Compartment-resolved Proteomic Analysis of Mouse Aorta during Atherosclerotic Plaque Formation Reveals Osteoclast-specific Protein Expression.

Atherosclerosis leads to vascular lesions that involve major rearrangements of the vascular proteome, especially of the extracellular matrix (ECM). Using single aortas from ApoE knock out mice, we quantified formation of plaques by single-run, high-resolution mass spectrometry (MS)-based proteomics. To probe localization on a proteome-wide scale we employed quantitative detergent solubility profiling. This compartment- and time-resolved resource of atherogenesis comprised 5117 proteins, 182 of which changed their expression status in response to vessel maturation and atherosclerotic plaque development. In the insoluble ECM proteome, 65 proteins significantly changed, including relevant collagens, matrix metalloproteinases and macrophage derived proteins. Among novel factors in atherosclerosis, we identified matrilin-2, the collagen IV crosslinking enzyme peroxidasin as well as the poorly characterized MAM-domain containing 2 (Mamdc2) protein as being up-regulated in the ECM during atherogenesis. Intriguingly, three subunits of the osteoclast specific V-ATPase complex were strongly increased in mature plaques with an enrichment in macrophages thus implying an active de-mineralization function.

Phosphoproteome Analysis Reveals Differential Mode of Action of Sorafenib in Wildtype and Mutated FLT3 Acute Myeloid Leukemia (AML) Cells.

Constitutively activating internal tandem duplication (ITD) alterations of the receptor tyrosine kinase FLT3 (Fms-like tyrosine kinase 3) are common in acute myeloid leukemia (AML) and classifies FLT3 as an attractive therapeutic target. So far, applications of FLT3 small molecule inhibitors have been investigated primarily in FLT3-ITD+ patients. Only recently, a prolonged event-free survival has been observed in AML patients who were treated with the multikinase inhibitor sorafenib in addition to standard therapy. Here, we studied the sorafenib effect on proliferation in a panel of 13 FLT3-ITD- and FLT3-ITD+ AML cell lines. Sorafenib IC50 values ranged from 0.001 to 5.6 µm, whereas FLT3-ITD+ cells (MOLM-13, MV4-11) were found to be more sensitive to sorafenib than FLT3-ITD- cells. However, we identified two FLT3-ITD- cell lines (MONO-MAC-1 and OCI-AML-2) which were also sorafenib sensitive. Phosphoproteome analyses revealed that the affected pathways differed in sorafenib sensitive FLT3-ITD- and FLT3-ITD+ cells. In MV4-11 cells sorafenib suppressed mTOR signaling by direct inhibition of FLT3. In MONO-MAC-1 cells sorafenib inhibited the MEK/ERK pathway. These data suggest that the FLT3 status in AML patients might not be the only factor predicting response to treatment with sorafenib.

Time- and compartment-resolved proteome profiling of the extracellular niche in lung injury and repair.

The extracellular matrix (ECM) is a key regulator of tissue morphogenesis and repair. However, its composition and architecture are not well characterized. Here, we monitor remodeling of the extracellular niche in tissue repair in the bleomycin-induced lung injury mouse model. Mass spectrometry quantified 8,366 proteins from total tissue and bronchoalveolar lavage fluid (BALF) over the course of 8 weeks, surveying tissue composition from the onset of inflammation and fibrosis to its full recovery. Combined analysis of proteome, secretome, and transcriptome highlighted post-transcriptional events during tissue fibrogenesis and defined the composition of airway epithelial lining fluid. To comprehensively characterize the ECM, we developed a quantitative detergent solubility profiling (QDSP) method, which identified Emilin-2 and collagen-XXVIII as novel constituents of the provisional repair matrix. QDSP revealed which secreted proteins interact with the ECM, and showed drastically altered association of morphogens to the insoluble matrix upon injury. Thus, our proteomic systems biology study assigns proteins to tissue compartments and uncovers their dynamic regulation upon lung injury and repair, potentially contributing to the development of anti-fibrotic strategies.

Systematic evaluation of label-free and super-SILAC quantification for proteome expression analysis.

RATIONALE: Advanced implementations of mass spectrometry (MS)-based proteomics allow for comprehensive proteome expression profiling across many biological samples. The outcome of such studies critically depends on accurate and precise quantification, which has to be ensured for high-coverage proteome analysis possible on fast and sensitive mass spectrometers such as quadrupole orbitrap instruments. METHODS: We conducted ultra-high-performance liquid chromatography (UHPLC)/MS experiments on a Q Exactive to systematically compare label-free proteome quantification across six human cancer cell lines with quantification against a shared reference mix generated by stable isotope labeling with amino acids in cell culture (super-SILAC). RESULTS: Single-shot experiments identified on average about 5000 proteins in the label-free compared to about 3500 in super-SILAC experiments. Label-free quantification was slightly less precise than super-SILAC in replicate measurements, verifying previous results obtained for lower proteome coverage. Due to the higher number of quantified proteins, more significant differences were detected in label-free cell line comparisons, whereas a higher percentage of quantified proteins was identified as differentially expressed in super-SILAC experiments. Additional label-free replicate analyses effectively compensated for lower precision of quantification. Finally, peptide fractionation by high pH reversed-phase chromatography prior to LC/MS analysis further increased the robustness and precision of label-free quantification in conjunction with higher proteome coverage. CONCLUSIONS: Our results benchmark and highlight the utility of label-free proteome quantification for applications such as target and biomarker discovery on state-of-the-art UHPLC/MS workflows.

A SILAC-based approach identifies substrates of caspase-dependent cleavage upon TRAIL-induced apoptosis.

The extracellular ligand-induced extrinsic pathway of apoptosis is executed via caspase protease cascades that activate downstream effectors by means of site-directed proteolysis. Here we identify proteome changes upon the induction of apoptosis by the cytokine tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) in a Jurkat T cell line. We detected caspase-dependent cleavage substrates by quantifying protein intensities before and after TRAIL induction in SDS gel slices. Apoptotic protein cleavage events are identified by a characteristic stable isotope labeling with amino acids in cell culture (SILAC) ratio pattern across gel slices that results from differential migration of the cleaved and uncleaved proteins. We applied a statistical test to define apoptotic substrates in the proteome. Our approach identified more than 650 of these cleaved proteins in response to TRAIL-induced apoptosis, including many previously unknown substrates and cleavage sites. Inhibitor treatment combined with triple SILAC demonstrated that the detected cleavage events were caspase dependent. Proteins located in the lumina of organelles such as mitochondria and endoplasmic reticulum were significantly underrepresented in the substrate population. Interestingly, caspase cleavage is generally observed in not only one but several members of stable complexes, but often with lower stoichiometry. For instance, all five proteins of the condensin I complex were cleaved upon TRAIL treatment. The apoptotic substrate proteome data can be accessed and visualized in the MaxQB database and might prove useful for basic and clinical research into TRAIL-induced apoptosis. The technology described here is extensible to a wide range of other proteolytic cleavage events.

Comparative proteomic analysis of eleven common cell lines reveals ubiquitous but varying expression of most proteins.

Deep proteomic analysis of mammalian cell lines would yield an inventory of the building blocks of the most commonly used systems in biological research. Mass spectrometry-based proteomics can identify and quantify proteins in a global and unbiased manner and can highlight the cellular processes that are altered between such systems. We analyzed 11 human cell lines using an LTQ-Orbitrap family mass spectrometer with a "high field" Orbitrap mass analyzer with improved resolution and sequencing speed. We identified a total of 11,731 proteins, and on average 10,361 ± 120 proteins in each cell line. This very high proteome coverage enabled analysis of a broad range of processes and functions. Despite the distinct origins of the cell lines, our quantitative results showed surprisingly high similarity in terms of expressed proteins. Nevertheless, this global similarity of the proteomes did not imply equal expression levels of individual proteins across the 11 cell lines, as we found significant differences in expression levels for an estimated two-third of them. The variability in cellular expression levels was similar for low and high abundance proteins, and even many of the most highly expressed proteins with household roles showed significant differences between cells. Metabolic pathways, which have high redundancy, exhibited variable expression, whereas basic cellular functions such as the basal transcription machinery varied much less. We harness knowledge of these cell line proteomes for the construction of a broad coverage "super-SILAC" quantification standard. Together with the accompanying paper (Schaab, C. MCP 2012, PMID: 22301388) (17) these data can be used to obtain reference expression profiles for proteins of interest both within and across cell line proteomes.

Analysis of high accuracy, quantitative proteomics data in the MaxQB database.

MS-based proteomics generates rapidly increasing amounts of precise and quantitative information. Analysis of individual proteomic experiments has made great strides, but the crucial ability to compare and store information across different proteome measurements still presents many challenges. For example, it has been difficult to avoid contamination of databases with low quality peptide identifications, to control for the inflation in false positive identifications when combining data sets, and to integrate quantitative data. Although, for example, the contamination with low quality identifications has been addressed by joint analysis of deposited raw data in some public repositories, we reasoned that there should be a role for a database specifically designed for high resolution and quantitative data. Here we describe a novel database termed MaxQB that stores and displays collections of large proteomics projects and allows joint analysis and comparison. We demonstrate the analysis tools of MaxQB using proteome data of 11 different human cell lines and 28 mouse tissues. The database-wide false discovery rate is controlled by adjusting the project specific cutoff scores for the combined data sets. The 11 cell line proteomes together identify proteins expressed from more than half of all human genes. For each protein of interest, expression levels estimated by label-free quantification can be visualized across the cell lines. Similarly, the expression rank order and estimated amount of each protein within each proteome are plotted. We used MaxQB to calculate the signal reproducibility of the detected peptides for the same proteins across different proteomes. Spearman rank correlation between peptide intensity and detection probability of identified proteins was greater than 0.8 for 64% of the proteome, whereas a minority of proteins have negative correlation. This information can be used to pinpoint false protein identifications, independently of peptide database scores. The information contained in MaxQB, including high resolution fragment spectra, is accessible to the community via a user-friendly web interface at http://www.biochem.mpg.de/maxqb.

Phosphosignature predicts dasatinib response in non-small cell lung cancer.

Targeted drugs are less toxic than traditional chemotherapeutic therapies; however, the proportion of patients that benefit from these drugs is often smaller. A marker that confidently predicts patient response to a specific therapy would allow an individual therapy selection most likely to benefit the patient. Here, we used quantitative mass spectrometry to globally profile the basal phosphoproteome of a panel of non-small cell lung cancer cell lines. The effect of the kinase inhibitor dasatinib on cellular growth was tested against the same panel. From the phosphoproteome profiles, we identified 58 phosphorylation sites, which consistently differ between sensitive and resistant cell lines. Many of the corresponding proteins are involved in cell adhesion and cytoskeleton organization. We showed that a signature of only 12 phosphorylation sites is sufficient to accurately predict dasatinib sensitivity. Four of the phosphorylation sites belong to integrin ß4, a protein that mediates cell-matrix or cell-cell adhesion. The signature was validated in cross-validation and label switch experiments and in six independently profiled breast cancer cell lines. The study supports that the phosphorylation of integrin ß4, as well as eight further proteins comprising the signature, are candidate biomarkers for predicting response to dasatinib in solid tumors. Furthermore, our results show that identifying predictive phosphorylation signatures from global, quantitative phosphoproteomic data is possible and can open a new path to discovering molecular markers for response prediction.

Comparison of SILAC and mTRAQ quantification for phosphoproteomics on a quadrupole orbitrap mass spectrometer.

Advances in mass spectrometric methodology and instrumentation have promoted a continuous increase in analytical performance in the field of phosphoproteomics. Here, we employed the recently introduced quadrupole Orbitrap (Q Exactive) mass spectrometer for quantitative signaling analysis to a depth of more than 15 000 phosphorylation sites. In parallel to the commonly used SILAC approach, we evaluated the nonisobaric chemical labeling reagent mTRAQ as an alternative quantification technique. Both enabled high phosphoproteome coverage in H3122 lung cancer cells. Replicate quantifications by mTRAQ identified almost as many significant phosphorylation changes upon treatment with ALK kinase inhibitor crizotinib as found by SILAC quantification. Overall, mTRAQ was slightly less precise than SILAC as evident from a somewhat higher variance of replicate phosphosite ratios. Direct comparison of SILAC- and mTRAQ-quantified phosphosites revealed that the majority of changes were detected by either quantification techniques, but also highlighted the aspect of false negative identifications in quantitative proteomics applications. Further inspection of crizotinib-regulated phosphorylation changes unveiled interference with multiple antioncogenic mechanisms downstream of ALK fusion kinase in H3122 cells. In conclusion, our results demonstrate a strong analytical performance of the Q Exactive in global phosphoproteomics, and establish mTRAQ quantification as a useful alternative to metabolic isotope labeling.

Delayed times to tissue fixation result in unpredictable global phosphoproteome changes.

Protein phosphorylation controls the activity of signal transduction pathways regulated by kinases and phosphatases. Little is known, however, about the impact of preanalytical factors, for example, delayed times to tissue fixation, on global phosphoprotein levels in tissues. The aim of this study was to characterize the potential effects of delayed tissue preservation (cold ischemia) on the levels of phosphoproteins using targeted and nontargeted proteomic approaches. Rat and murine liver samples were exposed to different cold ischemic conditions ranging from 10 to 360 min prior to cryopreservation. The phosphoproteome was analyzed using reverse phase protein array (RPPA) technology and phosphoprotein-enriched quantitative tandem mass spectrometry (LC-MS/MS). RPPA analysis of rat liver tissues with long (up to 360 min) cold ischemia times did not reveal statistically significant alterations of specific phosphoproteins even though nonphosphorylated cytokeratin 18 (CK18) showed increased levels after 360 min of delay to freezing. Keeping the samples on ice prior to cryopreservation prevented this effect. LC-MS/MS-based quantification of 1684 phosphorylation sites in rat liver tissues showed broadening of their distribution compared to time point zero, but without reaching statistical significance for individual phosphosites. Similarly, RPPA analysis of mouse liver tissues with short (<60 min) cold ischemia times did not reveal directed or predictable changes of protein and phosphoprotein levels. Using LC-MS/MS and quantification of 791 phosphorylation sites, we found that the distribution of ratios compared to time point zero broadens with prolonged ischemia times, but these were rather undirected and diffuse changes, as we could not detect significant alterations of individual phosphosites. On the basis of our results from RPPA and LC-MS/MS analysis of rat and mouse liver tissues, we conclude that prolonged cold ischemia results in unspecific phosphoproteome changes that can be neither predicted nor assigned to individual proteins. On the other hand, we identified a number of phosphosites which were extraordinarily stable even after 360 min of cold ischemia and, therefore, may be used as general reference markers for future companion diagnostics for kinase inhibitors.

Proteome-wide analysis of temporal phosphorylation dynamics in lysophosphatidic acid-induced signaling.

Most growth factor receptors trigger phosphorylation-based signal transduction to translate environmental stimuli into defined biological responses. In addition to comprehensive and reliable assessment of growth factor-induced phosphoregulation, temporal resolution is needed to gain insights into the organizing principles of the cellular signaling machinery. Here, we introduce a refined experimental design for MS-based phosphoproteomics to reconcile the need for high comprehensiveness and temporal resolution with the key requirement of monitoring biological reproducibility. We treated SILAC-labeled SCC-9 cells with the seven transmembrane receptor ligand lysophosphatidic acid (LPA) and identified more than 17 000 phosphorylation sites. Filtering for biological replicate quantification yielded five-time point profiles for 6292 site-specific phosphorylations, which we analyzed for statistically significant regulation. Notably, about 30% of these sites changed significantly upon LPA stimulation, indicating extensive phosphoproteome regulation in response to this growth factor. Analysis of time series data identified distinct temporal profiles for different kinase substrate motifs, likely reflecting temporal orchestration of cellular kinase activities. Our data further indicated coordinated regulation of biological processes and phosphoprotein networks upon LPA stimulation. Finally, we detected regulation of functionally characterized phosphorylation sites not yet implicated in LPA signaling, which may foster a better understanding how LPA regulates cellular physiology on the molecular level.

Identifying differentially regulated subnetworks from phosphoproteomic data.

BACKGROUND: Various high throughput methods are available for detecting regulations at the level of transcription, translation or posttranslation (e.g. phosphorylation). Integrating these data with protein networks should make it possible to identify subnetworks that are significantly regulated. Furthermore, such integration can support identification of regulated entities from often noisy high throughput data. In particular, processing mass spectrometry-based phosphoproteomic data in this manner may expose signal transduction pathways and, in the case of experiments with drug-treated cells, reveal the drug's mode of action. RESULTS: Here, we introduce SubExtractor, an algorithm that combines phosphoproteomic data with protein network information from STRING to identify differentially regulated subnetworks and individual proteins. The method is based on a Bayesian probabilistic model combined with a genetic algorithm and rigorous significance testing. The Bayesian model accounts for information about both differential regulation and network topology. The method was tested with artificial data and subsequently applied to a comprehensive phosphoproteomics study investigating the mode of action of sorafenib, a small molecule kinase inhibitor. CONCLUSIONS: SubExtractor reliably identifies differentially regulated subnetworks from phosphoproteomic data by integrating protein networks. The method can also be applied to gene or protein expression data.

Translation and evaluation of a pre-clinical 5-protein response prediction signature in a breast cancer phase Ib clinical trial.

Human protein biomarker discovery relies heavily on pre-clinical models, in particular established cell lines and patient-derived xenografts, but confirmation studies in primary tissue are essential to demonstrate clinical relevance. We describe in this study the process that was followed to clinically translate a 5-protein response signature predictive for the activity of an anti-HER3 monoclonal antibody (lumretuzumab) originally measured in fresh frozen xenograft tissue. We detail the development, qualification, and validation of the multiplexed targeted mass spectrometry assay used to assess the signature performance in formalin-fixed, paraffin-embedded human clinical samples collected in a phase Ib trial designed to evaluate lumretuzumab in patients with metastatic breast cancer. We believe that the strategy delineated here provides a path forward to avoid the time- and cost-consuming step of having to develop immunological reagents against unproven targets. We expect that mass spectrometry-based platforms may become part of a rational process to rapidly test and qualify large number of candidate biomarkers to identify the few that stand a chance for further development and validation.

Improving DNA array data quality by minimising 'neighbourhood' effects.

Gene expression studies using cDNA arrays require robust and sensitive detection methods. Being extremely sensitive, radioactive detection suffers from the influence of signals positioned in each other's vicinity, the 'neighbourhood' effect. This limits the gene density of arrays and the quality of the results obtained. We have investigated the quantitative influence of different parameters on the 'neighbourhood' effect. By using a model experimental system, we could show that the effect is linear and depends only on the intensity of the hybridisation signal. We identified a common factor that can describe the influence of the neighbour spots based on their intensities. This factor is <1%, but it has to be taken into account if a high dynamic range of gene expression is to be detected. We could also derive the factor, although with less precision, from comparison of duplicate spots on arrays of 4565 different clones and replication of the hybridisation experiments. The calculated coefficient applied to our actual experimental results not only revealed previously undetected tissue or cell-specific expression differences, but also increased the dynamic range of detection. It thus provides a relatively simple way of improving DNA array data quality with few experimental modifications.

Comparative proteome analysis across non-small cell lung cancer cell lines.

Non-small cell lung cancer (NSCLC) cell lines are widely used model systems to study molecular aspects of lung cancer. Comparative and in-depth proteome expression data across many NSCLC cell lines has not been generated yet, but would be of utility for the investigation of candidate targets and markers in oncogenesis. We employed a SILAC reference approach to perform replicate proteome quantifications across 23 distinct NSCLC cell lines. On average, close to 4000 distinct proteins were identified and quantified per cell line. These included many known targets and diagnostic markers, indicating that our proteome expression data represents a useful resource for NSCLC pre-clinical research. To assess proteome diversity within the NSCLC cell line panel, we performed hierarchical clustering and principal component analysis of proteome expression data. Our results indicate that general proteome diversity among NSCLC cell lines supersedes potential effects common to K-Ras or epidermal growth factor receptor (EGFR) oncoprotein expression. However, we observed partial segregation of EGFR or KRAS mutant cell lines for certain principal components, which reflected biological differences according to gene ontology enrichment analyses. Moreover, statistical analysis revealed several proteins that were significantly overexpressed in KRAS or EGFR mutant cell lines.

Integrated genomics and proteomics define huntingtin CAG length-dependent networks in mice.

To gain insight into how mutant huntingtin (mHtt) CAG repeat length modifies Huntington's disease (HD) pathogenesis, we profiled mRNA in over 600 brain and peripheral tissue samples from HD knock-in mice with increasing CAG repeat lengths. We found repeat length-dependent transcriptional signatures to be prominent in the striatum, less so in cortex, and minimal in the liver. Coexpression network analyses revealed 13 striatal and 5 cortical modules that correlated highly with CAG length and age, and that were preserved in HD models and sometimes in patients. Top striatal modules implicated mHtt CAG length and age in graded impairment in the expression of identity genes for striatal medium spiny neurons and in dysregulation of cyclic AMP signaling, cell death and protocadherin genes. We used proteomics to confirm 790 genes and 5 striatal modules with CAG length-dependent dysregulation at the protein level, and validated 22 striatal module genes as modifiers of mHtt toxicities in vivo.

Phosphodiesterase 10A Inhibition Improves Cortico-Basal Ganglia Function in Huntington's Disease Models.

Huntington's disease (HD) symptoms are driven to a large extent by dysfunction of the basal ganglia circuitry. HD patients exhibit reduced striatal phoshodiesterase 10 (PDE10) levels. Using HD mouse models that exhibit reduced PDE10, we demonstrate the benefit of pharmacologic PDE10 inhibition to acutely correct basal ganglia circuitry deficits. PDE10 inhibition restored corticostriatal input and boosted cortically driven indirect pathway activity. Cyclic nucleotide signaling is impaired in HD models, and PDE10 loss may represent a homeostatic adaptation to maintain signaling. Elevation of both cAMP and cGMP by PDE10 inhibition was required for rescue. Phosphoproteomic profiling of striatum in response to PDE10 inhibition highlighted plausible neural substrates responsible for the improvement. Early chronic PDE10 inhibition in Q175 mice showed improvements beyond those seen with acute administration after symptom onset, including partial reversal of striatal deregulated transcripts and the prevention of the emergence of HD neurophysiological deficits. VIDEO ABSTRACT.

Pareto Optimization Identifies Diverse Set of Phosphorylation Signatures Predicting Response to Treatment with Dasatinib.

Multivariate biomarkers that can predict the effectiveness of targeted therapy in individual patients are highly desired. Previous biomarker discovery studies have largely focused on the identification of single biomarker signatures, aimed at maximizing prediction accuracy. Here, we present a different approach that identifies multiple biomarkers by simultaneously optimizing their predictive power, number of features, and proximity to the drug target in a protein-protein interaction network. To this end, we incorporated NSGA-II, a fast and elitist multi-objective optimization algorithm that is based on the principle of Pareto optimality, into the biomarker discovery workflow. The method was applied to quantitative phosphoproteome data of 19 non-small cell lung cancer (NSCLC) cell lines from a previous biomarker study. The algorithm successfully identified a total of 77 candidate biomarker signatures predicting response to treatment with dasatinib. Through filtering and similarity clustering, this set was trimmed to four final biomarker signatures, which then were validated on an independent set of breast cancer cell lines. All four candidates reached the same good prediction accuracy (83%) as the originally published biomarker. Although the newly discovered signatures were diverse in their composition and in their size, the central protein of the originally published signature - integrin ß4 (ITGB4) - was also present in all four Pareto signatures, confirming its pivotal role in predicting dasatinib response in NSCLC cell lines. In summary, the method presented here allows for a robust and simultaneous identification of multiple multivariate biomarkers that are optimized for prediction performance, size, and relevance.

Identification of significant features by the Global Mean Rank test.

With the introduction of omics-technologies such as transcriptomics and proteomics, numerous methods for the reliable identification of significantly regulated features (genes, proteins, etc.) have been developed. Experimental practice requires these tests to successfully deal with conditions such as small numbers of replicates, missing values, non-normally distributed expression levels, and non-identical distributions of features. With the MeanRank test we aimed at developing a test that performs robustly under these conditions, while favorably scaling with the number of replicates. The test proposed here is a global one-sample location test, which is based on the mean ranks across replicates, and internally estimates and controls the false discovery rate. Furthermore, missing data is accounted for without the need of imputation. In extensive simulations comparing MeanRank to other frequently used methods, we found that it performs well with small and large numbers of replicates, feature dependent variance between replicates, and variable regulation across features on simulation data and a recent two-color microarray spike-in dataset. The tests were then used to identify significant changes in the phosphoproteomes of cancer cells induced by the kinase inhibitors erlotinib and 3-MB-PP1 in two independently published mass spectrometry-based studies. MeanRank outperformed the other global rank-based methods applied in this study. Compared to the popular Significance Analysis of Microarrays and Linear Models for Microarray methods, MeanRank performed similar or better. Furthermore, MeanRank exhibits more consistent behavior regarding the degree of regulation and is robust against the choice of preprocessing methods. MeanRank does not require any imputation of missing values, is easy to understand, and yields results that are easy to interpret. The software implementing the algorithm is freely available for academic and commercial use.