Precision mapping of an in vivo N-glycoproteome reveals rigid topological and sequence constraints.

N-linked glycosylation is a biologically important protein modification, but only a small fraction of modification sites have been mapped. We developed a "filter aided sample preparation" (FASP)-based method in which glycopeptides are enriched by binding to lectins on the top of a filter and mapped 6367 N-glycosylation sites on 2352 proteins in four mouse tissues and blood plasma using high-accuracy mass spectrometry. We found 74% of known mouse N-glycosites and discovered an additional 5753 sites on a diverse range of proteins. Sites almost always have the N-!P-[S|T]-!P (where !P is not proline) and rarely the N-X-C motif or nonconsensus sequences. Combining the FASP approach with analysis of subcellular glycosite localization reveals that the sites always orient toward the extracellular space or toward the lumen of ER, Golgi, lysosome, or peroxisome. The N-glycoproteome contains a plethora of modification sites on factors important in development, organ-specific functions, and disease.

USP7 small-molecule inhibitors interfere with ubiquitin binding.

The ubiquitin system regulates essential cellular processes in eukaryotes. Ubiquitin is ligated to substrate proteins as monomers or chains and the topology of ubiquitin modifications regulates substrate interactions with specific proteins. Thus ubiquitination directs a variety of substrate fates including proteasomal degradation. Deubiquitinase enzymes cleave ubiquitin from substrates and are implicated in disease; for example, ubiquitin-specific protease-7 (USP7) regulates stability of the p53 tumour suppressor and other proteins critical for tumour cell survival. However, developing selective deubiquitinase inhibitors has been challenging and no co-crystal structures have been solved with small-molecule inhibitors. Here, using nuclear magnetic resonance-based screening and structure-based design, we describe the development of selective USP7 inhibitors GNE-6640 and GNE-6776. These compounds induce tumour cell death and enhance cytotoxicity with chemotherapeutic agents and targeted compounds, including PIM kinase inhibitors. Structural studies reveal that GNE-6640 and GNE-6776 non-covalently target USP7 12 Å distant from the catalytic cysteine. The compounds attenuate ubiquitin binding and thus inhibit USP7 deubiquitinase activity. GNE-6640 and GNE-6776 interact with acidic residues that mediate hydrogen-bond interactions with the ubiquitin Lys48 side chain, suggesting that USP7 preferentially interacts with and cleaves ubiquitin moieties that have free Lys48 side chains. We investigated this idea by engineering di-ubiquitin chains containing differential proximal and distal isotopic labels and measuring USP7 binding by nuclear magnetic resonance. This preferential binding protracted the depolymerization kinetics of Lys48-linked ubiquitin chains relative to Lys63-linked chains. In summary, engineering compounds that inhibit USP7 activity by attenuating ubiquitin binding suggests opportunities for developing other deubiquitinase inhibitors and may be a strategy more broadly applicable to inhibiting proteins that require ubiquitin binding for full functional activity.

15 years of PhosphoSitePlus®: integrating post-translationally modified sites, disease variants and isoforms.

For 15 years the mission of PhosphoSitePlus® (PSP, https://www.phosphosite.org) has been to provide comprehensive information and tools for the study of mammalian post-translational modifications (PTMs). The number of unique PTMs in PSP is now more than 450 000 from over 22 000 articles and thousands of MS datasets. The most important areas of growth in PSP are in disease and isoform informatics. Germline mutations associated with inherited diseases and somatic cancer mutations have been added to the database and can now be viewed along with PTMs and associated quantitative information on novel 'lollipop' plots. These plots enable researchers to interactively visualize the overlap between disease variants and PTMs, and to identify mutations that may alter phenotypes by rewiring signaling networks. We are expanding the sequence space to include over 30 000 human and mouse isoforms to enable researchers to explore the important but understudied biology of isoforms. This represents a necessary expansion of sequence space to accommodate the growing precision and depth of coverage enabled by ongoing advances in mass spectrometry. Isoforms are aligned using a new algorithm. Exploring the worlds of PTMs and disease mutations in the entire isoform space will hopefully lead to new biomarkers, therapeutic targets, and insights into isoform biology.

Mapping N-glycosylation sites across seven evolutionarily distant species reveals a divergent substrate proteome despite a common core machinery.

N-linked glycosylation is an important posttranslational modification in all eukaryotes, but little is known about the N-glycoproteomes in nonmammalian systems. Here, we measure N-glycoproteomes of the major model organisms Arabidopsis thaliana, Schizosaccharomyces pombe, Saccharomyces cerevisiae, Caenorhabditis elegans, Drosophila melanogaster, and Danio rerio, representatively spanning the eukaryotic domain of life. The number of detected N-glycosylation sites varied between 425 in fission yeast, 516 in budding yeast, 1,794 in worm, 2,186 in plant, 2,229 in fly, and 2,254 in zebrafish. We find that all eukaryotic N-glycoproteomes have invariant characteristics including sequence recognition patterns, structural constraints, and subcellular localization. However, a surprisingly large percentage of the N-glycoproteome evolved after the phylogenetic divergences between plants, fungi, nematodes, insects, and vertebrates. Many N-glycosylated proteins coevolved with the rise of extracellular processes that are specific within corresponding phylogenetic groups and essential for organismal development, body growth, and organ formation.

Kras mutant genetically engineered mouse models of human cancers are genomically heterogeneous.

KRAS mutant tumors are largely recalcitrant to targeted therapies. Genetically engineered mouse models (GEMMs) of Kras mutant cancer recapitulate critical aspects of this disease and are widely used for preclinical validation of targets and therapies. Through comprehensive profiling of exomes and matched transcriptomes of >200 KrasG12D-initiated GEMM tumors from one lung and two pancreatic cancer models, we discover that significant intratumoral and intertumoral genomic heterogeneity evolves during tumorigenesis. Known oncogenes and tumor suppressor genes, beyond those engineered, are mutated, amplified, and deleted. Unlike human tumors, the GEMM genomic landscapes are dominated by copy number alterations, while protein-altering mutations are rare. However, interspecies comparative analyses of the genomic landscapes demonstrate fidelity between genes altered in KRAS mutant human and murine tumors. Genes that are spontaneously altered during murine tumorigenesis are also among the most prevalent found in human indications. Using targeted therapies, we also demonstrate that this inherent tumor heterogeneity can be exploited preclinically to discover cancer-specific and genotype-specific therapeutic vulnerabilities. Focusing on Kras allelic imbalance, a feature shared by all three models, we discover that MAPK pathway inhibition impinges uniquely on this event, indicating distinct susceptibility and fitness advantage of Kras-mutant cells. These data reveal previously unknown genomic diversity among KrasG12D-initiated GEMM tumors, places them in context of human patients, and demonstrates how to exploit this inherent tumor heterogeneity to discover therapeutic vulnerabilities.

Role of the E3 ubiquitin ligase RNF157 as a novel downstream effector linking PI3K and MAPK signaling pathways to the cell cycle.

The interconnected PI3K and MAPK signaling pathways are commonly perturbed in cancer. Dual inhibition of these pathways by the small-molecule PI3K inhibitor pictilisib (GDC-0941) and the MEK inhibitor cobimetinib (GDC-0973) suppresses cell proliferation and induces cell death better than either single agent in several preclinical models. Using mass spectrometry-based phosphoproteomics, we have identified the RING finger E3 ubiquitin ligase RNF157 as a target at the intersection of PI3K and MAPK signaling. We demonstrate that RNF157 phosphorylation downstream of the PI3K and MAPK pathways influences the ubiquitination and stability of RNF157 during the cell cycle in an anaphase-promoting complex/cyclosome-CDH1-dependent manner. Deletion of these phosphorylation-targeted residues on RNF157 disrupts binding to CDH1 and protects RNF157 from ubiquitination and degradation. Expression of the cyclin-dependent kinase 2 (CDK2), itself a downstream target of PI3K/MAPK signaling, leads to increased phosphorylation of RNF157 on the same residues modulated by PI3K and MAPK signaling. Inhibition of PI3K and MEK in combination or of CDK2 by their respective small-molecule inhibitors reduces RNF157 phosphorylation at these residues and attenuates RNF157 interaction with CDH1 and its subsequent degradation. Knockdown of endogenous RNF157 in melanoma cells leads to late S phase and G2/M arrest and induces apoptosis, the latter further potentiated by concurrent PI3K/MEK inhibition, consistent with a role for RNF157 in the cell cycle. We propose that RNF157 serves as a novel node integrating oncogenic signaling pathways with the cell cycle machinery and promoting optimal cell cycle progression in transformed cells.

Recurrent R-spondin fusions in colon cancer.

Identifying and understanding changes in cancer genomes is essential for the development of targeted therapeutics. Here we analyse systematically more than 70 pairs of primary human colon tumours by applying next-generation sequencing to characterize their exomes, transcriptomes and copy-number alterations. We have identified 36,303 protein-altering somatic changes that include several new recurrent mutations in the Wnt pathway gene TCF7L2, chromatin-remodelling genes such as TET2 and TET3 and receptor tyrosine kinases including ERBB3. Our analysis for significantly mutated cancer genes identified 23 candidates, including the cell cycle checkpoint kinase ATM. Copy-number and RNA-seq data analysis identified amplifications and corresponding overexpression of IGF2 in a subset of colon tumours. Furthermore, using RNA-seq data we identified multiple fusion transcripts including recurrent gene fusions involving R-spondin family members RSPO2 and RSPO3 that together occur in 10% of colon tumours. The RSPO fusions were mutually exclusive with APC mutations, indicating that they probably have a role in the activation of Wnt signalling and tumorigenesis. Consistent with this we show that the RSPO fusion proteins were capable of potentiating Wnt signalling. The R-spondin gene fusions and several other gene mutations identified in this study provide new potential opportunities for therapeutic intervention in colon cancer.

Site-specific identification of SUMO-2 targets in cells reveals an inverted SUMOylation motif and a hydrophobic cluster SUMOylation motif.

Reversible protein modification by small ubiquitin-like modifiers (SUMOs) is critical for eukaryotic life. Mass spectrometry-based proteomics has proven effective at identifying hundreds of potential SUMO target proteins. However, direct identification of SUMO acceptor lysines in complex samples by mass spectrometry is still very challenging. We have developed a generic method for the identification of SUMO acceptor lysines in target proteins. We have identified 103 SUMO-2 acceptor lysines in endogenous target proteins. Of these acceptor lysines, 76 are situated in the SUMOylation consensus site [VILMFPC]KxE. Interestingly, eight sites fit the inverted SUMOylation consensus motif [ED]xK[VILFP]. In addition, we found direct mass spectrometric evidence for crosstalk between SUMOylation and phosphorylation with a preferred spacer between the SUMOylated lysine and the phosphorylated serine of four residues. In 16 proteins we identified a hydrophobic cluster SUMOylation motif (HCSM). SUMO conjugation of RanGAP1 and ZBTB1 via HCSMs is remarkably efficient.

Uncovering a Dual Regulatory Role for Caspases During Endoplasmic Reticulum Stress-induced Cell Death.

Many diseases are associated with endoplasmic reticulum (ER) stress, which results from an accumulation of misfolded proteins. This triggers an adaptive response called the "unfolded protein response" (UPR), and prolonged exposure to ER stress leads to cell death. Caspases are reported to play a critical role in ER stress-induced cell death but the underlying mechanisms by which they exert their effect continue to remain elusive. To understand the role caspases play during ER stress, a systems level approach integrating analysis of the transcriptome, proteome, and proteolytic substrate profile was employed. This quantitative analysis revealed transcriptional profiles for most human genes, provided information on protein abundance for 4476 proteins, and identified 445 caspase substrates. Based on these data sets many caspase substrates were shown to be downregulated at the protein level during ER stress suggesting caspase activity inhibits their cellular function. Additionally, RNA sequencing revealed a role for caspases in regulation of ER stress-induced transcriptional pathways and gene set enrichment analysis showed expression of multiple gene targets of essential transcription factors to be upregulated during ER stress upon inhibition of caspases. Furthermore, these transcription factors were degraded in a caspase-dependent manner during ER stress. These results indicate that caspases play a dual role in regulating the cellular response to ER stress through both post-translational and transcriptional regulatory mechanisms. Moreover, this study provides unique insight into progression of the unfolded protein response into cell death, which may help identify therapeutic strategies to treat ER stress-related diseases.

Cellular Interrogation: Exploiting Cell-to-Cell Variability to Discriminate Regulatory Mechanisms in Oscillatory Signalling.

The molecular complexity within a cell may be seen as an evolutionary response to the external complexity of the cell's environment. This suggests that the external environment may be harnessed to interrogate the cell's internal molecular architecture. Cells, however, are not only nonlinear and non-stationary, but also exhibit heterogeneous responses within a clonal, isogenic population. In effect, each cell undertakes its own experiment. Here, we develop a method of cellular interrogation using programmable microfluidic devices which exploits the additional information present in cell-to-cell variation, without requiring model parameters to be fitted to data. We focussed on Ca2+ signalling in response to hormone stimulation, which exhibits oscillatory spiking in many cell types and chose eight models of Ca2+ signalling networks which exhibit similar behaviour in simulation. We developed a nonlinear frequency analysis for non-stationary responses, which could classify models into groups under parameter variation, but found that this question alone was unable to distinguish critical feedback loops. We further developed a nonlinear amplitude analysis and found that the combination of both questions ruled out six of the models as inconsistent with the experimentally-observed dynamics and heterogeneity. The two models that survived the double interrogation were mathematically different but schematically identical and yielded the same unexpected predictions that we confirmed experimentally. Further analysis showed that subtle mathematical details can markedly influence non-stationary responses under parameter variation, emphasising the difficulty of finding a "correct" model. By developing questions for the pathway being studied, and designing more versatile microfluidics, cellular interrogation holds promise as a systematic strategy that can complement direct intervention by genetics or pharmacology.

Quantitative phosphoproteomic analysis of the PI3K-regulated signaling network.

The PI3K pathway is commonly activated in cancer. Only a few studies have attempted to explore the spectrum of phosphorylation signaling downstream of the PI3K cascade. Such insight, however, is imperative to understand the mechanisms responsible for oncogenic phenotypes. By applying MS-based phosphoproteomics, we mapped 2509 phosphorylation sites on 1096 proteins, and quantified their responses to activation or inhibition of PIK3CA using isogenic knock-in derivatives and a series of targeted inhibitors. We uncovered phosphorylation changes in a wide variety of proteins involved in cell growth and proliferation, many of which have not been previously associated with PI3K signaling. A significant update of the posttranslational modification database PHOSIDA (http://www.phosida.com) allows efficient use of the data. All MS data have been deposited in the ProteomeXchange with identifier PXD003899 (http://proteomecentral.proteomexchange.org/dataset/PXD003899).

Phosphoproteome analysis of the MAPK pathway reveals previously undetected feedback mechanisms.

The RAS-RAF-MEK-ERK (MAPK) pathway is prevalently perturbed in cancer. Recent large-scale sequencing initiatives profiled thousands of tumors providing insight into alterations at the DNA and RNA levels. These efforts confirmed that key nodes of the MAPK pathway, in particular KRAS and BRAF, are among the most frequently altered proteins in cancer. The establishment of targeted therapies, however, has proven difficult. To decipher the underlying challenges, it is essential to decrypt the phosphorylation network spanned by the MAPK core axis. Using mass spectrometry we identified 2241 phosphorylation sites on 1020 proteins, and measured their responses to inhibition of MEK or ERK. Multiple phosphorylation patterns revealed previously undetected feedback, as upstream signaling nodes, including receptor kinases, showed changes at the phosphorylation level. We provide a dataset rich in potential therapeutic targets downstream of the MAPK cascade. By integrating TCGA (The Cancer Genome Atlas) data, we highlight some downstream phosphoproteins that are frequently altered in cancer. All MS data have been deposited in the ProteomeXchange with identifier PXD003908 (http://proteomecentral.proteomexchange.org/dataset/PXD003908).

Kinase-selective enrichment enables quantitative phosphoproteomics of the kinome across the cell cycle.

Protein kinases are pivotal regulators of cell signaling that modulate each other's functions and activities through site-specific phosphorylation events. These key regulatory modifications have not been studied comprehensively, because low cellular abundance of kinases has resulted in their underrepresentation in previous phosphoproteome studies. Here, we combine kinase-selective affinity purification with quantitative mass spectrometry to analyze the cell-cycle regulation of protein kinases. This proteomics approach enabled us to quantify 219 protein kinases from S and M phase-arrested human cancer cells. We identified more than 1000 phosphorylation sites on protein kinases. Intriguingly, half of all kinase phosphopeptides were upregulated in mitosis. Our data reveal numerous unknown M phase-induced phosphorylation sites on kinases with established mitotic functions. We also find potential phosphorylation networks involving many protein kinases not previously implicated in mitotic progression. These results provide a vastly extended knowledge base for functional studies on kinases and their regulation through site-specific phosphorylation.

Bioinformatics analysis of thousands of TCGA tumors to determine the involvement of epigenetic regulators in human cancer.

BACKGROUND: Many cancer cells show distorted epigenetic landscapes. The Cancer Genome Atlas (TCGA) project profiles thousands of tumors, allowing the discovery of somatic alterations in the epigenetic machinery and the identification of potential cancer drivers among members of epigenetic protein families. METHODS: We integrated mutation, expression, and copy number data from 5943 tumors from 13 cancer types to train a classification model that predicts the likelihood of being an oncogene (OG), tumor suppressor (TSG) or neutral gene (NG). We applied this predictor to epigenetic regulator genes (ERGs), and used differential expression and correlation network analysis to identify dysregulated ERGs along with co-expressed cancer genes. Furthermore, we quantified global proteomic changes by mass spectrometry after EZH2 inhibition. RESULTS: Mutation-based classifiers uncovered the OG-like profile of DNMT3A and TSG-like profiles for several ERGs. Differential gene expression and correlation network analyses revealed that EZH2 is the most significantly over-expressed ERG in cancer and is co-regulated with a cell cycle network. Proteomic analysis showed that EZH2 inhibition induced down-regulation of cell cycle regulators in lymphoma cells. CONCLUSIONS: Using classical driver genes to train an OG/TSG predictor, we determined the most predictive features at the gene level. Our predictor uncovered one OG and several TSGs among ERGs. Expression analyses elucidated multiple dysregulated ERGs including EZH2 as member of a co-expressed cell cycle network.

Genome and transcriptome sequencing of lung cancers reveal diverse mutational and splicing events.

Lung cancer is a highly heterogeneous disease in terms of both underlying genetic lesions and response to therapeutic treatments. We performed deep whole-genome sequencing and transcriptome sequencing on 19 lung cancer cell lines and three lung tumor/normal pairs. Overall, our data show that cell line models exhibit similar mutation spectra to human tumor samples. Smoker and never-smoker cancer samples exhibit distinguishable patterns of mutations. A number of epigenetic regulators, including KDM6A, ASH1L, SMARCA4, and ATAD2, are frequently altered by mutations or copy number changes. A systematic survey of splice-site mutations identified 106 splice site mutations associated with cancer specific aberrant splicing, including mutations in several known cancer-related genes. RAC1b, an isoform of the RAC1 GTPase that includes one additional exon, was found to be preferentially up-regulated in lung cancer. We further show that its expression is significantly associated with sensitivity to a MAP2K (MEK) inhibitor PD-0325901. Taken together, these data present a comprehensive genomic landscape of a large number of lung cancer samples and further demonstrate that cancer-specific alternative splicing is a widespread phenomenon that has potential utility as therapeutic biomarkers. The detailed characterizations of the lung cancer cell lines also provide genomic context to the vast amount of experimental data gathered for these lines over the decades, and represent highly valuable resources for cancer biology.

PAK1 mediates pancreatic cancer cell migration and resistance to MET inhibition.

Pancreatic adenocarcinoma (PDAC) is a major unmet medical need and a deeper understanding of molecular drivers is needed to advance therapeutic options for patients. We report here that p21-activated kinase 1 (PAK1) is a central node in PDAC cells downstream of multiple growth factor signalling pathways, including hepatocyte growth factor (HGF) and MET receptor tyrosine kinase. PAK1 inhibition blocks signalling to cytoskeletal effectors and tumour cell motility driven by HGF/MET. MET antagonists, such as onartuzumab and crizotinib, are currently in clinical development. Given that even highly effective therapies have resistance mechanisms, we show that combination with PAK1 inhibition overcomes potential resistance mechanisms mediated either by activation of parallel growth factor pathways or by direct amplification of PAK1. Inhibition of PAK1 attenuated in vivo tumour growth and metastasis in a model of pancreatic adenocarcinoma. In human tissues, PAK1 is highly expressed in a proportion of PDACs (33% IHC score 2 or 3; n = 304) and its expression is significantly associated with MET positivity (p < 0.0001) and linked to a widespread metastatic pattern in patients (p = 0.067). Taken together, our results provide evidence for a functional role of MET/PAK1 signalling in pancreatic adenocarcinoma and support further characterization of therapeutic inhibitors in this indication.

Systems-wide analysis of K-Ras, Cdc42, and PAK4 signaling by quantitative phosphoproteomics.

Although K-Ras, Cdc42, and PAK4 signaling are commonly deregulated in cancer, only a few studies have sought to comprehensively examine the spectrum of phosphorylation-mediated signaling downstream of each of these key signaling nodes. In this study, we completed a label-free quantitative analysis of oncogenic K-Ras, activated Cdc42, and PAK4-mediated phosphorylation signaling, and report relative quantitation of 2152 phosphorylated peptides on 1062 proteins. We define the overlap in phosphopeptides regulated by K-Ras, Cdc42, and PAK4, and find that perturbation of these signaling components affects phosphoproteins associated with microtubule depolymerization, cytoskeletal organization, and the cell cycle. These findings provide a resource for future studies to characterize novel targets of oncogenic K-Ras signaling and validate biomarkers of PAK4 inhibition.

Phosphoproteome of Pristionchus pacificus provides insights into architecture of signaling networks in nematode models.

Pristionchus pacificus is a nematode that is increasingly used as a model organism in evolutionary biology. The genome of P. pacificus differs markedly from that of C. elegans, with a high number of orphan genes that are restricted to P. pacificus and have no homologs in other species. To gain insight into the architecture of signal transduction networks in model nematodes, we performed a large-scale qualitative phosphoproteome analysis of P. pacificus. Using two-stage enrichment of phosphopeptides from a digest of P. pacificus proteins and their subsequent analysis via high accuracy MS, we detected and localized 6,809 phosphorylation events on 2,508 proteins. We compared the detected P. pacificus phosphoproteome to the recently published phosphoproteome of C. elegans. The overall numbers and functional classes of phosphoproteins were similar between the two organisms. Interestingly, the products of orphan genes were significantly underrepresented among the detected P. pacificus phosphoproteins. We defined the theoretical kinome of P. pacificus and compared it to that of C. elegans. While tyrosine kinases were slightly underrepresented in the kinome of P. pacificus, all major classes of kinases were present in both organisms. Application of our kinome annotation to a recent transcriptomic study of dauer and mixed stage populations showed that Ser/Thr and Tyr kinases show similar expression levels in P. pacificus but not in C. elegans. This study presents the first systematic comparison of phosphoproteomes and kinomes of two model nematodes and, as such, will be a useful resource for comparative studies of their signal transduction networks.

Assessment of computational methods for predicting the effects of missense mutations in human cancers.

BACKGROUND: Recent advances in sequencing technologies have greatly increased the identification of mutations in cancer genomes. However, it remains a significant challenge to identify cancer-driving mutations, since most observed missense changes are neutral passenger mutations. Various computational methods have been developed to predict the effects of amino acid substitutions on protein function and classify mutations as deleterious or benign. These include approaches that rely on evolutionary conservation, structural constraints, or physicochemical attributes of amino acid substitutions. Here we review existing methods and further examine eight tools: SIFT, PolyPhen2, Condel, CHASM, mCluster, logRE, SNAP, and MutationAssessor, with respect to their coverage, accuracy, availability and dependence on other tools. RESULTS: Single nucleotide polymorphisms with high minor allele frequencies were used as a negative (neutral) set for testing, and recurrent mutations from the COSMIC database as well as novel recurrent somatic mutations identified in very recent cancer studies were used as positive (non-neutral) sets. Conservation-based methods generally had moderately high accuracy in distinguishing neutral from deleterious mutations, whereas the performance of machine learning based predictors with comprehensive feature spaces varied between assessments using different positive sets. MutationAssessor consistently provided the highest accuracies. For certain combinations metapredictors slightly improved the performance of included individual methods, but did not outperform MutationAssessor as stand-alone tool. CONCLUSIONS: Our independent assessment of existing tools reveals various performance disparities. Cancer-trained methods did not improve upon more general predictors. No method or combination of methods exceeds 81% accuracy, indicating there is still significant room for improvement for driver mutation prediction, and perhaps more sophisticated feature integration is needed to develop a more robust tool.

Proteus: a web-based, context-specific modelling tool for molecular networks.

SUMMARY: Molecular networks are often studied in diverse cellular or experimental contexts, with highly context-specific details. Modelling introduces further choices as to levels of mathematical description. The resulting possibilities are difficult to explore rapidly, hampering the integration of modelling and experiment. We have developed Proteus, a web-based, context-specific tool for building compartmentalized, ordinary differential equation (ODE) models. It is inspired by the idea of a molecular 'toolkit' for Ca(2+) signalling. Toolkits in Proteus are context-independent representations of biological systems as sets of components, which may correspond to mechanisms of differing levels of complexity. Users pick and choose components from a toolkit and, for each component, pick and choose from different mechanisms, each of which describes a different instantiation of the component's mechanism. Proteus combines these choices into a system of ODEs, which may then be downloaded in SBML (Systems Biology Markup Language), Matlab or Fortran format and independently analyzed. Toolkits, components and mechanisms are user-constructible, either de novo or by cannibalizing existing models, including all those in the Biomodels database. A wide variety of context-specific models may thereby be rapidly built, modified and explored. AVAILABILITY AND IMPLEMENTATION: Proteus, implemented in C#, and a prototype toolkit for modelling calcium signalling are freely and universally available at www.modularmodeling.com CONTACT: gnad.florian@gene.com; jeremy@hms.harvard.edu SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.