MOTIVATION: Several gene expression-based risk scores and subtype classifiers for breast cancer were developed to distinguish high- and low-risk patients. Evaluating the performance of these classifiers helps to decide which classifiers should be used in clinical practice for personal therapeutic recommendations. So far, studies that compared multiple classifiers in large independent patient cohorts mostly used microarray measurements. qPCR-based classifiers were not included in the comparison or had to be adapted to the different experimental platforms. RESULTS: We used a prospective study of 726 early breast cancer patients from seven certified German breast cancer centers. Patients were treated according to national guidelines and the expressions of 94 selected genes were measured by the mid-throughput qPCR platform Fluidigm. Clinical and pathological data including outcome over five years is available. Using these data, we could compare the performance of six classifiers (scmgene and research versions of PAM50, ROR-S, recurrence score, EndoPredict and GGI). Similar to other studies, we found a similar or even higher concordance between most of the classifiers and most were also able to differentiate high- and low-risk patients. The classifiers that were originally developed for microarray data still performed similarly using the Fluidigm data. Therefore, Fluidigm can be used to measure the gene expressions needed by several classifiers for a large cohort with little effort. In addition, we provide an interactive report of the results, which enables a transparent, in-depth comparison of classifiers and their prediction of individual patients. AVAILABILITY AND IMPLEMENTATION: https://services.bio.ifi.lmu.de/pia/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: The goal of many genome-wide experiments is to explain the changes between the analyzed conditions. Typically, the analysis is started with a set of differential genes DG and the first step is to identify the set of relevant biological processes BP . Current enrichment methods identify the involved biological process via statistically significant overrepresentation of differential genes in predefined sets, but do not further explain how the differential genes interact with each other or which other genes might be important for the enriched process. Other network-based methods determine subnetworks of interacting genes containing many differential genes, but do not employ process knowledge for a more focused analysis. RESULTS: RelExplain is a method to analyze a given biological process bp (e.g. identified by enrichment) in more detail by computing an explanation using the measured DG and a given network. An explanation is a subnetwork that contains the differential genes in the process bp and connects them in the best way given the experimental data using also genes that are not differential or not in bp . RelExplain takes into account the functional annotations of nodes and the edge consistency of the measurements. Explanations are compact networks of the relevant part of the bp and additional nodes that might be important for the bp . Our evaluation showed that RelExplain is better suited to retrieve manually curated subnetworks from unspecific networks than other algorithms. The interactive RelExplain tool allows to compute and inspect sub-optimal and alternative optimal explanations. AVAILABILITY AND IMPLEMENTATION: A webserver is available at https://services.bio.ifi.lmu.de/relexplain . CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: A recent large-scale analysis of Gene Expression Omnibus (GEO) data found frequent evidence for spatial defects in a substantial fraction of Affymetrix microarrays in the GEO. Nevertheless, in contrast to quality assessment, artefact detection is not widely used in standard gene expression analysis pipelines. Furthermore, although approaches have been proposed to detect diverse types of spatial noise on arrays, the correction of these artefacts is mostly left to either summarization methods or the corresponding arrays are completely discarded. RESULTS: We show that state-of-the-art robust summarization procedures are vulnerable to artefacts on arrays and cannot appropriately correct for these. To address this problem, we present a simple approach to detect artefacts with high recall and precision, which we further improve by taking into account the spatial layout of arrays. Finally, we propose two correction methods for these artefacts that either substitute values of defective probes using probeset information or filter corrupted probes. We show that our approach can identify and correct defective probe measurements appropriately and outperforms existing tools. CONCLUSIONS: While summarization is insufficient to correct for defective probes, this problem can be addressed in a straightforward way by the methods we present for identification and correction of defective probes. As these methods output CEL files with corrected probe values that serve as input to standard normalization and summarization procedures, they can be easily integrated into existing microarray analysis pipelines as an additional pre-processing step. An R package is freely available from http://www.bio.ifi.lmu.de/artefact-correction.