The participants performed endurance training, walking/running 'uphill' on a treadmill 3 times per week for 16 weeks. They warmed-up for 10 minutes at 70% of maximal heart frequency (Hfmax) before performing 4 intervals of 4 minutes at 90-95% of Hfmax, with 3 minutes of active recovery at 70% of Hfmax between each interval. In the end, they had a 5-minute cool-down period, creating a total of 40 minutes
Extracted molecule
total RNA
Extraction protocol
RNA was isolated from whole blood using the PAXgene Blood RNA Kit (Qiagen, Germantown, MD) and then globin RNA was removed with GLOBINclear (Ambion, Austin, TX) according to the manufacturer's instructions. RNA integrity, purity and quantity were assessed by Bioanalyzer (Agilent Technologies, Santa Clara, CA) and Nanodrop (NanoDrop Technologies, Baltimore, MD). The concentration of total RNA was measured by Nanodrop with ultraviolet spectrophotometry at 260/280 nm. RNA quality was assessed by electrophoresis on Bioanalyzer chips (Agilent Technologies Santa Clara, CA). High quality RNA was classified as a 260/280 ratio above 1.8. Only samples with a 260/280 ratio between 1.8-2.2 and no signs of degradation were used for analysis.
Label
Biotin
Label protocol
According to manufacture's instructions
Hybridization protocol
RNA from each sample was hybridized to Applied Biosystem (AB) Human Genome Survey microarray v.2.0. The order of labeling and hybridization was randomised to avoid technical batch effects. The microarray analyses were performed on the AB 1700 Array Expression System.
Scan protocol
According to manufacture's instructions
Description
22 Pre exercise
Data processing
The data files from the AB 1700 Chemiluminescent Microarray Analyzer Software were processed using J-Express Pro v.2.7 to filter and normalize the data from each hybridisation and compile gene expression profile matrix (gene by sample) data sets for further analysis. The pre-normalized intensity values were extracted per spot from the data files, and all flagged, weak and control spots were filtered out. An unsupervised analysis of all samples using the Correspondence Analysis tool in JExpress was performed in order to spot any microarray-outliers in the dataset. A filtering value of signal to noise ratio (S/N) of minimum 3.0 was chosen to reduces most of the noise The S/N filter creates missing values. Genes with low S/N for several samples had less confidence and was removed from the analysis. Since this dataset contained a relative small number of paired samples, we decided to be strict on the number of missing values allowed, in order to have enough values to calculate the missing values from, in the imputation step. Before compiled into an expression profile data matrix, all arrays were quantile normalized to be comparable. Quantile normalization is a widely used method for 1-channel data. Genes with at most 10% missing values were allowed in the final data set. The signal intensities in the dataset were further log transformed (base 2) and missing values were replaced by imputation using Adaptive LSimpute
