transcript per gene and did not account for alternative splicing. However, an extended version of AUGUSTUS has been developed to predict multiple splice variants, making it the first ab initio gene finder capable of predicting multiple transcripts. This new version also includes a motif searching facility where user-defined regular expressions can be searched against predicted proteins. The AUGUSTUS web interface and downloadable open-source program are freely available at http://augustus.gobics.de. The introduction highlights the limitations of existing gene prediction tools and the importance of handling alternative splicing, which affects 40–60% of human genes. The new version of AUGUSTUS addresses this by allowing users to control the number of predicted splice variants per gene, balancing sensitivity and specificity. The materials and methods section explains the sampling and posterior probability estimation techniques used to predict alternative transcripts. The web server description details the input and output formats, including the ability to search for regular expressions to analyze putative protein products. The results section demonstrates the accuracy of AUGUSTUS on a large test dataset from the EGASP workshop, showing that the sensitivity increases with the number of predicted transcripts, while specificity decreases. The gene-level sensitivity improves significantly when using the 'many transcripts' option, making it particularly useful for finding at least one correct splice form per gene. The acknowledgments section credits the development of the multiple transcript prediction option to the Tobacco Genome Initiative and acknowledges funding sources.