270 likes | 399 Views
http://www.dgemap.org/. Jano van Hemert. 3 years EU-funded design study.
E N D
http://www.dgemap.org/ Jano van Hemert
3 years EU-funded design study Goal: design theorganisational & collaborative structures, ethical framework, and molecular genetic & informatics technologies necessary for a new research infrastructure which will accelerate an integrated European approach to gene expression in early human development
National e-Science Centre MRC Human Genetics Unit University of Newcastle Consortium
Three major goals • Facilitate collaboration over multiple laboratories • Improve ways for handling spatial-temporal data from gene expression studies • Provideintegrationwithother technologies and databases to help biologists advance their studies
Framework: Edinburgh Mouse Atlas Space and Anatomy
Space and Anatomy anatomical name
Gene Expression Database • Query: by both space and text...
Data mining Human-mouse link Other data sources (OMIM, GDX, …) Visualisation Silicon processes
Where do workflows fit in? • Advanced queries incorporating other DBs • Linking genes with diseases (OMIM) • Genetic pathways (Kegg) • Mouse-human interoperability • Using anatomical terms • Using direct 3D to 3D model mapping • Using spatial-temporal ontologies • Data mining and processes • Hierarchical Clustering • Association rules
Hierarchical clustering ‘McMahon’ Data TS17
Hierarchical clustering ‘McMahon’ Data TS17 Myt1l Dlx5
What are association rules? • Based on a set of transactions • We want to derive rules of the form X => Y • Meaning, if X happens then Y happens • X and • X and Y are sets of items appearing in the transactions • The rules come with numbers to express their quality with respect to the set of transactions (most common: support and confidence)
Association Rules • In the context of gene expression: if Gene1 and Gene2 then Gene3 where a transaction equals a set of genes expressing together at the same time in the same anatomical component • Alternative: if Component1 then Component2 and Component3where a transaction equals a number of components expressing the same gene at the same time
Association Rules Results Transaction: genes expressing in the same anatomical component in the same Theiler stage Association rules with a minimum confidence of 90% Wnt1, Bmp4 => Shh 0.053 0.91 Vcam1 => Kdr 0.057 0.93 Emx2 => Otx2 0.054 0.95 Otx1, Pax6 => Otx2 0.051 0.92 Techo-fact: extracted using web services called from a Perl script… Source: the EMAGE database, using the editorial spatial annotations extracted on 2006/08/28
Main issues while using Taverna • Need for more data mangling functions • Need for more data formatting controls • Pipelining and memory concerns • Library of useful translations services • Interaction Plug-in Architecture…? • What about Axis version 2?
Thanks for your attention Susan Lindsay Demetrius Vouyiouklis Marie-Laure Muiras Xunxian Wang Mark Scott Alina Andras Malcolm Atkinson Jano van Hemert Yin Chen Richard Baldock Simon Woods Ken Taylor