This author interview is by Dr David Johnson, of Imperial College London.  Dr Johnson's full paper, Semantically Linking In Silico Cancer Models, is available for download in Cancer Informatics.

Please summarise for readers the content of your article.

Our paper describes a new approach to thinking about the informatics behind cancer models. We think of computational models as any other data, which can be linked to metadata. We use connected property graphs as a means to representing data in such a way that meaningful questions can be asked of a database containing many cancer models. Models can be composed into larger models, even where different research groups have developed them. What our property-graph approach enables is a way of exploring possible model combinations based on semantic links, primarily through computaitonal interfaces, where compatibilities may be reasoned based on common units, computaional types, and biological concepts. The actual composition of models would still require a lot of rework and validation. In our paper we show two examples of describing cancer models as property graphs: one based on an EGFR-ERK pathway module, the other on a decomposed vascular tumour growth modelling framework.

How did you come to be involved in your area of study?

I have worked on life science and biological informatics since beginning my postdoctoral research, initially working on developing software for phylogenetics (the study of evolutionary relationships among populations and species) research in Professor Mark Pagel's Evolutionary Biology Group at the Univeresity of Reading, UK. Following this, I was appointed to the University of Oxford's Computing Laboratory (now called the Department of Computer Science) to work on the ‘Transatlantic Tumor Model Repositories' project developing interoperable databases between the US and Europe. This body of work eventually led to the work described in our contributions in this paper.

What was previously known about the topic of your article?

The idea of using linked data is not novel - in fact in biology in general, relating data and models to domain knowledge is commonplace, where there is a mature ecosytem of biological standards and ontologies to manage the diversity of data in biomedical sciences. There are established Web standards, such as RDF, OWL and SPARQL, are used to link datasets together. However these technologies primary purpose is to enable interoperation across distributed systems on the Web. Our approach is to link data in the first instance, within a cancer model database.

How has your work in this area advanced understanding of the topic?

Biological data management is creating new use cases for ‘NoSQL' databases including those that adopt the property graph data model, such as Neo4j that we used in our study. What we have proven is that linking model data, in particular model interfaces (input and output parameters) has the potential to create a system by which we can explore cancer model compositions. The challenge now is to use property graphs to build up a comprehensive resource of cancer models linked to metadata, to hopefully have an environment by which we can discover new candidate model compositions that could steer multiscale model research.

What do you regard as being the most important aspect of the results reported in the article?

Linking cancer models to metadata is a first step to producing an ecosystem of models and data. What we envisage is that we could link datasets into our property graphs, where not only model compositions could be inferred, but data-model combinations that may have previously not been identified could be found. Our focus was very much on the data management aspect of cancer informatics that we hope will in-turn help further the science in cancer research.

If you would like to include a link to a departmental webpage, LinkedIn profile, or other webpage where readers can learn more about your work paste it below:

There is a live demo of some of the work presented in the paper available here: http://gist.neo4j.org/?6038a7b526bfa48da2c0

LinkedIn: http://uk.linkedin.com/in/drdavidjohnson

Twitter: @NuDataScientist

ORCiD: http://orcid.org/0000-0002-2323-6847

share on