Understanding cellular responses to external stimuli is critical for parsing biological mechanisms and advancing therapeutic development. High-content image-based assays provide a cost-effective approach to examine cellular phenotypes induced by diverse interventions, which offers valuable insights into biological processes and cellular states. We introduce MorphoDiff, a generative pipeline to predict high-resolution cell morphological responses under different conditions based on perturbation encoding. To the best of our knowledge, MorphoDiff is the first framework capable of producing guided, high-resolution predictions of cell morphology that generalize across both chemical and genetic interventions. The model integrates perturbation embeddings as guiding signals within a 2D latent diffusion model. The comprehensive computational, biological, and visual validations across three open-source Cell Painting datasets show that MorphoDiff can generate high-fidelity images and produce meaningful biology signals under various interventions. We envision the model will facilitate efficient in silico exploration of perturbational landscapes towards more effective drug discovery studies.

BACKGROUND: Identifying associations among biological variables is a major challenge in modern quantitative biological research, particularly given the systemic and statistical noise endemic to biological systems. Drug sensitivity data has proven to be a particularly challenging field for identifying associations to inform patient treatment. RESULTS: To address this, we introduce two semi-parametric variations on the commonly used concordance index: the robust concordance index and the kernelized concordance index (rCI, kCI), which incorporate measurements about the noise distribution from the data. We demonstrate that common statistical tests applied to the concordance index and its variations fail to control for false positives, and introduce efficient implementations to compute p-values using adaptive permutation testing. We then evaluate the statistical power of these coefficients under simulation and compare with Pearson and Spearman correlation coefficients. Finally, we evaluate the various statistics in matching drugs across pharmacogenomic datasets. CONCLUSIONS: We observe that the rCI and kCI are better powered than the concordance index in simulation and show some improvement on real data. Surprisingly, we observe that the Pearson correlation was the most robust to measurement noise among the different metrics.

The goal of precision oncology is to tailor treatment for patients individually using the genomic profile of their tumors. Pharmacogenomics datasets such as cancer cell lines are among the most valuable resources for drug sensitivity prediction, a crucial task of precision oncology. Machine learning methods have been employed to predict drug sensitivity based on the multiple omics data available for large panels of cancer cell lines. However, there are no comprehensive guidelines on how to properly train and validate such machine learning models for drug sensitivity prediction. In this paper, we introduce a set of guidelines for different aspects of training gene expression-based predictors using cell line datasets. These guidelines provide extensive analysis of the generalization of drug sensitivity predictors and challenge many current practices in the community including the choice of training dataset and measure of drug sensitivity. The application of these guidelines in future studies will enable the development of more robust preclinical biomarkers.

Performance of neural network models relies on the availability of large datasets with minimal levels of uncertainty. Transfer Learning (TL) models have been proposed to resolve the issue of small dataset size by letting the model train on a bigger, task-related reference dataset and then fine-tune on a smaller, task-specific dataset. In this work, we apply a transfer learning approach to improve predictive power in noisy data systems with large variable confidence datasets. We propose a deep neural network method called Filtered Transfer Learning (FTL) that defines multiple tiers of data confidence as separate tasks in a transfer learning setting. The deep neural network is fine-tuned in a hierarchical process by iteratively removing (filtering) data points with lower label confidence, and retraining. In this report we use FTL for predicting the interaction of drugs and proteins. We demonstrate that using FTL to learn stepwise, across the label confidence distribution, results in higher performance compared to deep neural network models trained on a single confidence range. We anticipate that this approach will enable the machine learning community to benefit from large datasets with uncertain labels in fields such as biology and medicine.

MOTIVATION: Individualized drug response prediction is a fundamental part of personalized medicine for cancer. Great effort has been made to discover biomarkers or to develop machine learning methods for accurate drug response prediction in cancers. Incorporating prior knowledge of biological systems into these methods is a promising avenue to improve prediction performance. High-throughput cell line assays of drug-induced transcriptomic perturbation effects are a prior knowledge that has not been fully incorporated into a drug response prediction model yet. RESULTS: We introduce a unified probabilistic approach, Drug Response Variational Autoencoder (Dr.VAE), that simultaneously models both drug response in terms of viability and transcriptomic perturbations. Dr.VAE is a deep generative model based on variational autoencoders. Our experimental results showed Dr.VAE to do as well or outperform standard classification methods for 23 out of 26 tested Food and Drug Administration-approved drugs. In a series of ablation experiments we showed that the observed improvement of Dr.VAE can be credited to the incorporation of drug-induced perturbation effects with joint modeling of treatment sensitivity. AVAILABILITY AND IMPLEMENTATION: Processed data and software implementation using PyTorch (Paszke et al., 2017) are available at: https://github.com/rampasek/DrVAE. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: High-throughput molecular profiles of human cells have been used in predictive computational approaches for stratification of healthy and malignant phenotypes and identification of their biological states. In this regard, pathway activities have been used as biological features in unsupervised and supervised learning schemes. RESULTS: We developed SIGN (Similarity Identification in Gene expressioN), a flexible open-source R package facilitating the use of pathway activities and their expression patterns to identify similarities between biological samples. We defined a new measure, the transcriptional similarity coefficient, which captures similarity of gene expression patterns, instead of quantifying overall activity, in biological pathways between the samples. To demonstrate the utility of SIGN in biomedical research, we establish that SIGN discriminates subtypes of breast tumors and patients with good or poor overall survival. SIGN outperforms the best models in DREAM challenge in predicting survival of breast cancer patients using the data from the Molecular Taxonomy of Breast Cancer International Consortium. In summary, SIGN can be used as a new tool for interrogating pathway activity and gene expression patterns in unsupervised and supervised learning schemes to improve prognostic risk estimation for cancer patients by the biomedical research community. AVAILABILITY AND IMPLEMENTATION: An open-source R package is available (https://cran.r-project.org/web/packages/SIGN/).

Pancreatic ductal adenocarcinoma (PDAC) has the worst prognosis among solid malignancies and improved therapeutic strategies are needed to improve outcomes. Patient-derived xenografts (PDX) and patient-derived organoids (PDO) serve as promising tools to identify new drugs with therapeutic potential in PDAC. For these preclinical disease models to be effective, they should both recapitulate the molecular heterogeneity of PDAC and validate patient-specific therapeutic sensitivities. To date however, deep characterization of the molecular heterogeneity of PDAC PDX and PDO models and comparison with matched human tumour remains largely unaddressed at the whole genome level. We conducted a comprehensive assessment of the genetic landscape of 16 whole-genome pairs of tumours and matched PDX, from primary PDAC and liver metastasis, including a unique cohort of 5 'trios' of matched primary tumour, PDX, and PDO. We developed a pipeline to score concordance between PDAC models and their paired human tumours for genomic events, including mutations, structural variations, and copy number variations. Tumour-model comparisons of mutations displayed single-gene concordance across major PDAC driver genes, but relatively poor agreement across the greater mutational load. Genome-wide and chromosome-centric analysis of structural variation (SV) events highlights previously unrecognized concordance across chromosomes that demonstrate clustered SV events. We found that polyploidy presented a major challenge when assessing copy number changes; however, ploidy-corrected copy number states suggest good agreement between donor-model pairs. Collectively, our investigations highlight that while PDXs and PDOs may serve as tractable and transplantable systems for probing the molecular properties of PDAC, these models may best serve selective analyses across different levels of genomic complexity.

Large-scale perturbation databases, such as Connectivity Map (CMap) or Library of Integrated Network-based Cellular Signatures (LINCS), provide enormous opportunities for computational pharmacogenomics and drug design. A reason for this is that in contrast to classical pharmacology focusing at one target at a time, the transcriptomics profiles provided by CMap and LINCS open the door for systems biology approaches on the pathway and network level. In this article, we provide a review of recent developments in computational pharmacogenomics with respect to CMap and LINCS and related applications.

Drug combinations have been proposed as a promising therapeutic strategy to overcome drug resistance and improve efficacy of monotherapy regimens in cancer. This strategy aims at targeting multiple components of this complex disease. Despite the increasing number of drug combinations in use, many of them were empirically found in the clinic, and the molecular mechanisms underlying these drug combinations are often unclear. These challenges call for rational, systematic approaches for drug combination discovery. Although high-throughput screening of single-agent therapeutics has been successfully implemented, it is not feasible to test all possible drug combinations, even for a reduced subset of anticancer drugs. Hence, in vitro and in vivo screening of a large number of drug combinations are not practical. Therefore, devising computational methods to efficiently explore the space of drug combinations and to discover efficacious combinations has attracted a lot of attention from the scientific community in the past few years. Nevertheless, in the absence of consensus regarding the computational approaches used to predict efficacious drug combinations, a plethora of methods, techniques and hypotheses have been developed to date, while the research field lacks an elaborate categorization of the existing computational methods and the available data sources. In this manuscript, we review and categorize the state-of-the-art computational approaches for drug combination prediction, and elaborate on the limitations of these methods and the existing challenges. We also discuss about the recent pan-cancer drug combination data sets and their importance in revising the available methods or developing more performant approaches.

Objectives: We sought to investigate the tissue specificity of drug sensitivities in large-scale pharmacological studies and compare these associations to those found in drug clinical indications. Materials and Methods: We leveraged the curated cell line response data from PharmacoGx and applied an enrichment algorithm on drug sensitivity values' area under the drug dose-response curves (AUCs) with and without adjustment for general level of drug sensitivity. Results: We observed tissue specificity in 63% of tested drugs, with 8% of total interactions deemed significant (false discovery rate <0.05). By restricting the drug-tissue interactions to those with AUC > 0.2, we found that in 52% of interactions, the tissue was predictive of drug sensitivity (concordance index > 0.65). When compared with clinical indications, the observed overlap was weak (Matthew correlation coefficient, MCC = 0.0003, P > .10). Discussion: While drugs exhibit significant tissue specificity in vitro, there is little overlap with clinical indications. This can be attributed to factors such as underlying biological differences between in vitro models and patient tumors, or the inability of tissue-specific drugs to bring additional benefits beyond gold standard treatments during clinical trials. Conclusion: Our meta-analysis of pan-cancer drug screening datasets indicates that most tested drugs exhibit tissue-specific sensitivities in a large panel of cancer cell lines. However, the observed preclinical results do not translate to the clinical setting. Our results suggest that additional research into showing parallels between preclinical and clinical data is required to increase the translational potential of in vitro drug screening.

Briefings in Bioinformatics (2017) doi: 10.1093/bib/bbw112 In the article above, the last name of ‘Galina Glazko’ was misspelled as ‘Galina Galzko’. The name has been corrected online and in print. The authors apologize for this error.

Molecular interrogation of a biological sample through DNA sequencing, RNA and microRNA profiling, proteomics and other assays, has the potential to provide a systems level approach to predicting treatment response and disease progression, and to developing precision therapies. Large publicly funded projects have generated extensive and freely available multi-assay data resources; however, bioinformatic and statistical methods for the analysis of such experiments are still nascent. We review multi-assay genomic data resources in the areas of clinical oncology, pharmacogenomics and other perturbation experiments, population genomics and regulatory genomics and other areas, and tools for data acquisition. Finally, we review bioinformatic tools that are explicitly geared toward integrative genomic data visualization and analysis. This review provides starting points for accessing publicly available data and tools to support development of needed integrative methods.

UNLABELLED: Breast cancer is one of the most frequent cancers among women. Extensive studies into the molecular heterogeneity of breast cancer have produced a plethora of molecular subtype classification and prognosis prediction algorithms, as well as numerous gene expression signatures. However, reimplementation of these algorithms is a tedious but important task to enable comparison of existing signatures and classification models between each other and with new models. Here, we present the genefu R/Bioconductor package, a multi-tiered compendium of bioinformatics algorithms and gene signatures for molecular subtyping and prognostication in breast cancer. AVAILABILITY AND IMPLEMENTATION: The genefu package is available from Bioconductor. http://www.bioconductor.org/packages/devel/bioc/html/genefu.html Source code is also available on Github https://github.com/bhklab/genefu CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Sequencing and microarray samples often are collected or processed in multiple batches or at different times. This often produces technical biases that can lead to incorrect results in the downstream analysis. There are several existing batch adjustment tools for '-omics' data, but they do not indicate a priori whether adjustment needs to be conducted or how correction should be applied. We present a software pipeline, BatchQC, which addresses these issues using interactive visualizations and statistics that evaluate the impact of batch effects in a genomic dataset. BatchQC can also apply existing adjustment tools and allow users to evaluate their benefits interactively. We used the BatchQC pipeline on both simulated and real data to demonstrate the effectiveness of this software toolkit. AVAILABILITY AND IMPLEMENTATION: BatchQC is available through Bioconductor: http://bioconductor.org/packages/BatchQC and GitHub: https://github.com/mani2012/BatchQC CONTACT: [email protected] information: Supplementary data are available at Bioinformatics online.

UNLABELLED: Pharmacogenomics holds great promise for the development of biomarkers of drug response and the design of new therapeutic options, which are key challenges in precision medicine. However, such data are scattered and lack standards for efficient access and analysis, consequently preventing the realization of the full potential of pharmacogenomics. To address these issues, we implemented PharmacoGx, an easy-to-use, open source package for integrative analysis of multiple pharmacogenomic datasets. We demonstrate the utility of our package in comparing large drug sensitivity datasets, such as the Genomics of Drug Sensitivity in Cancer and the Cancer Cell Line Encyclopedia. Moreover, we show how to use our package to easily perform Connectivity Map analysis. With increasing availability of drug-related data, our package will open new avenues of research for meta-analysis of pharmacogenomic data. AVAILABILITY AND IMPLEMENTATION: PharmacoGx is implemented in R and can be easily installed on any system. The package is available from CRAN and its source code is available from GitHub. CONTACT: [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Detection of allelic imbalances in ChIP-Seq reads is a powerful approach to identify functional non-coding single nucleotide variants (SNVs), either polymorphisms or mutations, which modulate the affinity of transcription factors for chromatin. We present ABC, a computational tool that identifies allele-specific binding of transcription factors from aligned ChIP-Seq reads at heterozygous SNVs. ABC controls for potential false positives resulting from biases introduced by the use of short sequencing reads in ChIP-Seq and can efficiently process a large number of heterozygous SNVs. RESULTS: ABC successfully identifies previously characterized functional SNVs, such as the rs4784227 breast cancer risk associated SNP that modulates the affinity of FOXA1 for the chromatin. AVAILABILITY AND IMPLEMENTATION: The code is open-source under an Artistic-2.0 license and versioned on GitHub (https://github.com/mlupien/ABC/). ABC is written in PERL and can be run on any platform with both PERL (≥5.18.1) and R (≥3.1.1) installed. The script requires the PERL Statistics::R module. CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Prior to applying genomic predictors to clinical samples, the genomic data must be properly normalized to ensure that the test set data are comparable to the data upon which the predictor was trained. The most effective normalization methods depend on data from multiple patients. From a biomedical perspective, this implies that predictions for a single patient may change depending on which other patient samples they are normalized with. This test set bias will occur when any cross-sample normalization is used before clinical prediction. RESULTS: We demonstrate that results from existing gene signatures which rely on normalizing test data may be irreproducible when the patient population changes composition or size using a set of curated, publicly available breast cancer microarray experiments. As an alternative, we examine the use of gene signatures that rely on ranks from the data and show why signatures using rank-based features can avoid test set bias while maintaining highly accurate classification, even across platforms. AVAILABILITY AND IMPLEMENTATION: The code, data and instructions necessary to reproduce our entire analysis is available at https://github.com/prpatil/testsetbias.

BACKGROUND: We recently identified two robust ovarian cancer subtypes, defined by the expression of genes involved in angiogenesis, with significant differences in clinical outcome. To identify potential regulatory mechanisms that distinguish the subtypes we applied PANDA, a method that uses an integrative approach to model information flow in gene regulatory networks. RESULTS: We find distinct differences between networks that are active in the angiogenic and non-angiogenic subtypes, largely defined by a set of key transcription factors that, although previously reported to play a role in angiogenesis, are not strongly differentially-expressed between the subtypes. Our network analysis indicates that these factors are involved in the activation (or repression) of different genes in the two subtypes, resulting in differential expression of their network targets. Mechanisms mediating differences between subtypes include a previously unrecognized pro-angiogenic role for increased genome-wide DNA methylation and complex patterns of combinatorial regulation. CONCLUSIONS: The models we develop require a shift in our interpretation of the driving factors in biological networks away from the genes themselves and toward their interactions. The observed regulatory changes between subtypes suggest therapeutic interventions that may help in the treatment of ovarian cancer.

Cancer is a complex disease that has proven to be difficult to understand on the single-gene level. For this reason a functional elucidation needs to take interactions among genes on a systems-level into account. In this study, we infer a colon cancer network from a large-scale gene expression data set by using the method BC3Net. We provide a structural and a functional analysis of this network and also connect its molecular interaction structure with the chromosomal locations of the genes enabling the definition of cis- and trans-interactions. Furthermore, we investigate the interaction of genes that can be found in close neighborhoods on the chromosomes to gain insight into regulatory mechanisms. To our knowledge this is the first study analyzing the genome-scale colon cancer network.

MOTIVATION: Feature selection is one of the main challenges in analyzing high-throughput genomic data. Minimum redundancy maximum relevance (mRMR) is a particularly fast feature selection method for finding a set of both relevant and complementary features. Here we describe the mRMRe R package, in which the mRMR technique is extended by using an ensemble approach to better explore the feature space and build more robust predictors. To deal with the computational complexity of the ensemble approach, the main functions of the package are implemented and parallelized in C using the openMP Application Programming Interface. RESULTS: Our ensemble mRMR implementations outperform the classical mRMR approach in terms of prediction accuracy. They identify genes more relevant to the biological context and may lead to richer biological interpretations. The parallelized functions included in the package show significant gains in terms of run-time speed when compared with previously released packages. AVAILABILITY: The R package mRMRe is available on Comprehensive R Archive Network and is provided open source under the Artistic-2.0 License. The code used to generate all the results reported in this application note is available from Supplementary File 1. CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

SUMMARY: The R/Bioconductor package RamiGO is an R interface to AmiGO that enables visualization of Gene Ontology (GO) trees. Given a list of GO terms, RamiGO uses the AmiGO visualize API to import Graphviz-DOT format files into R, and export these either as images (SVG, PNG) or into Cytoscape for extended network analyses. RamiGO provides easy customization of annotation, highlighting of specific GO terms, colouring of terms by P-value or export of a simplified summary GO tree. We illustrate RamiGO functionalities in a genome-wide gene set analysis of prognostic genes in breast cancer. AVAILABILITY AND IMPLEMENTATION: RamiGO is provided in R/Bioconductor, is open source under the Artistic-2.0 License and is available with a user manual containing installation, operating instructions and tutorials. It requires R version 2.15.0 or higher. URL: http://bioconductor.org/packages/release/bioc/html/RamiGO.html

BACKGROUND: An enduring challenge in personalized medicine lies in selecting the right drug for each individual patient. While testing of drugs on patients in large trials is the only way to assess their clinical efficacy and toxicity, we dramatically lack resources to test the hundreds of drugs currently under development. Therefore the use of preclinical model systems has been intensively investigated as this approach enables response to hundreds of drugs to be tested in multiple cell lines in parallel. METHODS: Two large-scale pharmacogenomic studies recently screened multiple anticancer drugs on over 1000 cell lines. We propose to combine these datasets to build and robustly validate genomic predictors of drug response. We compared five different approaches for building predictors of increasing complexity. We assessed their performance in cross-validation and in two large validation sets, one containing the same cell lines present in the training set and another dataset composed of cell lines that have never been used during the training phase. RESULTS: Sixteen drugs were found in common between the datasets. We were able to validate multivariate predictors for three out of the 16 tested drugs, namely irinotecan, PD-0325901, and PLX4720. Moreover, we observed that response to 17-AAG, an inhibitor of Hsp90, could be efficiently predicted by the expression level of a single gene, NQO1. CONCLUSION: These results suggest that genomic predictors could be robustly validated for specific drugs. If successfully validated in patients' tumor cells, and subsequently in clinical trials, they could act as companion tests for the corresponding drugs and play an important role in personalized medicine.

A major goal in translational cancer research is to identify biological signatures driving cancer progression and metastasis. A common technique applied in genomics research is to cluster patients using gene expression data from a candidate prognostic gene set, and if the resulting clusters show statistically significant outcome stratification, to associate the gene set with prognosis, suggesting its biological and clinical importance. Recent work has questioned the validity of this approach by showing in several breast cancer data sets that "random" gene sets tend to cluster patients into prognostically variable subgroups. This work suggests that new rigorous statistical methods are needed to identify biologically informative prognostic gene sets. To address this problem, we developed Significance Analysis of Prognostic Signatures (SAPS) which integrates standard prognostic tests with a new prognostic significance test based on stratifying patients into prognostic subtypes with random gene sets. SAPS ensures that a significant gene set is not only able to stratify patients into prognostically variable groups, but is also enriched for genes showing strong univariate associations with patient prognosis, and performs significantly better than random gene sets. We use SAPS to perform a large meta-analysis (the largest completed to date) of prognostic pathways in breast and ovarian cancer and their molecular subtypes. Our analyses show that only a small subset of the gene sets found statistically significant using standard measures achieve significance by SAPS. We identify new prognostic signatures in breast and ovarian cancer and their corresponding molecular subtypes, and we show that prognostic signatures in ER negative breast cancer are more similar to prognostic signatures in ovarian cancer than to prognostic signatures in ER positive breast cancer. SAPS is a powerful new method for deriving robust prognostic biological signatures from clinically annotated genomic datasets.

Abstract Summary: The survcomp package provides functions to assess and statistically compare the performance of survival/risk prediction models. It implements state-of-the-art statistics to (i) measure the performance of risk prediction models; (ii) combine these statistical estimates from multiple datasets using a meta-analytical framework; and (iii) statistically compare the performance of competitive models. Availability: The R/Bioconductor package survcomp is provided open source under the Artistic-2.0 License with a user manual containing installation, operating instructions and use case scenarios on real datasets. survcomp requires R version 2.13.0 or higher. http://bioconductor.org/packages/release/bioc/html/survcomp.html Contact: [email protected]; [email protected] Supplementary Information: Supplementary data are available at Bioinformatics online.

BACKGROUND: Traditional strategies for selecting variables in high dimensional classification problems aim to find sets of maximally relevant variables able to explain the target variations. If these techniques may be effective in generalization accuracy they often do not reveal direct causes. The latter is essentially related to the fact that high correlation (or relevance) does not imply causation. In this study, we show how to efficiently incorporate causal information into gene selection by moving from a single-input single-output to a multiple-input multiple-output setting. RESULTS: We show in synthetic case study that a better prioritization of causal variables can be obtained by considering a relevance score which incorporates a causal term. In addition we show, in a meta-analysis study of six publicly available breast cancer microarray datasets, that the improvement occurs also in terms of accuracy. The biological interpretation of the results confirms the potential of a causal approach to gene selection. CONCLUSIONS: Integrating causal information into gene selection algorithms is effective both in terms of prediction accuracy and biological interpretation.

BACKGROUND: Genome scale expression profiling of human tumor samples is likely to yield improved cancer treatment decisions. However, identification of clinically predictive or prognostic classifiers can be challenging when a large number of genes are measured in a small number of tumors. RESULTS: We describe an unsupervised method to extract robust, consistent metagenes from multiple analogous data sets. We applied this method to expression profiles from five "double negative breast cancer" (DNBC) (not expressing ESR1 or HER2) cohorts and derived four metagenes. We assessed these metagenes in four similar but independent cohorts and found strong associations between three of the metagenes and agent-specific response to neoadjuvant therapy. Furthermore, we applied the method to ovarian and early stage lung cancer, two tumor types that lack reliable predictors of outcome, and found that the metagenes yield predictors of survival for both. CONCLUSIONS: These results suggest that the use of multiple data sets to derive potential biomarkers can filter out data set-specific noise and can increase the efficiency in identifying clinically accurate biomarkers.

MOTIVATION: Survival prediction of breast cancer (BC) patients independently of treatment, also known as prognostication, is a complex task since clinically similar breast tumors, in addition to be molecularly heterogeneous, may exhibit different clinical outcomes. In recent years, the analysis of gene expression profiles by means of sophisticated data mining tools emerged as a promising technology to bring additional insights into BC biology and to improve the quality of prognostication. The aim of this work is to assess quantitatively the accuracy of prediction obtained with state-of-the-art data analysis techniques for BC microarray data through an independent and thorough framework. RESULTS: Due to the large number of variables, the reduced amount of samples and the high degree of noise, complex prediction methods are highly exposed to performance degradation despite the use of cross-validation techniques. Our analysis shows that the most complex methods are not significantly better than the simplest one, a univariate model relying on a single proliferation gene. This result suggests that proliferation might be the most relevant biological process for BC prognostication and that the loss of interpretability deriving from the use of overcomplex methods may be not sufficiently counterbalanced by an improvement of the quality of prediction. AVAILABILITY: The comparison study is implemented in an R package called survcomp and is available from http://www.ulb.ac.be/di/map/bhaibeka/software/survcomp/.