From a Biomedical viewpoint: Biomarkers, High Density Microarray Data Analysis, High Throughput Data Analysis
Nowadays, medical research may require the insight of computer scientists and statisticians to mine useful knowledge from large biological or medical databases. Our research deals with the analysis of high-throughput data such as gene expression data coming from microarrays experiments. Microarrays are measuring gene expressions at the level of messenger RNA to infer knowledge about genes function.
We develop machine learning algorithms to predict the medical outcome of a treatment or, more generally, to help clinicians to set a clinical diagnosis or prognosis. Learning algorithms build predictive models from observed data. A particularly important result of such a modeling process is a list of genes, also known as reporters or biomarkers, which are predicted to matter in the pathologies or treatments under study.
Constructing a biomarker list is, in machine learning terms, a special case of feature selection. It aims at reducing the dimensionality of the prediction problem without losing useful information and, possibly, while filtering out noise in the data. The question on how to estimate the quality of the predictive models and the statistical significance of the reporter list is also addressed.
In many cases, the constructed gene list needs to be further investigated to assess its relevance from a medical or biological point of view. The final biomedical outcome is typically the design of a prognosis or diagnosis kit to be used routinely in hospitals. Concrete case studies are described here for several ongoing projects.
From a Biomedical viewpoint: Biomarkers, Microarray Data, Allergy, Pediatry
From a Biomedical viewpoint: Biomarkers, High Density Microarray Data, RT-PCR, Inflammation, Arthritis
From a Biomedical viewpoint: Biomarkers, High Density Microarray Data Analysis, Hypoxia, Cancer Metastasis
Pierre Dupont Last modified: Tue Oct 1 20:13:04 CEST 2013