diff --git a/README.md b/README.md
index 35ec151c885749472b045827e553257575fa7cfd..e5d5100f245c7f665f8fa0f9defb9abe54784d42 100644
--- a/README.md
+++ b/README.md
@@ -82,8 +82,8 @@ Launch _MSTclust_ without option to read the following documentation:
 * Specific rows can be selected using option `-r`.
 * A [minimum spanning tree](https://en.wikipedia.org/wiki/Minimum_spanning_tree) and a [single-linkage hierarchical classification tree](https://en.wikipedia.org/wiki/Single-linkage_clustering) can be written in [GraphML](http://graphml.graphdrawing.org/index.html) and [Newick](http://evolution.genetics.washington.edu/phylip/newicktree.html) output files using options `-m` and `-h`, respectively. These two options require important running times when considering large datasets (e.g. > 5,000 rows).
 * By definition, setting small cutoff values yield clustering with a large number of classes with fast running times.
-* Profile length is required to carry out data perturbation analyses (option `-L`).
-* Data subsampling analyses progressively subsample raw data with rate values _b_/(_B_+1) where _b_ = 1 ... _B_ and _B_ > 1 is the number of bins specified using option `-B`. At least _B_ = 9 bins (i.e. rates = 10%, 20% ... 90%) are recommended.
+* Profile length is required to carry out data perturbation analyses (option `-L`). For each statistics, _MSTclust_ will compute the average value and the 95% CI based on different replicates (option `-R`).
+* Data subsampling analyses progressively subsample raw data with rate values _b_/(_B_+1) where _b_ = 1 ... _B_ and _B_ > 1 is the number of bins specified using option `-B`. At least _B_ = 9 bins (i.e. rates = 10%, 20% ... 90%) are recommended. For each statistics, _MSTclust_ will compute the average value and the 95% CI based on different replicates (option `-R`).
 * For more details on the profile distance computation, clustering algorithm and descriptive statistics, see the technical notes pdf file.