GSAn: an alternative to enrichment analysis for annotating gene sets

Abstract

The revolution in new sequencing technologies, by strongly improving the production of omics data, is greatly leading to new understandings of the relations between genotype and phenotype. To interpret and analyze these massive data * To whom correspondence should be addressed. Tel: +44 000 0000000; Fax: +44 000 0000000; Email:xxx@yyyy.ac.zz † These authors contributed equally that are grouped according to a phenotype of interest, methods based on statistical enrichment became a standard in biology. However, these methods synthesize the biological information by a priori selecting the over-represented terms and may suffer from focusing on the most studied genes that represent a limited coverage of annotated genes within the gene set.To address these limitations, we developed GSAn, a novel gene set annotationWeb server that uses semantic similarity measures to reduce a priori Gene Ontology annotation terms. The originality of this new approach is to identify the best compromise between the number of retained annotation terms that has to be drastically reduced and the number of related genes that has to be as large as possible.Moreover, GSAn offers interactive visualization facilities dedicated to the multi-scale analysis of gene set annotations. GSAn is available at: https://gsan.labri.fr.

Activation of a neural stem cell transcriptional program in parenchymal astrocytes

“…For gene ontology (GO) analysis, we used DAVID v. 6.8 ( Huang et al, 2009 ) and GSAn ( Ayllon-Benitez et al, 2020 ). To analyze which subcellular compartments genes belonged to ( Figure 2e , S4a), DAVID’s GOTERM_CC_FAT option was used.…”

Distinct cellular origins and differentiation process account for distinct oncogenic and clinical behaviors of leiomyosarcomas

“…We annotated these 100 genes ( Figure S2B) using GSAn (Ayllon-Benitez et al, 2020) ( Table S1E) and found that 50 of them are part of the extracellular exosome, which are molecules (mRNA or proteins) exported to the extracellular space. This highlights the role of the extracellular matrix (ECM) and of cell-to-cell communication in hLMS pathology.…”

“…We found 14620 significant miRNA-mRNA interactions predicted in both ICGC and TCGA cohorts (negative Pearson's correlation coefficient (PCC), adjusted P<0.01), of which 210 were present in at least one database ( Table S2C). We annotated the 158 corresponding target genes (35 down-and 123 up-regulated in hLMS) with GSAn (19). Twenty-one terms with high specificity were mapped and none of them was specific to DIO3-DLK1 miRNA cluster, up-or down-regulated target genes, except "response to glucose" which was represented only by up-regulated genes ( Figure 2D, detailed in Table S1F).…”

simplifyEnrichment: an R/Bioconductor package for Clustering and Visualizing Functional Enrichment Results

“…Simplifying the enrichment results can be transformed into the task of clustering the similarity matrix of gene sets. Various clustering methods can be applied, for example, hierarchical clustering (Ayllon-Benitez et al , 2020) , affinity propagation clustering (Zhao and Wang, 2018) or density-based clustering (Huang et al , 2007) that have already been implemented in current tools. The similarity matrix can also be treated as an adjacency matrix and be converted to a graph where the gene sets are the nodes and the similarity values are the weights of edges.…”