Bioinformatics

Research in the Bioinformatics group combines interests in data and software integration with systems biology. Our approach is to use problems, at present taken mainly from research into crop improvement and plant pathogen interactions, to motivate the development of new bioinformatics methods that address current biological questions. In particular we focus on the use of a graph based approach to integrate biological databases, ontologies and experimental data to create a data warehouse. This warehouse will first be used to facilitate the analysis and interpretation of large scale transcriptomics experiments. The approach, however, is completely general and could equally be applied to integrated analysis of other post-genomic datasets such as those arising from proteomics and metabolomics.

The availability of a well-structured and generic biological information resource is also the basis for other projects in systems biology within the group, which include: analysis/prediction of transcription factors and their interactions, the representation, modelling and simulation of the dynamic behaviour of biological networks at the molecular and genomic level.

Our research is highly collaborative in nature and the group has built strong links with other research programmes in Rothamsted Research. We also benefit by working closely with the Bioinformatics consultants in the group who are in a position to exploit new methods as they appear and influence their development for maximum benefit of other scientists.

Research Activities

Databases and data integration

An important challenge for bioinformatics research and infrastructure groups is to provide integrated access to a wide variety of different types of data from many different species. Therefore, database development and data integration is one of the core research activities of the Bioinformatics group.

• The DPV (Descriptions of Plant Viruses) and PHI-base (Pathogen Host Interactome database) are examples of joint research activities with the PPM Department.
• A Wheat EST database of 20,000 ESTs, based on the original database developed at University of Bristol for the wheat IGF cDNA microarray , has been extended to include gene ontology terms, functional classification and locus information from sequence similarity.A wheat gene database is being developed to collect assemblies of all public-domain wheat ESTs that will store inferred function and links to the closest genomic sequence for locus information.
• The metabolites database (MetDB) has been developed to assemble publicly available data of known reactions and compounds into a highly searchable collection to aid the determination of biosynthetic pathways.

Database integration has the potential to provide deeper insights into plant systems by linking and integrating complimentary information from different data sources. ONDEX-integration is a system that is developed to overcome both technical and semantic data heterogeneity using ontologies and generalised data structures.

Text mining

The origins of the graph based approach to data integration lie in the ontology based a text mining system (ONDEX). The extraction of information from unstructured text is an important component in our research and there are mutual benefits to developing these methods alongside data integration research. This is because a comprehensive vocabulary and semantics for biological concepts underpins both effective text mining and data integration. ONDEX is being used in several projects to extract relevant data from literature through concept based relationship mining.

Advanced analysis methods for post genomic data sets

Post genomics technologies such as microarrays, high-throughput NMR and high-throughput EST library sequencing generate large quantities of data on a daily basis. These technologies help biologists at RRes to identify and characterise genes, metabolites and gene products. The bioinformatics research group is developing methods to support these activities through analysis and interpretation. Currently the focus is on methods for microarray data and reflects the importance this technology has within the biology research groups at Rothamsted.

Our research on microarray analysis complements work on statistical analysis methods being developed in the Biostatistics Group.

Recognising that the actual methods used for quantifying microarray chips have a significant impact on all subsequent analysis steps we have investigated methods (print tip optimisation) for maximising signal to noise ratios. Statistical analyses of microarray data result in significantly differentially expressed or co-expressed gene sets that often consist of hundreds or even thousands of genes. However, several bioinformatics methods can be applied to help biologists interpret and analyse such lists. AutoMATRIX is a tool for transcription factor analysis that can be used to identify key transcription factors that may be causally implicated by the microarray results themselves. The front-end of the database integration system ONDEX is being developed to support analysis and visualisation of experimental results (microarrays, EST libraries, large-scale spectroscopy etc.) in the context of background knowledge on metabolic networks, gene regulatory networks and ontologies.

Modelling and simulation

Modelling and simulation of biological systems is used as a predictive method to investigate the potential impact of manipulations at the genetic or molecular level. Modelling itself can help to capture the knowledge of biological systems in an unambiguous way to obtain a comprehensive yet precise (within limits of our knowledge) and detailed overview of any given system.

An example of research we are undertaking is the development of a pathway model of the Gibberellin biosynthesis in collaboration with the group of Peter Hedden, PSC. Another example is the Carbon-Nitrate partitioning model which is being developed in collaboration with PSC and the Institut de Biotechnologie des Plantes, France. Further examples are the regulation of Nitrogen uptake from soil and the development of methods for the parameterisation of pathway models.

Most models are developed using Hybrid Petri Nets although other formalisms and tools to assist the modelling and simulation process are also being developed and applied.

Sequence analysis

We have developed a complex motif search pipeline to scan, sort and cluster sequences (ESTs, nucleotides, amino acid sequences etc) for given motifs., This is now being used by the bioinformatics consultation team to support ongoing research activities of several biological groups. Further, we developed transmembrane protein prediction of Fusarium proteins for initial use within the PPM Department.

Other high throughput sequence analysis methodologies (multiple alignments, homology searches etc.) are applied for several purposes such as homology based GO classification to support, for example, the Fusarium genome annotation. Methods are also being developed to make cross species comparisons of transcription factors and promoter regions.