Image-based classification of histology sections plays an important role in predicting clinical outcomes. However this task is very challenging due to the presence of large technical variations (e.g., fixation, staining) and biological heterogeneities (e.g., cell type, cell state). In the field of biomedical imaging, for the purposes of visualization and/or quantification, different stains are typically used for different targets of interest (e.g., cellular/subcellular events), which generates multi-spectrum data (images) through various types of microscopes and, as a result, provides the possibility of learning biological-component-specific features by exploiting multispectral information. We propose a multispectral feature learning model that automatically learns a set of convolution filter banks from separate spectra to efficiently discover the intrinsic tissue morphometric signatures, based on convolutional sparse coding (CSC). The learned feature representations are then aggregated through the spatial pyramid matching framework (SPM) and finally classified using a linear SVM. The proposed system has been evaluated using two large-scale tumor cohorts, collected from The Cancer Genome Atlas (TCGA). Experimental results show that the proposed model 1) outperforms systems utilizing sparse coding for unsupervised feature learning (e.g., PSD-SPM [5]); 2) is competitive with systems built upon features with biological prior knowledge (e.g., SMLSPM [4]).

Image-based classification of tissue histology, in terms of different components (e.g., normal signature, categories of aberrant signatures), provides a series of indices for tumor composition. Subsequently, aggregation of these indices in each whole slide image (WSI) from a large cohort can provide predictive models of clinical outcome. However, the performance of the existing techniques is hindered as a result of large technical and biological variations that are always present in a large cohort. In this paper, we propose two algorithms for classification of tissue histology based on robust representations of morphometric context, which are built upon nuclear level morphometric features at various locations and scales within the spatial pyramid matching (SPM) framework. These methods have been evaluated on two distinct datasets of different tumor types collected from The Cancer Genome Atlas (TCGA), and the experimental results indicate that our methods are (i) extensible to different tumor types; (ii) robust in the presence of wide technical and biological variations; (iii) invariant to different nuclear segmentation strategies; and (iv) scalable with varying training sample size. In addition, our experiments suggest that enforcing sparsity, during the construction of morphometric context, further improves the performance of the system.

Image-based classification of tissue histology, in terms of distinct histopathology (e.g., tumor or necrosis regions), provides a series of indices for tumor composition. Furthermore, aggregation of these indices from each whole slide image (WSI) in a large cohort can provide predictive models of clinical outcome. However, the performance of the existing techniques is hindered as a result of large technical variations (e.g., fixation, staining) and biological heterogeneities (e.g., cell type, cell state) that are always present in a large cohort. We suggest that, compared with human engineered features widely adopted in existing systems, unsupervised feature learning is more tolerant to batch effect (e.g., technical variations associated with sample preparation) and pertinent features can be learned without user intervention. This leads to a novel approach for classification of tissue histology based on unsupervised feature learning and spatial pyramid matching (SPM), which utilize sparse tissue morphometric signatures at various locations and scales. This approach has been evaluated on two distinct datasets consisting of different tumor types collected from The Cancer Genome Atlas (TCGA), and the experimental results indicate that the proposed approach is (i) extensible to different tumor types; (ii) robust in the presence of wide technical variations and biological heterogeneities; and (iii) scalable with varying training sample sizes.

MOTIVATION: Tumors acquire many chromosomal amplifications, and those acquired early in the lifespan of the tumor may be not only important for tumor growth but also can be used for diagnostic purposes. Many methods infer the order of the accumulation of abnormalities based on their occurrence in a large cohort of patients. Recently, Durinck et al. (2011) and Greenman et al. (2012) developed methods to order a single tumor's chromosomal amplifications based on the patterns of mutations accumulated within those regions. This method offers an unprecedented opportunity to assess the etiology of a single tumor sample, but has not been widely evaluated. RESULTS: We show that the model for timing chromosomal amplifications is limited in scope, particularly for regions with high levels of amplification. We also show that the estimation of the order of events can be sensitive for events that occur early in the progression of the tumor and that the partial maximum likelihood method of Greenman et al. (2012) can give biased estimates, particularly for moderate read coverage or normal contamination. We propose a maximum-likelihood estimation procedure that fully accounts for sequencing variability and show that it outperforms the partial maximum-likelihood estimation method. We also propose a Bayesian estimation procedure that stabilizes the estimates in certain settings. We implement these methods on a small number of ovarian tumors, and the results suggest possible differences in how the tumors acquired amplifications. AVAILABILITY AND IMPLEMENTATION: We provide implementation of these methods in an R package cancerTiming, which is available from the Comprehensive R Archive Network (CRAN) at http://CRAN.R-project.org/.

Automated analysis of whole mount tissue sections can provide insights into tumor subtypes and the underlying molecular basis of neoplasm. However, since tumor sections are collected from different laboratories, inherent technical and biological variations impede analysis for very large datasets such as The Cancer Genome Atlas (TCGA). Our objective is to characterize tumor histopathology, through the delineation of the nuclear regions, from hematoxylin and eosin (H&E) stained tissue sections. Such a representation can then be mined for intrinsic subtypes across a large dataset for prediction and molecular association. Furthermore, nuclear segmentation is formulated within a multi-reference graph framework with geodesic constraints, which enables computation of multidimensional representations, on a cell-by-cell basis, for functional enrichment and bioinformatics analysis. Here, we present a novel method, multi-reference graph cut (MRGC), for nuclear segmentation that overcomes technical variations associated with sample preparation by incorporating prior knowledge from manually annotated reference images and local image features. The proposed approach has been validated on manually annotated samples and then applied to a dataset of 377 Glioblastoma Multiforme (GBM) whole slide images from 146 patients. For the GBM cohort, multidimensional representation of the nuclear features and their organization have identified 1) statistically significant subtypes based on several morphometric indexes, 2) whether each subtype can be predictive or not, and 3) that the molecular correlates of predictive subtypes are consistent with the literature. Data and intermediaries for a number of tumor types (GBM, low grade glial, and kidney renal clear carcinoma) are available at: http://tcga.lbl.gov for correlation with TCGA molecular data. The website also provides an interface for panning and zooming of whole mount tissue sections with/without overlaid segmentation results for quality control.

MOTIVATION: Protein signaling networks play a key role in cellular function, and their dysregulation is central to many diseases, including cancer. To shed light on signaling network topology in specific contexts, such as cancer, requires interrogation of multiple proteins through time and statistical approaches to make inferences regarding network structure. RESULTS: In this study, we use dynamic Bayesian networks to make inferences regarding network structure and thereby generate testable hypotheses. We incorporate existing biology using informative network priors, weighted objectively by an empirical Bayes approach, and exploit a connection between variable selection and network inference to enable exact calculation of posterior probabilities of interest. The approach is computationally efficient and essentially free of user-set tuning parameters. Results on data where the true, underlying network is known place the approach favorably relative to existing approaches. We apply these methods to reverse-phase protein array time-course data from a breast cancer cell line (MDA-MB-468) to predict signaling links that we independently validate using targeted inhibition. The methods proposed offer a general approach by which to elucidate molecular networks specific to biological context, including, but not limited to, human cancers. AVAILABILITY: http://mukherjeelab.nki.nl/DBN (code and data).

BACKGROUND: An important question in the analysis of biochemical data is that of identifying subsets of molecular variables that may jointly influence a biological response. Statistical variable selection methods have been widely used for this purpose. In many settings, it may be important to incorporate ancillary biological information concerning the variables of interest. Pathway and network maps are one example of a source of such information. However, although ancillary information is increasingly available, it is not always clear how it should be used nor how it should be weighted in relation to primary data. RESULTS: We put forward an approach in which biological knowledge is incorporated using informative prior distributions over variable subsets, with prior information selected and weighted in an automated, objective manner using an empirical Bayes formulation. We employ continuous, linear models with interaction terms and exploit biochemically-motivated sparsity constraints to permit exact inference. We show an example of priors for pathway- and network-based information and illustrate our proposed method on both synthetic response data and by an application to cancer drug response data. Comparisons are also made to alternative Bayesian and frequentist penalised-likelihood methods for incorporating network-based information. CONCLUSIONS: The empirical Bayes method proposed here can aid prior elicitation for Bayesian variable selection studies and help to guard against mis-specification of priors. Empirical Bayes, together with the proposed pathway-based priors, results in an approach with a competitive variable selection performance. In addition, the overall procedure is fast, deterministic, and has very few user-set parameters, yet is capable of capturing interplay between molecular players. The approach presented is general and readily applicable in any setting with multiple sources of biological prior knowledge.

MOTIVATION: High-throughput techniques facilitate the simultaneous measurement of DNA copy number at hundreds of thousands of sites on a genome. Older techniques allow measurement only of total copy number, the sum of the copy number contributions from the two parental chromosomes. Newer single nucleotide polymorphism (SNP) techniques can in addition enable quantifying parent-specific copy number (PSCN). The raw data from such experiments are two-dimensional, but are unphased. Consequently, inference based on them necessitates development of new analytic methods. METHODS: We have adapted and enhanced the circular binary segmentation (CBS) algorithm for this purpose with focus on paired test and reference samples. The essence of paired parent-specific CBS (Paired PSCBS) is to utilize the original CBS algorithm to identify regions of equal total copy number and then to further segment these regions where there have been changes in PSCN. For the final set of regions, calls are made of equal parental copy number and loss of heterozygosity (LOH). PSCN estimates are computed both before and after calling. RESULTS: The methodology is evaluated by simulation and on glioblastoma data. In the simulation, PSCBS compares favorably to established methods. On the glioblastoma data, PSCBS identifies interesting genomic regions, such as copy-neutral LOH. AVAILABILITY: The Paired PSCBS method is implemented in an open-source R package named PSCBS, available on CRAN (http://cran.r-project.org/).

BACKGROUND: Our goals are to develop a computational histopathology pipeline for characterizing tumor types that are being generated by The Cancer Genome Atlas (TCGA) for genomic association. TCGA is a national collaborative program where different tumor types are being collected, and each tumor is being characterized using a variety of genome-wide platforms. Here, we have developed a tumor-centric analytical pipeline to process tissue sections stained with hematoxylin and eosin (H&E) for visualization and cell-by-cell quantitative analysis. Thus far, analysis is limited to Glioblastoma Multiforme (GBM) and kidney renal clear cell carcinoma tissue sections. The final results are being distributed for subtyping and linking the histology sections to the genomic data. RESULTS: A computational pipeline has been designed to continuously update a local image database, with limited clinical information, from an NIH repository. Each image is partitioned into blocks, where each cell in the block is characterized through a multidimensional representation (e.g., nuclear size, cellularity). A subset of morphometric indices, representing potential underlying biological processes, can then be selected for subtyping and genomic association. Simultaneously, these subtypes can also be predictive of the outcome as a result of clinical treatments. Using the cellularity index and nuclear size, the computational pipeline has revealed five subtypes, and one subtype, corresponding to the extreme high cellularity, has shown to be a predictor of survival as a result of a more aggressive therapeutic regime. Further association of this subtype with the corresponding gene expression data has identified enrichment of (i) the immune response and AP-1 signaling pathways, and (ii) IFNG, TGFB1, PKC, Cytokine, and MAPK14 hubs. CONCLUSION: While subtyping is often performed with genome-wide molecular data, we have shown that it can also be applied to categorizing histology sections. Accordingly, we have identified a subtype that is a predictor of the outcome as a result of a therapeutic regime. Computed representation has become publicly available through our Web site.

BACKGROUND: Methylation of CpG islands within the DNA promoter regions is one mechanism that leads to aberrant gene expression in cancer. In particular, the abnormal methylation of CpG islands may silence associated genes. Therefore, using high-throughput microarrays to measure CpG island methylation will lead to better understanding of tumor pathobiology and progression, while revealing potentially new biomarkers. We have examined a recently developed high-throughput technology for measuring genome-wide methylation patterns called mTACL. Here, we propose a computational pipeline for integrating gene expression and CpG island methylation profiles to identify epigenetically regulated genes for a panel of 45 breast cancer cell lines, which is widely used in the Integrative Cancer Biology Program (ICBP). The pipeline (i) reduces the dimensionality of the methylation data, (ii) associates the reduced methylation data with gene expression data, and (iii) ranks methylation-expression associations according to their epigenetic regulation. Dimensionality reduction is performed in two steps: (i) methylation sites are grouped across the genome to identify regions of interest, and (ii) methylation profiles are clustered within each region. Associations between the clustered methylation and the gene expression data sets generate candidate matches within a fixed neighborhood around each gene. Finally, the methylation-expression associations are ranked through a logistic regression, and their significance is quantified through permutation analysis. RESULTS: Our two-step dimensionality reduction compressed 90% of the original data, reducing 137,688 methylation sites to 14,505 clusters. Methylation-expression associations produced 18,312 correspondences, which were used to further analyze epigenetic regulation. Logistic regression was used to identify 58 genes from these correspondences that showed a statistically significant negative correlation between methylation profiles and gene expression in the panel of breast cancer cell lines. Subnetwork enrichment of these genes has identified 35 common regulators with 6 or more predicted markers. In addition to identifying epigenetically regulated genes, we show evidence of differentially expressed methylation patterns between the basal and luminal subtypes. CONCLUSIONS: Our results indicate that the proposed computational protocol is a viable platform for identifying epigenetically regulated genes. Our protocol has generated a list of predictors including COL1A2, TOP2A, TFF1, and VAV3, genes whose key roles in epigenetic regulation is documented in the literature. Subnetwork enrichment of these predicted markers further suggests that epigenetic regulation of individual genes occurs in a coordinated fashion and through common regulators.

MOTIVATION: The rapid expansion of whole-genome copy number (CN) studies brings a demand for increased precision and resolution of CN estimates. Recent studies have obtained CN estimates from more than one platform for the same set of samples, and it is natural to want to combine the different estimates in order to meet this demand. Estimates from different platforms show different degrees of attenuation of the true CN changes. Similar differences can be observed in CNs from the same platform run in different labs, or in the same lab, with different analytical methods. This is the reason why it is not straightforward to combine CN estimates from different sources (platforms, labs and analysis methods). RESULTS: We propose a single-sample multi source normalization that brings full-resolution CN estimates to the same scale across sources. The normalized CNs are such that for any underlying CN level, their mean level is the same regardless of the source, which make them better suited for being combined across sources, e.g. existing segmentation methods may be used to identify aberrant regions. We use microarray-based CN estimates from 'The Cancer Genome Atlas' (TCGA) project to illustrate and validate the method. We show that the normalized and combined data better separate two CN states at a given resolution. We conclude that it is possible to combine CNs from multiple sources such that the resolution becomes effectively larger, and when multiple platforms are combined, they also enhance the genome coverage by complementing each other in different regions. AVAILABILITY: A bounded-memory implementation is available in aroma.cn.

MOTIVATION: Combinatorial effects, in which several variables jointly influence an output or response, play an important role in biological systems. In many settings, Boolean functions provide a natural way to describe such influences. However, biochemical data using which we may wish to characterize such influences are usually subject to much variability. Furthermore, in high-throughput biological settings Boolean relationships of interest are very often sparse, in the sense of being embedded in an overall dataset of higher dimensionality. This motivates a need for statistical methods capable of making inferences regarding Boolean functions under conditions of noise and sparsity. RESULTS: We put forward a statistical model for sparse, noisy Boolean functions and methods for inference under the model. We focus on the case in which the form of the underlying Boolean function, as well as the number and identity of its inputs are all unknown. We present results on synthetic data and on a study of signalling proteins in cancer biology.

BACKGROUND: Biological studies involve a growing number of distinct high-throughput experiments to characterize samples of interest. There is a lack of methods to visualize these different genomic datasets in a versatile manner. In addition, genomic data analysis requires integrated visualization of experimental data along with constantly changing genomic annotation and statistical analyses. RESULTS: We developed GenomeGraphs, as an add-on software package for the statistical programming environment R, to facilitate integrated visualization of genomic datasets. GenomeGraphs uses the biomaRt package to perform on-line annotation queries to Ensembl and translates these to gene/transcript structures in viewports of the grid graphics package. This allows genomic annotation to be plotted together with experimental data. GenomeGraphs can also be used to plot custom annotation tracks in combination with different experimental data types together in one plot using the same genomic coordinate system. CONCLUSION: GenomeGraphs is a flexible and extensible software package which can be used to visualize a multitude of genomic datasets within the statistical programming environment R.

Many dynamical processes can be represented as directed attributed graphs or Petri nets where relationships between various entities are explicitly expressed. Signaling networks modeled as Petri nets are one class of such graphical modeling and representations. These networks encode how different protein in specific compartments, interact to create new protein products. Initially, the proteins and rules governing their interactions are curated from literature and then refined with experimental data. Variation in these networks occurs in topological structure, size, and weights associated on edges. Collectively, these variations are quite significant for manual and interactive analysis. Furthermore, as new information is added to these networks, the emergence of new computational models becomes paramount. From this perspective, hierarchical spectral methods are proposed and applied for inferring similarities and dissimilarities from an ensemble of graphs that corresponds to reaction networks. The technique has been implemented and tested on curated signaling networks that are derived for breast cancer cell lines

BACKGROUND: Sharing of microarray data within the research community has been greatly facilitated by the development of the disclosure and communication standards MIAME and MAGE-ML by the MGED Society. However, the complexity of the MAGE-ML format has made its use impractical for laboratories lacking dedicated bioinformatics support. RESULTS: We propose a simple tab-delimited, spreadsheet-based format, MAGE-TAB, which will become a part of the MAGE microarray data standard and can be used for annotating and communicating microarray data in a MIAME compliant fashion. CONCLUSION: MAGE-TAB will enable laboratories without bioinformatics experience or support to manage, exchange and submit well-annotated microarray data in a standard format using a spreadsheet. The MAGE-TAB format is self-contained, and does not require an understanding of MAGE-ML or XML.

Recently, the Microarray Gene Expression Data Society (MGED, www.mged.org) was awarded funds from NHGRI/NIH to continue development of MAGE, the MGED Ontology (MO) and related technologies for communication and interpretation of microarray data. This award reflects NIH's substantial commitment of resources to microarray experimentation and the recognition that these data should be preserved. In this editorial I discuss MGED's goals, show how MGED's efforts have helped the microarray community and describe MGED's future efforts in the context of the recent award. MGED's efforts are 2-fold. First, it is committed to ensure that microarray experiments are scientifically sound. Second, it is involved in building the infrastructure that will support experimental data, from millions of microarrays. This second goals allows microarray data to be treated similar to DNA sequences which are repeatedly queried, reanalyzed and amalgamated. This is obviously critical because the data are extremely valuable if they are computationally interpretable. MGED's solutions to achieve these two goals are the introduction of scientific guidelines (MIAME), data communications standards (MAGE) and biological annotationsystems (MO). Concerning the development of infrastructure for data storage, query and analysis, perhaps it is better to start from what I think MGED is not trying to do. Explicitly MGED's effort is not to build an interoperability and computational infrastructure for a few hundred or a few thousand experiments. This is the current size of the ArrayExpress and GEO data archives, which hold 700 and 2000 experiments, respectively. If a user requires information on the biology of one of these experiments it is simply a matter of reading the paper that describes the experiment. Data standards for interpretation and communication are unnecessary in this case. MGED's efforts are focused squarely on the future; a future where 100 million individual microarrays (or more) are archived and served to the research community. One hundred million publicly available microarrays is an enormous number, but I suspect it will be lower than that exists in 20 years, because as costs go down and experimental systems become more automated, microarrays will be increasingly used in biological research, just as sequencing technologies have. Both ArrayExpress and GEO have doubled in size every six months, which would mean that 100 million microarrays would be available in 2010; assuming a more modest doubling every year, 100 million microarrays will be available by 2015. As a ballpark estimate, 100 million microarrays would account for 2% of public global medical research spending in the next 10 years. In practice, MGED's efforts will become useful well before this point, typically once more data are available than a user can easily access and understand without computer assistance. For some data consumers, data surfeit has already occurred, for some, it may be imminent; whereas for others it may be several years in the future. No matter what the current strengths of any one research group may be, at some point there will be too much published microarray data for anyone to accurately translate between formats and to hand-annotate by reading research papers. What is a researcher, who is interested in viewing or analyzing data from several hundred or several thousand of these experiments, supposed to do without a common infrastructure? If each is in a different format with its own meaning, the advantage of public data sharing has been lost. Unlike DNA sequences where nearly all of the information is encoded in the data stream itself, interpretation of microarray results depend on context and the experimental variables, design and parameters all have a profound impact on what information can be extracted from the data. The result is that sharing microarray data is as much a matter of explaining the biology of the system as it is describing a table of numbers. The MGED solution to this problem has been the development of a publication standard, a communication model and an annotation system for microarray experiments. MIAME specifies what information is necessary to understand a microarray experiment. MAGE and its associated technologies provide a computational model for describing a microarray experiment. MO provides standardized descriptions of biological and experimental properties to better facilitate interpretation and comparison. These three efforts, in particular, have enormous benefits for the scientific community: MIAME, MAGE and MO provide a concrete framework for software developers to learn about the structure of microarray experiments—one that can evolve and adapt as our needs and the technology develop. MIAME, MAGE and MO have stimulated discussion of what is biologically important, how it should be modeled and how to work together with others in a productive and efficient manner. These efforts have accelerated the pace of software and database development by providing tools and standards for implementation. The funding provided by NIGMS will allow MGED to further develop MIAME, MAGE and MO as standards and work to bring them into wider acceptance and use. First, we are extending MAGE–MO in three areas that are recognized to be limiting: the organization of MAGE components that limits its application to other technologies; incomplete incorporation of the MO; and limited ability to communicate analysis and interpretation beyond the data. Second, we will provide better maintenance of the MAGE software toolkit (MAGEstk), to ensure that the software is kept up-to-date, to respond to feature requests and to develop new technologies, such as the ability to efficiently accommodate very large datasets. Third, we are building the MO as an integrated extension of the MAGE–OM. Full details of our efforts can be found on our website www.mged.org and are routinely provided at our annual meetings. Members of MGED have committed their efforts to ensure that the biomedical research community is able to take advantage of the results being generated by investigators utilizing microarrays. We realize that success in this endeavor is closely tied to the availability of software and tools that make these standards useful to the broader research community and to developing standards that are practical and acceptable to researchers generating the data. Our commitment during the coming years remains to involve both commercial and academic partners in developing and implementing these standards and to create open and accessible packages that facilitate biological inquiry. Our hope is these efforts will facilitate the exponential growth in the public repositories that we expect to see in the coming years. More importantly, we believe that our successes and the lessons we learn will allow similar standards to be developed and implemented in a wide range of functional genomics, proteomics and metabolomics experimentation so that we, as a community, can effectively describe computable biological records of experimentation and enable wide use and analysis of the data.

Catherine A. Ball, Gavin Sherlock, Helen Parkinson, Philippe Rocca-Sera, Catherine Brooksbank, Helen C. Causton, Duccio Cavalieri, Terry Gaasterland, Pascal Hin

Paul T Spellman; Editorial: The future of publishing microarray data, Briefings in Bioinformatics, Volume 2, Issue 4, 1 December 2001, Pages 316–318, https://do