Mélanie Courtine

Serum amyloid A: production by human white adipocyte and regulation by obesity and nutrition

Aims/hypothesisThe acute-phase proteins, serum amyloid As (SAA), are precursors of amyloid A, involved in the pathogenesis of AA amyloidosis. This work started with the characterisation of systemic AA amyloidosis concurrent with SAA overexpression in the subcutaneous white adipose tissue (sWAT) of an obese patient with a leptin receptor deficiency. In the present study a series of histopathological, cellular and gene expression studies was performed to assess the importance of SAA in common obesity and its possible production by mature adipocytes.Materials and methodsGene expression profiling was performed in the sWAT of two extremely obese patients with a leptin receptor deficiency. Levels of the mRNAs of the different SAA isoforms were quantified in sWAT cellular fractions from lean subjects and from obese subjects before and after a very-low-calorie diet. These values were subsequently compared with serum levels of SAA in these individuals. In addition, histopathological analyses of sWAT were performed in lean and obese subjects.ResultsIn sWAT, the expression of SAA is more than 20-fold higher in mature adipocytes than in the cells of the stroma vascular fraction (p<0.01). Levels of SAA mRNA expression and circulating levels of the protein are sixfold (p<0.001) and 3.5-fold (p<0.01) higher in obese subjects than in lean subjects, respectively. In lean subjects, 5% of adipocytes are immunoreactive for SAA, whereas the corresponding value is greater than 20% in obese subjects. Caloric restriction results in decreases of 45–75% in levels of the transcripts for the SAA isoforms and in circulating levels of the protein.Conclusions/interpretationThe results of the present study indicate that SAA is expressed by sWAT, and its production at this site is regulated by nutritional status. If amyloidosis is seen in the context of obesity, it is possible that production of SAA by adipocytes could be a contributory factor.

Improving classification of microarray data using prototype-based feature selection

This paper addresses the problem of improving accuracy in the machine-learning task of classification from microarray data. One of the known issues specifically related to microarray data is the large number of inputs (genes) versus the small number of available samples (conditions). A promising direction of research to decrease the generalization error of classification algorithms is to perform gene selection so as to identify those genes which are potentially most relevant for the classification. Classical feature selection methods are based on direct statistical methods. We present a reduction algorithm based on the notion of prototypegene. Each prototype represents a set of similar gene according to a given clustering method. We present experimental evidence of the usefulness of combining prototype-based feature selection with statistical gene selection methods for the task of classifying adenocarcinoma from gene expressions.

Abstractions for Knowledge Organization of Relational Descriptions

The goal of conceptual clustering is to construct a hierarchy of concepts which cluster objects based on their similarities. Knowledge organization aims at generating the set of maximally specific concepts for all possible classifications: the Generalization Space. Our research focuses on the organization of relational data represented using conceptual graphs. Unfortunately, the generalization of relational descriptions necessary to build the Generalization Space leads to a combinatorial explosion. This paper proposes to incrementally introduce the relations by using a sequence of languages that are more and more expressive. The algorithm proposed, called KIDS, is based upon an iterative reformulation of the objects descriptions. Initially represented as conceptual graphs, they are reformulated into abstract objects represented as 〈attribute, value〉 pairs. This representation allows us to use an efficient propositional knowledge organization algorithm. Experiments on Chinese character databases show the interest of using KIDS to build organizations of relational concepts.

KIDS: An Iterative Algorithm to Organize Relational Knowledge

The goal of conceptual clustering is to build a set of embedded classes, which cluster objects based on their similarities. Knowledge organization aims at generating the set of most specific classes: the Generalization Space. It has applications in the field of data mining, knowledge indexation or knowledge acquisition. Efficient algorithms have been proposed for data described in 〈attribute, value〉 pairs formalism and for taking into account domain knowledge. Our research focuses on the organization of relational knowledge represented using conceptual graphs. In order to avoid the combinatorial explosion due to the relations in the building of the Generalization Space, we progressively introduce the complexity of the relations. The KIDS algorithm is based upon an iterative data reformulation which allows us to use an efficient propositional knowledge organization algorithm. Experiments show that the KIDS algorithm builds an organization of relational concepts but remains with a complexity that grows linearly with the number of considered objects.

Propositionalization for clustering symbolic relational descriptions

Propositionalization has recently received much attention in the ILP community as a mean to learn efficiently non-determinate concepts using adapted propositional algorithms. This paper proposes to extend such an approach to unsupervised learning from symbolic relational description. To help deal with the known combinatorial explosion of the number of possible clusters and the size of their descriptions, we suggest an approach that gradually increases the expressivity of the relational language used to describe the classes. At each level, only the initial object descriptions that could benefit from such an enriched generalization language are propositionalized. This latter representation allows us to use an efficient propositional clustering algorithm. This approach is implemented in the CAC system. Experiments on a large Chinese character database show the interest of using KIDS to cluster relational descriptions and pinpoint current problems for analyzing relational classifications.

Biomarkers Discovery in Medical Genomics Data

This chapter presents a system, called DiscoCini, assisting the biology experts to explore medical genomics data. First, it computes all the correlations (based on ranks) between gene expression and bioclinical data. The amount of generated results is huge. In a second step, we propose an original visual approach to simply and efficiently explore these results. Thanks to sets of data generated during experiments in the field of the obesities genomics, we show how DiscoClini allows easily identification of complex disease biomarkers.