ROC (‘receiver operator characteristics’) analysis is a visual as well as numerical method used for assessing the performance of classification algorithms, such as those used for predicting structures and functions from sequence data. This review summarizes the fundamental concepts of ROC analysis and the interpretation of results using examples of sequence and structure comparison. We overview the available programs and provide evaluation guidelines for genomic/proteomic data, with particular regard to applications to large and heterogeneous databases used in bioinformatics.

The possibility of generating neural cells from human bone-marrow-derived mesenchymal stem cells (hMSCs) by simple in vitro treatments is appealing both conceptually and practically. However, whether phenotypic modulations observed after chemical manipulation of such stem cells truly represent a genuine trans-lineage differentiation remains to be established. We have re-evaluated the effects of a frequently reported biochemical approach, based on treatment with butylated hydroxyanisole and dimethylsulphoxide, to bring about such phenotypic conversion by monitoring the morphological changes induced by the treatment in real time, by analysing the expression of phenotype-specific protein markers and by assessing the modulation of transcriptome. Video time-lapse microscopy showed that conversion of mesenchymal stem cells to a neuron-like morphology could be reproduced in normal primary fibroblasts as well as mimicked by addition of drugs eliciting cytoskeletal collapse and disruption of focal adhesion contacts. Analysis of markers revealed that mesenchymal stem cells constitutively expressed multi-lineage traits, including several pertaining to the neural one. However, the applied `neural induction' protocol neither significantly modulated the expression of such markers, nor induced de novo translation of other neural-specific proteins. Similarly, global expression profiling of over 21,000 genes demonstrated that gene transcription was poorly affected. Most strikingly, we found that the set of genes whose expression was altered by the inductive treatment did not match those sets of genes differentially expressed when comparing untreated mesenchymal stem cells and immature neural tissues. Conversely, by comparing these gene expression profiles with that obtained from comparisons between the same cells and an unrelated non-neural organ, such as liver, we found that the adopted neural induction protocol was no more effective in redirecting human mesenchymal stem cells toward a neural phenotype than toward an endodermal hepatic pathway.

COLOMBOS is a database that integrates publicly available transcriptomics data for several prokaryotic model organisms. Compared to the previous version it has more than doubled in size, both in terms of species and data available. The manually curated condition annotation has been overhauled as well, giving more complete information about samples’ experimental conditions and their differences. Functionality-wise cross-species analyses now enable users to analyse expression data for all species simultaneously, and identify candidate genes with evolutionary conserved expression behaviour. All the expression-based query tools have undergone a substantial improvement, overcoming the limit of enforced co-expression data retrieval and instead enabling the return of more complex patterns of expression behaviour. COLOMBOS is freely available through a web application at http://colombos.net/. The complete database is also accessible via REST API or downloadable as tab-delimited text files.

The clinical course of cutaneous melanoma (CM) can differ significantly for patients with identical stages of disease, defined clinico-pathologically, and no molecular markers differentiate patients with such a diverse prognosis. This study aimed to define the prognostic value of whole genome DNA methylation profiles in stage III CM.

The COLOMBOS database (http://www.colombos.net) features comprehensive organism-specific cross-platform gene expression compendia of several bacterial model organisms and is supported by a fully interactive web portal and an extensive web API. COLOMBOS was originally published in PLoS One, and COLOMBOS v2.0 includes both an update of the expression data, by expanding the previously available compendia and by adding compendia for several new species, and an update of the surrounding functionality, with improved search and visualization options and novel tools for programmatic access to the database. The scope of the database has also been extended to incorporate RNA-seq data in our compendia by a dedicated analysis pipeline. We demonstrate the validity and robustness of this approach by comparing the same RNA samples measured in parallel using both microarrays and RNA-seq. As far as we know, COLOMBOS currently hosts the largest homogenized gene expression compendia available for seven bacterial model organisms.