Vous êtes ici : GIPSA-lab > Animation > Séminaires

Séminaire du département Images et Signal du 08/11/2013 à 10h00

 

Structure-revealing Data Fusion Model with Applications in Metabolomics

Intervenant : Evrim Acar,

Lieu : Chartreuse

 

Résumé :

Analysis of data from multiple sources has the potential to enhance knowledge discovery by capturing underlying structures, which are, otherwise, difficult to extract. For instance, in metabolomics, biological fluids, such as blood or urine, are often measured using different analytical techniques in order to identify the chemicals related to certain diseases. Data from different analytical techniques, e.g., LC-MS (Liquid Chromatography - Mass Spectrometry) and NMR (Nuclear Magnetic Resonance) spectroscopy, provide complementary data sets and their joint analysis may enable us to capture a larger proportion of the complete metabolome belonging to a specific biological system.

Fusing data from multiple sources has already proved useful in many applications in social network analysis, signal processing and bioinformatics. However, data fusion is a challenging task since data from multiple sources (i) have both shared and unshared components, and are often (ii) heterogeneous (i.e., in the form of higher-order tensors and matrices), (iii) incomplete, and (iv) high dimensional containing abundance of irrelevant features, which makes it difficult to extract interpretable patterns. In order to address these challenges, we formulate data fusion as a coupled matrix and tensor factorization problem. Through modeling constraints, the model can automatically reveal shared and unshared components, that are easily-interpretable. We demonstrate the effectiveness of our approach with applications from metabolomics.


GIPSA-lab, 11 rue des Mathématiques, Grenoble Campus BP46, F-38402 SAINT MARTIN D'HERES CEDEX - 33 (0)4 76 82 71 31