A statistical methodology to select covariates in high-dimensional data under dependence. Application to the classification of genetic profiles in oncology.

We propose a new methodology for selecting and ranking covariates associated with a variable of interest in a context of high-dimensional data under dependence but few observations. The methodology successively intertwines the clustering of covariates, decorrelation of covariates using Factor Latent Analysis, selection using aggregation of adapted methods and finally ranking. A simulation study shows the interest of the decorrelation inside the different clusters of covariates. We first apply our method to transcriptomic data of 37 patients with advanced non-small-cell lung cancer who have received chemotherapy, to select the transcriptomic covariates that explain the survival outcome of the treatment. Secondly, we apply our method to 79 breast tumor samples to define patient profiles for a new metastatic biomarker and associated gene network in order to personalize the treatments.

Spatial and Temporal Pattern Analyses of Esca Grapevine Disease in Vineyards in France.

To assess the capacity of esca to spread within vineyards of the Bordeaux region, over 8 years of annual records, containing between 1,200 and 2,300 contiguous Cabernet Sauvignon vines from 15 mature vineyards, were used for spatial statistical analyses. A group of nonparametric tests, based on join count statistics and on permutation methods, was developed to characterize the spatial structure of esca-symptomatic vines in terms of spread in any direction or within-row only. Among vineyards, a large range of spatial patterns, from random to strongly structured, associated with various prevalence rates that increased over time were observed. In four vineyards, the complex esca distribution pattern indicated different levels of clustering. By contrast, in other vineyards, only small clusters of two adjacent symptomatic vines were observed, and they were localized along rows, without enlargement over time, except in one vineyard. An analysis of spatial dependence between previously and newly symptomatic vines within k-order neighborhoods (k = 1 to 5), showed, for 5 of the 15 vineyards, that the newly symptomatic vines were located close to previously infected vines, without a favored orientation or neighbor order. All the results together suggested a limited potential for secondary local spread from neighboring symptomatic vines.

European reference equations for CO and NO lung transfer.

The aim of the present study was to calculate reference equations for carbon monoxide and nitric oxide transfer, measured in two distinct populations. The transfer factor of the lung for nitric oxide (T(L,NO)) and carbon monoxide (T(L,CO)) were measured in 303 people aged 18-94 yrs. Measurements were similarly made in two distant cities, using the single-breath technique. Capillary lung volume (V(c)) and membrane conductance, the diffusing capacity of the membrane (D(m)), for carbon monoxide (D(m,CO)) were derived. The transfer of both gases appeared to depend upon age, height, sex and localisation. The rate of decrease in both transfers increased after the age of 59 yrs. T(L,NO)/alveolar volume (V(A)) and T(L,CO)/V(A) were only age-dependent. The mean T(L,NO)/T(L,CO) was 4.75 and the mean D(m)/V(c) was 6.17 min(-1) x kPa(-1); these parameters were independent of any covariate. V(c) and D(m,CO) calculations depend upon the choice of coefficients included in the Roughton-Forster equation. Values of 1.97 for D(m,NO)/D(m,CO) ratio and 12.86 min x kPa(-1) for 1/red cell CO conductance are recommended. The scatter of transfer reference values in the literature, including the current study, is wide. The present results suggest that differences might be due to the populations themselves and not the methods alone.