MCAM: multiple clustering analysis methodology for deriving hypotheses and insights from high-throughput proteomic datasets

Advances in proteomic technologies continue to substantially accelerate capability for generating experimental data on protein levels, states, and activities in biological samples. For example, studies on receptor tyrosine kinase signaling networks can now capture the phosphorylation state of hundreds to thousands of proteins across multiple conditions. However, little is known about the function of many of these protein modifications, or the enzymes responsible for modifying them. To address this challenge, we have developed an approach that enhances the power of clustering techniques to infer functional and regulatory meaning of protein states in cell signaling networks. We have created a new computational framework for applying clustering to biological data in order to overcome the typical dependence on specific a priori assumptions and expert knowledge concerning the technical aspects of clustering. Multiple clustering analysis methodology ('MCAM') employs an array of diverse data transformations, distance metrics, set sizes, and clustering algorithms, in a combinatorial fashion, to create a suite of clustering sets. These sets are then evaluated based on their ability to produce biological insights through statistical enrichment of metadata relating to knowledge concerning protein functions, kinase substrates, and sequence motifs. We applied MCAM to a set of dynamic phosphorylation measurements of the ERRB network to explore the relationships between algorithmic parameters and the biological meaning that could be inferred and report on interesting biological predictions. Further, we applied MCAM to multiple phosphoproteomic datasets for the ERBB network, which allowed us to compare independent and incomplete overlapping measurements of phosphorylation sites in the network. We report specific and global differences of the ERBB network stimulated with different ligands and with changes in HER2 expression. Overall, we offer MCAM as a broadly-applicable approach for analysis of proteomic data which may help increase the current understanding of molecular networks in a variety of biological problems.     

Exploring biological processes involved in embryonic stem cell differentiation by analyzing proteomic data

Since, proteins carry out many functional roles in a cell with different concentrations, proteomics is likely a more appropriate approach to explain biological processes and cellular events than mRNA studies. Although, gene ontology provides a systematic and organized vocabulary of biological terms for proteins, we need more details to decide about the correct duty and annotation of proteins in a specific condition. One can assume that a change of protein concentration is related to a biological process of that protein with negligible error. Therefore, we can obtain more information about the function of proteins by studying these profiles. In this study, we used time-course proteomic data of a twenty day differentiation study of embryonic stem cells (ESCs) differentiating to embryoid bodies (EBs). Hierarchical clustering was used to cluster time-series concentration profile of proteins. Our results demonstrate that there are eleven active processes with distinct concentration profiles in this initial differentiation. According to the concentration profiles of proteins, we suggest new change points (or equivalently, new stages) in the course of embryonic differentiation.     

Large scale protein profiling by combination of protein fractionation and multidimensional protein identification technology (MudPIT)

In the past decade, shotgun proteomic analysis has been utilized extensively to answer complex biological questions. New challenges arise in large scale proteomic profiling when dealing with complex biological mixtures such as the mammalian cell lysate. In this study, we explored the approach of protein separation prior to the shotgun multidimensional protein identification technology (MudPIT) analysis. We fractionated the mammalian cancer cell lysate using the PF 2D ProteomeLab system and analyzed the distribution of molecular weight, isoelectric point, and cellular localization of the eluted proteins. As a result, we were able to reduce sample complexity by protein fractionation and increase the possibility of detecting proteins with lower abundance in the complex protein mixture.     

Challenges and solutions in proteomics

The accelerated growth of proteomics data presents both opportunities and challenges. Large-scale proteomic profiling of biological samples such as cells, organelles or biological fluids has led to discovery of numerous key and novel proteins involved in many biological/disease processes including cancers, as well as to the identification of novel disease biomarkers and potential therapeutic targets. While proteomic data analysis has been greatly assisted by the many bioinformatics tools developed in recent years, a careful analysis of the major steps and flow of data in a typical highthroughput analysis reveals a few gaps that still need to be filled to fully realize the value of the data. To facilitate functional and pathway discovery for large-scale proteomic data, we have developed an integrated proteomic expression analysis system, iProXpress, which facilitates protein identification using a comprehensive sequence library and functional interpretation using integrated data. With its modular design, iProXpress complements and can be integrated with other software in a proteomic data analysis pipeline. This novel approach to complex biological questions involves the interrogation of multiple data sources, thereby facilitating hypothesis generation and knowledge discovery from the genomic-scale studies and fostering disease diagnosis and drug development.     

Protein Neighbors and Proximity Proteomics

Within cells, proteins can co-assemble into functionally integrated and spatially restricted multicomponent complexes. Often, the affinities between individual proteins are relatively weak, and proteins within such clusters may interact only indirectly with many of their other protein neighbors. This makes proteomic characterization difficult using methods such as immunoprecipitation or cross-linking. Recently, several groups have described the use of enzyme-catalyzed proximity labeling reagents that covalently tag the neighbors of a targeted protein with a small molecule such as fluorescein or biotin. The modified proteins can then be isolated by standard pulldown methods and identified by mass spectrometry. Here we will describe the techniques as well as their similarities and differences. We discuss their applications both to study protein assemblies and to provide a new way for characterizing organelle proteomes. We stress the importance of proteomic quantitation and independent target validation in such experiments. Furthermore, we suggest that there are biophysical and cell-biological principles that dictate the appropriateness of enzyme-catalyzed proximity labeling methods to address particular biological questions of interest.     

Interaction profiling methods to map protein and pathway targets of bioactive ligands

Recent advances in -omic profiling technologies have ushered in an era where we no longer want to merely measure the presence or absence of a biomolecule of interest, but instead hope to understand its function and interactions within larger signaling networks. Here, we review several emerging proteomic technologies capable of detecting protein interaction networks in live cells and their integration to draft holistic maps of proteins that respond to diverse stimuli, including bioactive small molecules. Moreover, we provide a conceptual framework to combine so-called ‘top-down’ and ‘bottom-up’ interaction profiling methods and ensuing proteomic profiles to directly identify binding targets of small molecule ligands, as well as for unbiased discovery of proteins and pathways that may be directly bound or influenced by those first responders. The integrated, interaction-based profiling methods discussed here have the potential to provide a unique and dynamic view into cellular signaling networks for both basic and translational biological studies.     

Proteome profiling reveals potential toxicity and detoxification pathways following exposure of BEAS-2B cells to engineered nanoparticle titanium dioxide

Oxidative stress is known to play important roles in engineered nanomaterial‐induced cellular toxicity. However, the proteins and signaling pathways associated with the engineered nanomaterial‐mediated oxidative stress and toxicity are largely unknown. To identify these toxicity pathways and networks that are associated with exposure to engineered nanomaterials, an integrated proteomic study was conducted using human bronchial epithelial cells, BEAS‐2B and nanoscale titanium dioxide. Utilizing 2‐DE and MS, we identified 46 proteins that were altered at protein expression levels. The protein changes detected by 2‐DE/MS were verified by functional protein assays. These identified proteins include some key proteins involved in cellular stress response, metabolism, adhesion, cytoskeletal dynamics, cell growth, cell death, and cell signaling. The differentially expressed proteins were mapped using Ingenuity Pathway Analyses^? canonical pathways and Ingenuity Pathway Analyses tox lists to create protein‐interacting networks and proteomic pathways. Twenty protein canonical pathways and tox lists were generated, and these pathways were compared to signaling pathways generated from genomic analyses of BEAS‐2B cells treated with titanium dioxide. There was a significant overlap in the specific pathways and lists generated from the proteomic and the genomic data. In addition, we also analyzed the phosphorylation profiles of protein kinases in titanium dioxide‐treated BEAS‐2B cells for a better understanding of upstream signaling pathways in response to the titanium dioxide treatment and the induced oxidative stress. In summary, the present study provides the first protein‐interacting network maps and novel insights into the biological responses and potential toxicity and detoxification pathways of titanium dioxide.     

1-DE MS and 2-D LC-MS analysis of the mouse bronchoalveolar lavage proteome

Bronchoalveolar lavage fluid (BALF) is a complex mixture of proteins, which represents a unique clinically useful sampling of the lower respiratory tract. Many proteomic technologies can be used to characterize complex biological mixtures; however, it is not yet clear which technology(s) provide more information regarding the number of proteins identified and sequence coverage. In this study, we initially compared two common proteomic approaches, 2-D LC microESI MS/MS and 1-DE followed by gel slice digestion, peptide extraction and peptide identification by MS in characterization of the mouse BALF proteome; secondly, we identified 297 unique proteins from the mouse BALF proteome, greatly expanded the BALF proteome by about threefold regardless of species.     

On the functional and structural characterization of hubs in protein–protein interaction networks

A number of interesting issues have been addressed on biological networks about their global and local properties. The connection between the topological properties of proteins in Protein–Protein Interaction (PPI) networks and their biological relevance has been investigated focusing on hubs, i.e. proteins with a large number of interacting partners. We will survey the literature trying to answer the following questions: Do hub proteins have special biological properties? Do they tend to be more essential than non-hub proteins? Are they more evolutionarily conserved? Do they play a central role in modular organization of the protein interaction network? Are there structural properties that characterize hub proteins?     

LC/MS-Based Quantitative Proteomic Analysis of Paraffin-Embedded Archival Melanomas Reveals Potential Proteomic Biomarkers Associated with Metastasis

Background

Melanoma metastasis status is highly associated with the overall survival of patients; yet, little is known about proteomic changes during melanoma tumor progression. To better understand the changes in protein expression involved in melanoma progression and metastasis, and to identify potential biomarkers, we conducted a global quantitative proteomic analysis on archival metastatic and primary melanomas.

Methodology and Findings

A total of 16 metastatic and 8 primary cutaneous melanomas were assessed. Proteins were extracted from laser captured microdissected formalin fixed paraffin-embedded archival tissues by liquefying tissue cells. These preparations were analyzed by a LC/MS-based label-free protein quantification method. More than 1500 proteins were identified in the tissue lysates with a peptide ID confidence level of >75%. This approach identified 120 significant changes in protein levels. These proteins were identified from multiple peptides with high confidence identification and were expressed at significantly different levels in metastases as compared with primary melanomas (q-Value<0.05).

Conclusions and Significance

The differentially expressed proteins were classified by biological process or mapped into biological system networks, and several proteins were implicated by these analyses as cancer- or metastasis-related. These proteins represent potential biomarkers for tumor progression. The study successfully identified proteins that are differentially expressed in formalin fixed paraffin-embedded specimens of metastatic and primary melanoma.     

Reproducible quantitative proteotype data matrices for systems biology

Historically, many mass spectrometry–based proteomic studies have aimed at compiling an inventory of protein compounds present in a biological sample, with the long-term objective of creating a proteome map of a species. However, to answer fundamental questions about the behavior of biological systems at the protein level, accurate and unbiased quantitative data are required in addition to a list of all protein components. Fueled by advances in mass spectrometry, the proteomics field has thus recently shifted focus toward the reproducible quantification of proteins across a large number of biological samples. This provides the foundation to move away from pure enumeration of identified proteins toward quantitative matrices of many proteins measured across multiple samples. It is argued here that data matrices consisting of highly reproducible, quantitative, and unbiased proteomic measurements across a high number of conditions, referred to here as quantitative proteotype maps, will become the fundamental currency in the field and provide the starting point for downstream biological analysis. Such proteotype data matrices, for example, are generated by the measurement of large patient cohorts, time series, or multiple experimental perturbations. They are expected to have a large effect on systems biology and personalized medicine approaches that investigate the dynamic behavior of biological systems across multiple perturbations, time points, and individuals.     

Network-based auto-probit modeling for protein function prediction

Predicting the functional roles of proteins based on various genome-wide data, such as protein-protein association networks, has become a canonical problem in computational biology. Approaching this task as a binary classification problem, we develop a network-based extension of the spatial auto-probit model. In particular, we develop a hierarchical Bayesian probit-based framework for modeling binary network-indexed processes, with a latent multivariate conditional autoregressive Gaussian process. The latter allows for the easy incorporation of protein-protein association network topologies-either binary or weighted-in modeling protein functional similarity. We use this framework to predict protein functions, for functions defined as terms in the Gene Ontology (GO) database, a popular rigorous vocabulary for biological functionality. Furthermore, we show how a natural extension of this framework can be used to model and correct for the high percentage of false negative labels in training data derived from GO, a serious shortcoming endemic to biological databases of this type. Our method performance is evaluated and compared with standard algorithms on weighted yeast protein-protein association networks, extracted from a recently developed integrative database called Search Tool for the Retrieval of INteracting Genes/proteins (STRING). Results show that our basic method is competitive with these other methods, and that the extended method-incorporating the uncertainty in negative labels among the training data-can yield nontrivial improvements in predictive accuracy.     

Experimental and computational tools useful for (re)construction of dynamic kinase–substrate networks

The explosion of site‐ and context‐specific in vivo phosphorylation events presents a potentially rich source of biological knowledge and calls for novel data analysis and modeling paradigms. Perhaps the most immediate challenge is delineating detected phosphorylation sites to their effector kinases. This is important for (re)constructing transient kinase–substrate interaction networks that are essential for mechanistic understanding of cellular behaviors and therapeutic intervention, but has largely eluded high‐throughput protein‐interaction studies due to their transient nature and strong dependencies on cellular context. Here, we surveyed some of the computational approaches developed to dissect phosphorylation data detected in systematic proteomic experiments and reviewed some experimental and computational approaches used to map phosphorylation sites to their effector kinases in efforts aimed at reconstructing biological signaling networks.     

Alternative splicing takes shape during neuronal development

The spatial and temporal control of alternative splicing is a major mechanism used to generate proteomic diversity in the brain. Microarray and Next Generation Sequencing approaches reveal mechanistic insights about networks of tissue-specific RNA binding proteins and micro RNAs that coordinate suites of alternative splicing patterns during neuronal differentiation. In the context of large-scale changes, one alternative splicing switch during embryonic brain development is crucial for neuronal migration and the laminar organization of the cerebral cortex. A major challenge to understand alternative splicing at the systems level is now being approached by the design of integrative modeling approaches that predict the combinatorial control of brain-specific exons.     

Biological networks

Recent advances in high-throughput methods have provided us with a first glimpse of the overall structure of molecular interaction networks in biological systems. Ultimately, we expect that such information will change how we think about biological systems in a fundamental way. Instead of viewing the genetic parts list of an organism as a loose collection of biochemical activities, in the best case, we anticipate discrete networks of function to bridge the gap between genotype and phenotype, and to do so in a more profound way than the current qualitative classification of linked reactions into familiar pathways, such as glycolysis and the MAPK signal transduction cascades. At the present time, however, we are still far from a complete answer to the most basic question: what can we learn about biology by studying networks? Promising steps in this direction have come from such diverse approaches as mathematical analysis of global network structure, partitioning networks into functionally related modules and motifs, and even de novo design of networks. A complete picture will probably require integrating the data obtained from all of these approaches with modeling efforts at many different levels of detail.     

Examples of Mathematical Modeling: Tales from the Crypt

Mathematical modeling is being increasingly recognized within the biomedical sciences as an important tool that can aid the understanding of biological systems. The heavily regulated cell renewal cycle in the colonic crypt provides a good example of how modeling can be used to find out key features of the system kinetics, and help to explain both the breakdown of homeostasis and the initiation of tumorigenesis.We use the cell population model by Johnston et al. (2007) Proc. Natl. Acad. Sci. USA 104, 4008-4013, to illustrate the power of mathematical modeling by considering two key questions about the cell population dynamics in the colonic crypt. We ask: how can a model describe both homeostasis and unregulated growth in tumorigenesis; and to which parameters in the system is the model most sensitive? In order to address these questions, we discuss what type of modeling approach is most appropriate in the crypt.We use the model to argue why tumorigenesis is observed to occur in stages with long lag phases between periods of rapid growth, and we identify the key parameters.     

From proteomics toward systems biology: integration of different types of proteomics data into network models

Living organisms are comprised of various systems at different levels, i.e., organs, tissues, and cells. Each system carries out its diverse functions in response to environmental and genetic perturbations, by utilizing biological networks, in which nodal components, such as, DNA, mRNAs, proteins, and metabolites, closely interact with each other. Systems biology investigates such systems by producing comprehensive global data that represent different levels of biological information, i.e., at the DNA, mRNA, protein, or metabolite levels, and by integrating this data into network models that generate coherent hypotheses for given biological situations. This review presents a systems biology framework, called the 'Integrative Proteomics Data Analysis Pipeline' (IPDAP), which generates mechanistic hypotheses from network models reconstructed by integrating diverse types of proteomic data generated by mass spectrometry-based proteomic analyses. The devised framework includes a serial set of computational and network analysis tools. Here, we demonstrate its functionalities by applying these tools to several conceptual examples.