Novel approaches to develop community-built biological network models for potential drug discovery.

INTRODUCTION: Hundreds of thousands of data points are now routinely generated in clinical trials by molecular profiling and NGS technologies. A true translation of this data into knowledge is not possible without analysis and interpretation in a well-defined biology context. Currently, there are many public and commercial pathway tools and network models that can facilitate such analysis. At the same time, insights and knowledge that can be gained is highly dependent on the underlying biological content of these resources. Crowdsourcing can be employed to guarantee the accuracy and transparency of the biological content underlining the tools used to interpret rich molecular data. Areas covered: In this review, the authors describe crowdsourcing in drug discovery. The focal point is the efforts that have successfully used the crowdsourcing approach to verify and augment pathway tools and biological network models. Technologies that enable the building of biological networks with the community are also described. Expert opinion: A crowd of experts can be leveraged for the entire development process of biological network models, from ontologies to the evaluation of their mechanistic completeness. The ultimate goal is to facilitate biomarker discovery and personalized medicine by mechanistically explaining patients' differences with respect to disease prevention, diagnosis, and therapy outcome.

Information needs and the role of text mining in drug development.

Drug development generates information needs from groups throughout a company. Knowing where to look for high-quality information is essential for minimizing costs and remaining competitive. Using 1131 research requests that came to our library between 2001 and 2007, we show that drugs, diseases, and genes/proteins are the most frequently searched subjects, and journal articles, patents, and competitive intelligence literature are the most frequently consulted textual resources.

Suregene, a scalable system for automated term disambiguation of gene and protein names.

Researchers, hindered by a lack of standard gene and protein-naming conventions, endure long, sometimes fruitless, literature searches. A system that is able to automatically assign gene names to their LocusLink ID (LLID) in previously unseen MEDLINE abstracts is described. The system is based on supervised learning and builds a model for each LLID. The training sets for all LLIDs are extracted automatically from MEDLINE references in the LocusLink and SwissProt databases. A validation was done of the performance for all 20,546 human genes with LLIDs. Of these, 7344 produced good quality models (F-measure >0.7, nearly 60% of which were >0.9) and 13,202 did not, mainly due to insufficient numbers of known document references. A hand validation of MEDLINE documents for a set of 66 genes agreed well with the system's internal accuracy assessment. It is concluded that it is possible to achieve high quality gene disambiguation using scaleable automated techniques.

Advances in text analytics for drug discovery.

The automated extraction of biological and chemical information has improved over the past year, with advances in access to content, entity extraction of genes, chemicals, kinetic data and relationships, and algorithms for generating and testing hypotheses. As the systems for reading and understanding scientific literature grow more powerful, so must the infrastructure in which to assemble information. Advances in infrastructure systems are discussed in this review. Research efforts have flourished as a result of text analytics competitions that attract participants from various disciplines, from computer science to bioinformatics.

AZuRE, a scalable system for automated term disambiguation of gene and protein names.

Researchers, hindered by a lack of standard gene and protein-naming conventions, endure long, sometimes fruitless, literature searches. A system is described which is able to automatically assign gene names to their LocusLink ID (LLID) in previously unseen MEDLINE abstracts. The system is based on supervised learning and builds a model for each LLID. The training sets for all LLIDs are extracted automatically from MEDLINE references in the LocusLink and SwissProt databases. A validation was done of the performance for all 20,546 human genes with LLIDs. Of these, 7,344 produced good quality models (F-measure > 0.7, nearly 60% of which were > 0.9) and 13,202 did not, mainly due to insufficient numbers of known document references. A hand validation of MEDLINE documents for a set of 66 genes agreed well with the system's internal accuracy assessment. It is concluded that it is possible to achieve high quality gene disambiguation using scaleable automated techniques.