HER Imaging and Molecular Interaction Mapping in Breast Cancer
FP7 imagint
 
  Partner: King's College London
  Institute for Mathematical and
  Molecular Biomedicine
Project Leader
Prof. ACC Coolen

Group members
Dr. Katherine Lawler
Mr. James Barrett

Introduction

The Institute for Mathematical and Molecular Biomedicine (IMMB) is devoted to the development and application of advanced mathematical and computational methods for biomedical problems. The main foci of research are the analysis of stochastic signalling processes in molecular and inter-cellular networks (e.g. proteome signalling, gene regulation, immune networks), and on the statistical analysis of complex biomedical data (including survival analysis, Bayesian data integration, and outcome prediction). The members of the IMMB work closely with bioinformaticians, basic biomedical scientists, and clinical scientists, at King’s College and elsewhere.


Role in the “Imagint Consortium”

The Imagint consortium will generate multiple qualitatively different data sets, ranging from protein co-localization signals and microRNA and protein concentrations, through to protein interaction imaging based tissue biomarker measurements and clinical PET data. The primary role of the KCL-IMMB team is to generate reliable computational tools for analysing and integrating such data, for determining whether they contain meaningful patterns (and extract and visualize these if they exist), and for quantifying the clinical potential of candidate biomarkers and signatures as predictors of disease progression or treatment response, or as disease classifiers. To do this the team develops Bayesian methods for combine multiple types of data (disparate both in terms of nature and dimensionality) and search for low dimensional structure within datasets (see Fig. 1).

Figure 1. Example of binary classification of breast tumour data using Latent Variable Gaussian Process Classification. Red areas correspond to a high probability of belonging to the 'bad outcome' class. Similarly, blue regions mean that the tumour is likely to belong to the 'good outcome' class.
In addition the team members work on Bayesian and information-theoretic tools for decontaminating protein interaction data for experimental bias (see Fig 2), and on the development of novel survival analysis and regression methods that can quantify reliably (unlike most conventional approaches such as Cox regression or Kaplan-Meier estimators) the links between (molecular and genomic) biomarkers and breast cancer risk in the presence of strong disease or cohort heterogeneity.

Figure 2. Illustration of the precision of combinatorial and information-theoretic tools in predicting the degree correlation functions (shown in heat-maps) of protein-protein interaction data collected via imperfect and biased sampling protocols. These generate false-negative and false-positive nodes and links, with statistical properties that depend on the local topology of the true underlying network. Left: degree-degree correlation function of a protein interaction network before sampling. Middle: theoretical prediction of degree-degree correlation function after sampling. Right: actual observed degree-degree correlations after sampling.

The secondary role of the KCL-IMMB team is to use advanced mathematical techniques from the theory of complex signalling networks to identify candidate protein biomarkers on the basis of the toponome and protein interaction (FRET-FLIM and nanoscopy) data generated within the Imagint consortium.
 


Selected Publications

1. Shayeghi N, Ng T and Coolen ACC. Direct response analysis in cellular signalling networks. J Theor Biol 2012, 304: 219225.

2. Roberts ES and Coolen ACC. Unbiased degree-preserving randomization of directed binary networks. Phys Rev E 2012, 85: 046103.

3. Coolen ACC, Fraternali F, Annibale A, Fernandes LP and Kleinjung J. Modelling biological networks via tailored random graphs. Handbook of Statistical Systems Biology (Wiley; M Stumpf, DJ Balding and M Girolami, Eds) 2011, 309-330.

4. Fruhwirth GO et al. How Foerster resonance energy transfer imaging improves the understanding of protein interaction networks in cancer biology. Chem Phys Chem 2011, 12: 442-461.

5. Carlin LM et al. A targeted siRNA screen identifies regulators of Cdc42 activity at the natural killer cell immunological synapse. Science Signaling 2011, 4: 1-11.
 


   Copyright © 2011-15 - Imagint