Cloudgene: A graphical execution platform for MapReduce programs on private and public clouds

ABSTRACT: BACKGROUND: The MapReduce framework enables a scalable processing and analyzing of large datasets by distributing the computational load on connected computer nodes, referred to as a cluster. In Bioinformatics, MapReduce has already been adopted to various case scenarios such as mapping next generation sequencing data to a reference genome, finding SNPs from short read data or matching strings in genotype files. Nevertheless, tasks like installing and maintaining MapReduce on a cluster system, importing data into its distributed file system or executing MapReduce programs require advanced knowledge in computer science and could thus prevent scientists from usage of currently available and useful software solutions. RESULTS: Here we present Cloudgene, a freely available platform to improve the usability of MapReduce programs in Bioinformatics by providing a graphical user interface for the execution, the import and export of data and the reproducibility of workflows on in-house (private clouds) and rented clusters (public clouds). The aim of Cloudgene is to build a standardized graphical execution environment for currently available and future MapReduce programs, which can all be integrated by using its plug-in interface. Since Cloudgene can be executed on private clusters, sensitive datasets can be kept in house at all time and data transfer times are therefore minimized. CONCLUSIONS: Our results show that MapReduce programs can be integrated into Cloudgene with little effort and without adding any computational overhead to existing programs. This platform gives developers the opportunity to focus on the actual implementation task and provides scientists a platform with the aim to hide the complexity of MapReduce. In addition to MapReduce programs, Cloudgene can also be used to launch predefined systems (e.g. Cloud BioLinux, RStudio) in public clouds. Currently, five different bioinformatic programs using MapReduce and two systems are integrated and have been successfully deployed. Cloudgene is freely available at http://cloudgene.uibk.ac.at.     

A Self-Directed Method for Cell-Type Identification and Separation of Gene Expression Microarrays

Gene expression analysis is generally performed on heterogeneous tissue samples consisting of multiple cell types. Current methods developed to separate heterogeneous gene expression rely on prior knowledge of the cell-type composition and/or signatures - these are not available in most public datasets. We present a novel method to identify the cell-type composition, signatures and proportions per sample without need for a-priori information. The method was successfully tested on controlled and semi-controlled datasets and performed as accurately as current methods that do require additional information. As such, this method enables the analysis of cell-type specific gene expression using existing large pools of publically available microarray datasets.     

Is Model Fitting Necessary for Model-Based fMRI?

Model-based analysis of fMRI data is an important tool for investigating the computational role of different brain regions. With this method, theoretical models of behavior can be leveraged to find the brain structures underlying variables from specific algorithms, such as prediction errors in reinforcement learning. One potential weakness with this approach is that models often have free parameters and thus the results of the analysis may depend on how these free parameters are set. In this work we asked whether this hypothetical weakness is a problem in practice. We first developed general closed-form expressions for the relationship between results of fMRI analyses using different regressors, e.g., one corresponding to the true process underlying the measured data and one a model-derived approximation of the true generative regressor. Then, as a specific test case, we examined the sensitivity of model-based fMRI to the learning rate parameter in reinforcement learning, both in theory and in two previously-published datasets. We found that even gross errors in the learning rate lead to only minute changes in the neural results. Our findings thus suggest that precise model fitting is not always necessary for model-based fMRI. They also highlight the difficulty in using fMRI data for arbitrating between different models or model parameters. While these specific results pertain only to the effect of learning rate in simple reinforcement learning models, we provide a template for testing for effects of different parameters in other models.     

Quantifying compartment‐associated variations of protein abundance in proteomics data

Quantitative mass spectrometry enables to monitor the abundance of thousands of proteins across biological conditions. Currently, most data analysis approaches rely on the assumption that the majority of the observed proteins remain unchanged across compared samples. Thus, gross morphological differences between cell states, deriving from, e.g., differences in size or number of organelles, are often not taken into account. Here, we analyzed multiple published datasets and frequently observed that proteins associated with a particular cellular compartment collectively increase or decrease in their abundance between conditions tested. We show that such effects, arising from underlying morphological differences, can skew the outcome of differential expression analysis. We propose a method to detect and normalize morphological effects underlying proteomics data. We demonstrate the applicability of our method to different datasets and biological questions including the analysis of sub‐cellular proteomes in the context of Caenorhabditis elegans aging. Our method provides a complementary perspective to classical differential expression analysis and enables to uncouple overall abundance changes from stoichiometric variations within defined group of proteins.     

Odintifier - A computational method for identifying insertions of organellar origin from modern and ancient high-throughput sequencing data based on haplotype phasing

Background

Cellular organelles with genomes of their own (e.g. plastids and mitochondria) can pass genetic sequences to other organellar genomes within the cell in many species across the eukaryote phylogeny. The extent of the occurrence of these organellar-derived inserted sequences (odins) is still unknown, but if not accounted for in genomic and phylogenetic studies, they can be a source of error. However, if correctly identified, these inserted sequences can be used for evolutionary and comparative genomic studies. Although such insertions can be detected using various laboratory and bioinformatic strategies, there is currently no straightforward way to apply them as a standard organellar genome assembly on next-generation sequencing data. Furthermore, most current methods for identification of such insertions are unsuitable for use on non-model organisms or ancient DNA datasets.

Results

We present a bioinformatic method that uses phasing algorithms to reconstruct both source and inserted organelle sequences. The method was tested in different shotgun and organellar-enriched DNA high-throughput sequencing (HTS) datasets from ancient and modern samples. Specifically, we used datasets from lions (Panthera leo ssp. and Panthera leo leo) to characterize insertions from mitochondrial origin, and from common grapevine (Vitis vinifera) and bugle (Ajuga reptans) to characterize insertions derived from plastid genomes. Comparison of the results against other available organelle genome assembly methods demonstrated that our new method provides an improvement in the sequence assembly.

Conclusion

Using datasets from a wide range of species and different levels of complexity we showed that our novel bioinformatic method based on phasing algorithms can be used to achieve the next two goals: i) reference-guided assembly of chloroplast/mitochondrial genomes from HTS data and ii) identification and simultaneous assembly of odins. This method represents the first application of haplotype phasing for automatic detection of odins and reference-based organellar genome assembly.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-015-0682-1) contains supplementary material, which is available to authorized users.     

Emotional and cognitive reinforcement of rat hippocampal long-term potentiation by different learning paradigms

In earlier studies we have shown that a protein-synthesis-independent, early, long-term potentiaton (early-LTP) that lasts up to 4-5 hours can be transformed (reinforced) into a protein-synthesis-dependent late-LTP that lasts > or = 8 hours by either an emotional challenge (e.g. swim stress) or mastering a cognitive task (e.g. spatial learning). In the present study we show that LTP-reinforcement by spatial training depends on the specific constraints of the learning paradigm. In a holeboard paradigm,LTP-reinforcement is related to the formation of a lasting reference memory whereas water-maze training gives more heterogenous results. Thus, cognitive aspects interfere with emotionally challenging components of the latter paradigm. These data indicate that different spatial-learning tasks are weighted distinctly by the animal. Thus, we show that aspects of specific spatial learning paradigms such as shifts of attention and emotional content directly influence functional plasticity and memory formation.     

Mixture-of-Experts Variational Autoencoder for clustering and generating from similarity-based representations on single cell data

Clustering high-dimensional data, such as images or biological measurements, is a long-standing problem and has been studied extensively. Recently, Deep Clustering has gained popularity due to its flexibility in fitting the specific peculiarities of complex data. Here we introduce the Mixture-of-Experts Similarity Variational Autoencoder (MoE-Sim-VAE), a novel generative clustering model. The model can learn multi-modal distributions of high-dimensional data and use these to generate realistic data with high efficacy and efficiency. MoE-Sim-VAE is based on a Variational Autoencoder (VAE), where the decoder consists of a Mixture-of-Experts (MoE) architecture. This specific architecture allows for various modes of the data to be automatically learned by means of the experts. Additionally, we encourage the lower dimensional latent representation of our model to follow a Gaussian mixture distribution and to accurately represent the similarities between the data points. We assess the performance of our model on the MNIST benchmark data set and challenging real-world tasks of clustering mouse organs from single-cell RNA-sequencing measurements and defining cell subpopulations from mass cytometry (CyTOF) measurements on hundreds of different datasets. MoE-Sim-VAE exhibits superior clustering performance on all these tasks in comparison to the baselines as well as competitor methods.     

A framework for generalized subspace pattern mining in high-dimensional datasets

Background

A generalized notion of biclustering involves the identification of patterns across subspaces within a data matrix. This approach is particularly well-suited to analysis of heterogeneous molecular biology datasets, such as those collected from populations of cancer patients. Different definitions of biclusters will offer different opportunities to discover information from datasets, making it pertinent to tailor the desired patterns to the intended application. This paper introduces ‘GABi’, a customizable framework for subspace pattern mining suited to large heterogeneous datasets. Most existing biclustering algorithms discover biclusters of only a few distinct structures. However, by enabling definition of arbitrary bicluster models, the GABi framework enables the application of biclustering to tasks for which no existing algorithm could be used.

Results

First, a series of artificial datasets were constructed to represent three clearly distinct scenarios for applying biclustering. With a bicluster model created for each distinct scenario, GABi is shown to recover the correct solutions more effectively than a panel of alternative approaches, where the bicluster model may not reflect the structure of the desired solution. Secondly, the GABi framework is used to integrate clinical outcome data with an ovarian cancer DNA methylation dataset, leading to the discovery that widespread dysregulation of DNA methylation associates with poor patient prognosis, a result that has not previously been reported. This illustrates a further benefit of the flexible bicluster definition of GABi, which is that it enables incorporation of multiple sources of data, with each data source treated in a specific manner, leading to a means of intelligent integrated subspace pattern mining across multiple datasets.

Conclusions

The GABi framework enables discovery of biologically relevant patterns of any specified structure from large collections of genomic data. An R implementation of the GABi framework is available through CRAN (http://cran.r-project.org/web/packages/GABi/index.html).

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-014-0355-5) contains supplementary material, which is available to authorized users.     

OMERO: flexible, model-driven data management for experimental biology

Data-intensive research depends on tools that manage multidimensional, heterogeneous datasets. We built OME Remote Objects (OMERO), a software platform that enables access to and use of a wide range of biological data. OMERO uses a server-based middleware application to provide a unified interface for images, matrices and tables. OMERO's design and flexibility have enabled its use for light-microscopy, high-content-screening, electron-microscopy and even non-image-genotype data. OMERO is open-source software, available at http://openmicroscopy.org/.     

BIPSPI+: Mining Type-Specific Datasets of Protein Complexes to Improve Protein Binding Site Prediction

Computational approaches for predicting protein-protein interfaces are extremely useful for understanding and modelling the quaternary structure of protein assemblies. In particular, partner-specific binding site prediction methods allow delineating the specific residues that compose the interface of protein complexes. In recent years, new machine learning and other algorithmic approaches have been proposed to solve this problem. However, little effort has been made in finding better training datasets to improve the performance of these methods. With the aim of vindicating the importance of the training set compilation procedure, in this work we present BIPSPI+, a new version of our original server trained on carefully curated datasets that outperforms our original predictor. We show how prediction performance can be improved by selecting specific datasets that better describe particular types of protein interactions and interfaces (e.g. homo/hetero). In addition, our upgraded web server offers a new set of functionalities such as the sequence-structure prediction mode, hetero- or homo-complex specialization and the guided docking tool that allows to compute 3D quaternary structure poses using the predicted interfaces. BIPSPI+ is freely available at https://bipspi.cnb.csic.es.     

Structure Learning in a Sensorimotor Association Task

Learning is often understood as an organism''s gradual acquisition of the association between a given sensory stimulus and the correct motor response. Mathematically, this corresponds to regressing a mapping between the set of observations and the set of actions. Recently, however, it has been shown both in cognitive and motor neuroscience that humans are not only able to learn particular stimulus-response mappings, but are also able to extract abstract structural invariants that facilitate generalization to novel tasks. Here we show how such structure learning can enhance facilitation in a sensorimotor association task performed by human subjects. Using regression and reinforcement learning models we show that the observed facilitation cannot be explained by these basic models of learning stimulus-response associations. We show, however, that the observed data can be explained by a hierarchical Bayesian model that performs structure learning. In line with previous results from cognitive tasks, this suggests that hierarchical Bayesian inference might provide a common framework to explain both the learning of specific stimulus-response associations and the learning of abstract structures that are shared by different task environments.     

The GMOD Drupal bioinformatic server framework

MOTIVATION: Next-generation sequencing technologies have led to the widespread use of -omic applications. As a result, there is now a pronounced bioinformatic bottleneck. The general model organism database (GMOD) tool kit (http://gmod.org) has produced a number of resources aimed at addressing this issue. It lacks, however, a robust online solution that can deploy heterogeneous data and software within a Web content management system (CMS). RESULTS: We present a bioinformatic framework for the Drupal CMS. It consists of three modules. First, GMOD-DBSF is an application programming interface module for the Drupal CMS that simplifies the programming of bioinformatic Drupal modules. Second, the Drupal Bioinformatic Software Bench (biosoftware_bench) allows for a rapid and secure deployment of bioinformatic software. An innovative graphical user interface (GUI) guides both use and administration of the software, including the secure provision of pre-publication datasets. Third, we present genes4all_experiment, which exemplifies how our work supports the wider research community. Conclusion: Given the infrastructure presented here, the Drupal CMS may become a powerful new tool set for bioinformaticians. The GMOD-DBSF base module is an expandable community resource that decreases development time of Drupal modules for bioinformatics. The biosoftware_bench module can already enhance biologists' ability to mine their own data. The genes4all_experiment module has already been responsible for archiving of more than 150 studies of RNAi from Lepidoptera, which were previously unpublished. Availability and implementation: Implemented in PHP and Perl. Freely available under the GNU Public License 2 or later from http://gmod-dbsf.googlecode.com.     

Bio-Ontology and text: bridging the modeling gap

MOTIVATION: Natural language processing (NLP) techniques are increasingly being used in biology to automate the capture of new biological discoveries in text, which are being reported at a rapid rate. Yet, information represented in NLP data structures is classically very different from information organized with ontologies as found in model organisms or genetic databases. To facilitate the computational reuse and integration of information buried in unstructured text with that of genetic databases, we propose and evaluate a translational schema that represents a comprehensive set of phenotypic and genetic entities, as well as their closely related biomedical entities and relations as expressed in natural language. In addition, the schema connects different scales of biological information, and provides mappings from the textual information to existing ontologies, which are essential in biology for integration, organization, dissemination and knowledge management of heterogeneous phenotypic information. A common comprehensive representation for otherwise heterogeneous phenotypic and genetic datasets, such as the one proposed, is critical for advancing systems biology because it enables acquisition and reuse of unprecedented volumes of diverse types of knowledge and information from text. RESULTS: A novel representational schema, PGschema, was developed that enables translation of phenotypic, genetic and their closely related information found in textual narratives to a well-defined data structure comprising phenotypic and genetic concepts from established ontologies along with modifiers and relationships. Evaluation for coverage of a selected set of entities showed that 90% of the information could be represented (95% confidence interval: 86-93%; n = 268). Moreover, PGschema can be expressed automatically in an XML format using natural language techniques to process the text. To our knowledge, we are providing the first evaluation of a translational schema for NLP that contains declarative knowledge about genes and their associated biomedical data (e.g. phenotypes). AVAILABILITY: http://zellig.cpmc.columbia.edu/PGschema