Date(s) - 30/11/2012
3:00 pm - 4:00 pm
SGDP seminar room
Dr. Matt Davies – “Machine Learning Approaches to the Metabolome”
Human metabolic variation is distributed along a continuum reflecting the complex interplay of genetic and environmental factors. Technologies that allow the collection of highly dimensional datasets in standardized fashion are offering new promise for the discovery of disease biomarkers and for a better understanding of the role of genetic predisposition to complex chronic diseases. Metabolomics platforms measure small molecule metabolites from body fluids and tissues using untargeted approaches based on mass spectrometry (MS) and nuclear magnetic resonance (NMR). One limitation of the technique, however, is that the metabolome is not as well characterised as the human genome and the fact that many metabolites of interest happen to be present at low concentrations and suffer from a low signal to noise ratio. We are currently applying machine learning techniques to metabolomic data to identify profiles linked to various phenotypes. The tutorial will cover the use of the machine learning program WEKA and consider whether machine learning techniques may be preferable to standard statistical techniques in the analysis of this data.
Dr. James Swingland – “1-D Spatial modelling of gene expression data”
Microarrays represent a powerful technique for simultaneously investigating thousands of genes. However, lists of differentially expressed genes often do little to improve the understanding of disease mechanisms or the underlying biology. Despite theoretical and experimental reasons to think that spatial location is an important variable in gene expression, few modelling approaches include it. The ‘Chromowave’ model analyses data by first transforming it into wavelet space. This specifically highlights coherent expression of genes along chromosomes, providing fundamentally different results to analyses performed directly on probes. The approach can also be applied to other spatially organised data such as methylation data. This talk aims to justify the importance of spatial location in gene expression and the reasons for using a model such as Chromowave. Examples of Chromowave analysis in a variety of situations will also be provided to help better understand the method.
After his talk, Dr. Swingland will give a live demonstration of the Chromowave software. To attend this, simply stay in the seminar room after the main seminar.