|Application:||Gene expression microarray analysis|
|Number of samples:||96|
|Release date:||Mar 23 2016|
|Last update date:||Oct 11 2016|
|Diseases:||Multiple Myeloma, Neoplasms|
|Dataset link||Reconstruction of microRNA/genes transcriptional regulatory networks of multiple myeloma through in silico integrative genomics analysis [MM, gene]|
miRNA and transcripts expression data were analyzed using MAGIA2, to identify mixed circuits (triplets) involving miRNA/gene/transcription factor (TF; http://gencomp.bio.unipd.it/magia2/), as previously described [Bisognin A et al, 2012, Nucl Acid Res]. Specifically, Targetscan was used as target prediction algorithm, and Pearson coefficient was used to measure relationships between microRNA and target mRNA expression profiles. Only the most variable 75% genes according to the coefficient of variation were considered. Lower threshold for absolute correlation coefficients within circuits was set to 0.2; 0.4 was used for miRNA/target binary relationships. Micrographite pipeline allows integrating pathway topologies with predicted and validated miRNA–target interactions, to perform integrated analyses of miRNA and gene expression profiles, for the identification of modulated regulatory circuits involved in the disease in terms of both expression variations and differential strength of inferred interactions [Calura E et al, 2014, Nucl Acid Res]. Micrographite has two steps: i) the extension of pathway annotation using miRNA-target interaction and ii) recursive topological pathway analysis on these networks. We considered network topologies derived from KEGG database by Graphite package [Sales G et al, 2012, BMC Bioinformatics] and miRNA-target gene interactions identified by the above-described MAGIA2 analysis. Specifically, a miRNA was added to a pathway-derived network only if one (or more) of its validated or predicted target genes is a pathway component. Then, a modified recursive version of CliPPER topological pathway analysis [Martini P et al, 2013; Nucl Acid Res] was applied to the composite network, as previously described [Calura E et al, 2014, Nucl Acid Res] in order to identify the most important and non-redundant circuit modulated across groups. Briefly, (i) in the first step, the most significant pathways were selected using P<0.1 as cut-off value for significance; (ii) for each dataset, the upper-scored 10th percentile of the portion of these previously selected pathways (i.e. “paths”, calculated over a 10,000-permutation step) mostly associated with phenotype were selected; and (iii) for each dataset a meta-pathway was assembled using the paths extracted from previous step and finally re-analyzed. GSE70254 and GSE70319 Samples with the same patient number represent the same sample, profiled using two different Platforms.
Citations per year