Approche de la complexité du syndrome métabolique et de ses indicateurs de risque par la mise en oeuvre de méthodes numériques et symboliques de fouille de données

Abstract : Metabolic syndrome (MS) is a complex concept which is characterized by a constellation of cardiovascular risk factors. This cluster results from gene x environment interactions. The aim of this PhD thesis was to study the metabolic syndrome among an apparently healthy family population, namely the STANISLAS cohort. In order to approach the complexity of MS, we used statistics and data mining techniques, either separately or in combination. We developed a methodology for mining cohorts based on frequent itemsets and association mIes. It enables the expert of the domain to generate new research hypotheses validated by statistical tests or new lab experiments. Besides, given that the expert is interested in knowledge about MS, a poorly frequent state in the supposedly healthy subjects of the STANISLAS cohort, we have proposed a new algorithm which searches rare itemsets. Therefore, the major results of this thesis were obtained by performing epidemiological studies and/or data mining experiments, by considering the MS entity or by treating its related factors separately (insulin resistance, dyslipidemia and inflammation). After a descriptive analysis of MS in the STANISLAS cohort, we undertook a longitudinal study based on two visits. We found that MS is present in a non negligible percentage of the STANISLAS cohort participants. MS prevalence (NCEP-ATPIII definition) was, in men and women respectively, 5.9% and 2.1 % at time of inclusion and then rose to 7.2 and 5.4% five years later. Moreover, in children, parental MS seems to be predictive of higher values for TNF-a concentrations and lower levels ofHDL-C, two major indicators of cardiovascular risk. Based on the fact that the available definitions for MS do not fit to the STANISLAS population, we searched, with EGIR, NCEP-ATPIII and WHO definitions, the parameters that best discriminate individuals according to their MS status, by using a series of discriminant analyses. Moreover, a factor analysis enabled to give a global viewpoint of the factors that are associated to MS in our population. The following experimentations consisted in separate studies for the processes of insulin resistance (IR), dyslipidemia, and inflammation. IR was considered by the study of the subrogate fasting insulin concentration. We determined biological and genetic factors that influence insulin concentration in the subjects of the STANISLAS cohort, and established reference values for insulinemia in this population. Our results suggest an important role of the -455G/A polymorphism of the fibrinogen gene for determining insulinemia in men and women, either directly, or through interactions with the environment. The results on dyslipidemia were obtained by verifying statistically the hypotheses which were generated by data mining techniques. We found interactions involving the environment and genetic polymorphisms of the lipid metabolism (APOE codons 112/158 et APOB Thr71Ile). Additionally, an association was shown between the APOB Thr71Ile polymorphism and MS risk. Finally, we considered the pro-inflammatory state in MS by studying TNF-a and IL-6. We searched the biological variation factors of these two cytokines. Moreover, we established their familial correlations and we analyzed potential associations between genetic polymorphisms of the cytokines and the concentrations oftheir corresponding proteins. In conclusion, we showed that metabolic syndrome is present even in a supposedly healthy population. We tried to give some clues for understanding the complexity of MS and its associated factors thanks to the combination of data mining and statistics. The proposed methodologies for mining cohorts and for discovering rare itemsets are opened, and can be applied to other framework studies not only in biology but also in other research fields.
Document type :
Theses
File URL :
http://docnum.univ-lorraine.fr/prive/SCD_T_2005_0209_MAUMUS.pdf
Complete list of metadatas

https://hal.univ-lorraine.fr/tel-01746957
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 10:48:34 AM
Last modification on : Monday, April 16, 2018 - 10:40:42 AM

Identifiers

  • HAL Id : tel-01746957, version 1

Collections

Citation

Sandy Maumus. Approche de la complexité du syndrome métabolique et de ses indicateurs de risque par la mise en oeuvre de méthodes numériques et symboliques de fouille de données. Médecine humaine et pathologie. Université Henri Poincaré - Nancy 1, 2005. Français. ⟨NNT : 2005NAN12506⟩. ⟨tel-01746957⟩

Share

Metrics

Record views

7