The internet is not only a platform for publishing documents; it is a provider of data and services. Incr easingly, scientific disciplines are exposing their tools and data to the internet, as a result, some scientific problems have become essentially internet mining problems. We show that candidate gene prioritisation, a challenging problem in biology, is essentially an internet mining problem. Thus, improving our ability to mine Future Internet Knowledge Bases (FIKBs) will advance biology and other sciences.
The internet has been, and still is, prima rily concerned with publishing documents. However, it is clea rly a lso a provider of data and services: scientific data is increasingly accessible on the internet, and many scientific tools are made available via the web, as web services, web applications, or otherwise e xposed to the internet.
Th is is particularly evident in the Life Sc ience doma in, wh ich has embraced
theinternet as a mediu m for publishing data and tools. To cite a few e xa mples, for
mo lecular data, the journal Nucleic Acids Research has tracked 1,230 databases [
In light of the “internetisation” of biological data and resources, we assert that many biologica l problems are de facto internet mining problems, analogous to more conventional internet mining proble ms. Therefore imp roving our ability to mine Future Internet Knowledge Bases (FIKBs) will certain ly advance biology and other sciences. We demonstrate this by showing how the problem of gene prioritisation is analogous to automatic page tagging.
Finding causes that influence particular traits is an impo rtant challenge in bio logy;
whether it is locating disease genes affecting humans, factors decreasing food
production for cereals, or factors increasing industrial insulin production,
fundamentally, the goal is the same, to find causes of biological traits. Often the
causes under study are genetic actors, and the methods employed to exa mine them
invariably rely on dra wing para lle ls against the body of studied genes; that is to say,
given some new gene of study, the assumption is always that it works in a simila r way
to closely evolutionarily re lated genes [
Th is assertion underpins the choice to focus study on model organisms, usually
organisms which lend themselves to study (i.e. by virtue of having easily observed
characteristics or by being cheap to work with) which are representative of their
respective classes [
Using observations and statistical techniques, associations can be drawn between
comple x tra its and genomic regions, however, these regions can be large [
When analysing genes, whilst some useful knowledge may be gleaned from
analysing their sequences directly [
The pattern of the problem presented earlier is not unique to the biological domain. As an exa mple, the same pattern can also be found in an automatic page tagging system that works by comparing untagged pages against an existing corpus, where pages may be associated by relationships such as: “belonging to the same domain”, “being written in the same language”, or “belonging to the same web ring”. Furthermore, pages can be related by the simila rity of their structure, by shared keywords, by a shared audience of the pages, and by other page comparison methods. These relationships may be more or less informat ive, but will induce a semantically heterogeneous graph.
Suppose that the pages in the existing corpus have been assigned appropriate tags by curation, these tags may be from a controlled vocabulary, ontology, or free (which could still fall into a structure such as WordNet [13]). An automat ic tagging method then might be, for each untagged page, to determine the strength of association between that page and the existing tags, and then assign tags according to the strength of association (with some sensible threshold).
Then tag based search (by single tags or by collection of tags) of the new, automatically tagged pages would return an ordered list of those pages most associated to those tags, this order can utilise the semantic distances between tags, and ma kes the proble m analogous to the gene prioritisation proble m. An e xa mp le of a graph that this view of the tagging proble m may induce is shown in Figure 1. Fig. 1. Two e xa mp le graphs, candidate gene prioritizat ion based on studied genes, and automatic page tagging based on curated pages sharing the same graph topology. Bold type represent to the “gene version” of this graph topology, whereas regular type represents the “page version” of this graph topology. Where edge/node types are the same in both cases, they are italic ised.
In the previous section, we have shown how a typical bioinformatics problem, gene prioritisation, is analogous to a typical internet mining proble m. Beyond this analogy, this and other bioinformatics studies should also be considered internet mining problems in the own right.
Each of the node types shown in the gene priorit isation exa mple in Figure 1 is represented by one or more internet resources, as are the edge types. For each of the node types shown in Figure 1, one source of this type of data is given in Table 1. For each of the edge types shown in Figure 1, one source of data for this type of relationship, or one tool for asserting this type of relationship, is shown in Table 2. Biologica l entities and relationships are encoded in a variety of forms, as documents, in structured data, and combinations of both, for our purposes, we only wish to illustrate that at least one form is (and in general, many forms are) available on the internet.
Thus, heterogeneous graphs that can be used for solving the candidate gene prioritisation proble m are d irect ly availab le on the internet, and along with other scientific resources, will be part of Future Internet Knowledge Bases (FIKBs).
Genes
Ensemb l [14] URL http://www.ensemb l.org/inde x.ht ml
[15] Gene Ontology [12] http://www.genome .jp/kegg/pathway. html http://www.geneontology.org/ In conclusion, with the greater availab ility of scientific resources on the internet, tasks in mining scientific data will increasingly become internet min ing problems. Scientific research will increasingly rely on the design and availability of dedicated Future Internet Knowledge Bases (FIKBs), and on the development of associated methods to analyse them.
This brings with it, both new challenges and new opportunities. Whilst we have illustrated our case with a problem in the biological domain, the principles hold more widely for other sciences. Some scientific problems have parallels amongst existing internet mining p roblems, and it is reasonable to expect that advances in techniques in mining the future internet will provide solutions to scientific problems, and vice versa.
The authors would like to thank Julia Halder fo r valuable comments, the authors also gratefully acknowledge funding fro m the UK Biotechnology and Biological Sciences Research Council (DTG BB/F529038/1, SA BR Grant BB/F006039/1). 9. Altschul, S.F., Gish, W., Miller, W., Myers, E.W., Lip man, D.J.: Basic local align ment search tool. J Mol Bio l 215, 403-410 (1990) 10. Dale, J.M., Popescu, L., Karp, P.D.: Mach ine learning methods for metabolic pathway prediction. Bmc Bio informat ics 11, 15 (2010) 11. Krallinger, M., Valencia, A., Hirschman, L.: Linking genes to literature: text mining, informat ion extract ion, and retrieval applications for biology. Genome Biol 9 Suppl 2, S8 (2008) 12. Ashburner, M., Ball, C.A., Blake, J.A., Botstein, D., Butler, H., Cherry, J.M., Davis, A.P., Dolinski, K., Dwight, S.S., Eppig, J.T., Harris, M.A., Hill, D.P., IsselTarver, L., Kasarskis, A., Lewis, S., Matese, J.C., Richardson, J.E., Ringwald, M., Rubin, G.M., Sherlock, G.: Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 25, 25-29 (2000) 13. Sig man, M., Cecchi, G.A.: Global organization of the Wordnet lexicon. Proc Natl Acad Sci U S A 99, 1742-1747 (2002) 14. Hubbard, T., Barker, D., Birney, E., Cameron, G., Chen, Y., Clark, L., Co x, T., Cuff, J., Curwen, V., Down, T., Durbin, R., Eyras, E., Gilbert, J., Hammond, M., Hu min iecki, L., Kasprzyk, A., Lehvaslaiho, H., Lijnzaad, P., Melsopp, C., Mongin, E., Pettett, R., Pocock, M., Potter, S., Rust, A., Sch midt, E., Searle, S., Slater, G., Smith, J., Spooner, W., Stabenau, A., Stalker, J., Stupka, E., Ureta-Vidal, A., Vastrik, I., Clamp , M.: The Ensembl genome database project. Nucleic Acids Res 30, 38-41 (2002) 15. Kanehisa, M., Goto, S., Hattori, M ., Aoki-Kinoshita, K.F., Itoh, M., Kawashima, S., Katayama, T., Araki, M., Hirakawa, M.: Fro m genomics to chemical genomics: new developments in KEGG. Nucleic Acids Res 34, D354-357 (2006) 16. Johnson, M., Zaretskaya, I., Raytselis, Y., Merezhuk, Y., McGinn is, S., Madden, T.L.: NCBI BLAST: a better web interface. Nucleic Acids Res 36, W5-9 (2008) 17. Barrell, D., Dimmer, E., Huntley, R.P., Binns, D., O'Donovan, C., Apweiler, R.: The GOA database in 2009--an integrated Gene Ontology Annotation resource. Nucleic Acids Res 37, D396-403 (2009) 18. Go ffard, N., Weiller, G.: PathExpress: a web -based tool to identify relevant pathways in gene expression data. Nucleic Acids Res 35, W176-181 (2007) 19. He, M., Wang, Y., Li, W.: PPI finder: a mining tool for human proteinprotein interactions. Plos One 4, e4554 (2009)