Affordable Access

Automatic and manual functional annotation in a distributed web service environment

Publication Date
  • Life Sciences
  • Biology


While the number of genomic sequences becoming available is increasing exponentially, most genes are not functionally well characterized. Finding out more about the function of a gene and about functional relationships between genes will be the next big bottleneck in the post-genomic era. On the one hand improved pipelines and tools are needed in this context, because running experiments for all predicted genes is not feasible. On the other hand manual curation of the automatic predictions is necessary to judge the reliability of the automatic annotation and to get a more comprehensive view on the function of each individual gene. For the automatic functional annotation often a homology based function transfer from functionally characterized genes is applied using methods like Blast. However, this approach has many drawbacks and makes systematic errors by not taking care of speciation and duplication events. Phylogenomics has shown to improve the functional prediction accuracy by taking the evolutionary history of genes in a phylogenetic tree context into account. In this thesis the manual process from the assembly of the DNA sequence to the functional characterization of genes and the identification and comparison of shared syntenic regions, including the identification of candidate genes for pathogen resistance in potato chromosome V, is explained and problems discussed. To improve the automatic functional annotation in genome projects, a phylogenomic pipeline, which includes SIFTER one of the best phylogenomic tools in this area, is introduced, improved and tested in the Medicago truncatula, Sorghum bicolor and Solanum lycopersicum genome projects. To obtain new candidate genes for the development of new drugs and crop protection products, non-plant specific genes, like the transferrin family which is not known in plants yet, are extracted from the M. truncatula and S. bicolor genomes and further investigated. For further improvement of the annotation, a new phylogenomic approach is developed. This approach makes use of annotated functional attributes to calculate the functional mutation rate between genes and groups of genes in a phylogenetic tree and to find out if the function of a gene can be transferred or not. The new approach is integrated into the SIFTER tool and tested on the blue-light photoreceptor/photolyase family and on a test set of manually curated Arabidopsis thaliana genes. Using both test sets the prediction accuracy could be significantly improved and a more comprehensive view on the gene function could be obtained. But because still no tool is able to annotate all functions of a gene with 100% accuracy, I introduce a system for manual functional annotation, called AFAWE. AFAWE runs different web services for the functional annotation and displays the results and intermediate results in a comprehensive web interface that facilitates comparison. It can be used for any organism and any kind of gene. The inputs are the amino acid sequence and the corresponding organism. Because of its flexible structure, new web services and workflows can be easily integrated. Besides Blast searches against different databases and protein domain prediction tools, AFAWE also includes the phylogenomic pipeline. Different filters help to identify trustworthy results from each analysis. Furthermore a detailed manual annotation can be assigned to each protein, which will be used to update the functional annotation in public databases like MIPSPlantsDB.

There are no comments yet on this publication. Be the first to share your thoughts.