The dendrograms of sample relations display the distances or degree of relationship between experimental samples.
Data quality and sample relations were assessed using the Bioconductor lumi package [ 13].
The quality of the samples and the sample relations were further investigated using correlation analysis and clustering using Pearson's metrics (Additional file 1).
RNA-seq sample relations were evaluated based on multidimensional scaling using the plotMDS package of DESeq (Anders and Huber 2010) in R (www.R-project.org, last accessed May 23 , 2014.
In order to retrieve insight into group similarities, sample relations were then investigated by measuring the Euclidean distance between the miRNA expression profiles, followed by hierarchical clustering.
Despite the differences in target amplification, sample relations were preserved across the five target preparation methods as shown by unsupervised hierarchical clustering.
To inspect sample relations, variance-stabilizing transformed count data were used to build an Euclidean distance matrix, followed by hierarchical clustering analysis, as described elsewhere [ 16]. miRNA differential expression analysis was performed using Bioconductor's package DESeq [ 16].
Although having detailed metadata on content, provenance and sample relations enables both replication and reuse, thereby increasing the data's scientific impact, scientists will in reality weigh these benefits against the labour required to report and maintain the metadata.
Thus, it was important in our analyses that the sample-to-sample relations are maintained among each target preparation method.
