Start typing the gene name and suggestions will appear in the scroll menu. MetaMEx works with official gene symbols, for instance the official gene name of PGC1α is PPARGC1A.
In order to give a transparent overview of the currently available data, all studies are presented, even if genes are not detected. Older studies, or custom arrays often have a limited number of probes and therefore fewer detected genes. On the other hand, the more recent RNA sequencing datasets often have more depth and detect non-coding RNAs which are not present in gene arrays.
A forest plot is a graphical representation of results from several scientific studies and is typically used to plot meta-analyses. The left-hand columns list the names of the studies, followed by the fold-change (log2), false discovery rate (FDR) and sample size (n) for each individual study. The right-hand column is a plot of the fold-change (log2) represented by a square and the 95% confidence intervals represented by horizontal lines. The area of each square is proportional to the study's weight (sample size) in the meta-analysis. The overall meta-analysed score is represented by a diamond on the bottom line, the lateral points of which indicate confidence intervals.
MetaMEx compiles more than 90 studies which include volunteers of different age, sex, weight, fitness, weight and health. Studies can be included or excluded from the analysis by scrolling at the bottom of the page and checking the boxes. For instance, select males or females by checking the corresponding tick boxes.
After selecting either acute exercise, exercise training or inactivity, a specific menu will appear on the right of the page. This menu includes parameters such as exercise duration or time of biopsy collection after exercise cessation. Another list will appear under the forest plot to select or unselect specific datasets.
All studies were annotated with as much information as possible about age, weight, health, biopsy, muscle, etc. The title of the studies reflects the clinical data and protocol used for a specific study.
A detailed description of the labels is availabe in Datasets/Annotation.
The meta-analysis was created by collecting publicly available studies on mRNA expression levels in human skeletal muscle after exercise or inactivity. Statistics were first performed individually for each array.
- Robust multiarray averaging was used for affymetrix arrays (oligo package)
- Quantile normalization was used for other microarrays (limma package)
- Variance stabilizing transformation (VST) was used for RNA sequencing datasets (DESeq2 package).
- Moderated t-statistics were calculated for each study with empirical Bayes statistics for differential expression (limma package).
The meta-analysis summary was calculated using restricted maximum likelihood (metafor package). The analysis was weighted using sample size (n) to adjust for studies with small number of volunteers.
The timeline was calculated by collecting all data available in the database and annotate them by either time of the biopsy or inactivity duration. Moderated t-statistics were calculated with empirical Bayes statistics after blocking for other confounding parameters (sex, age, exercise type…).
Whenever possible, we downloaded the raw data and re-processed studies using the same pipeline. That means that the normalization methods that we used might differ from the ones used by the original authors. In addition, samples were often insufficiently annotated to allow us to run paired statistics comparing pre/post interventions. We therefore had to used unpaired statistics and lost power in the process. Finally, many studies pooled individuals of different age and BMI to have higher sample size. To allow proper comparison in th
