Learning generalizable behaviors from demonstration.

Generalizing prior experiences to complete new tasks is a challenging and unsolved problem in robotics. In this work, we explore a novel framework for control of complex systems called Primitive Imitation for Control (PICO). The approach combines ideas from imitation learning, task decomposition, and novel task sequencing to generalize from demonstrations to new behaviors. Demonstrations are automatically decomposed into existing or missing sub-behaviors which allows the framework to identify novel behaviors while not duplicating existing behaviors. Generalization to new tasks is achieved through dynamic blending of behavior primitives. We evaluated the approach using demonstrations from two different robotic platforms. The experimental results show that PICO is able to detect the presence of a novel behavior primitive and build the missing control policy.

Learning multi-agent cooperation.

Advances in reinforcement learning (RL) have resulted in recent breakthroughs in the application of artificial intelligence (AI) across many different domains. An emerging landscape of development environments is making powerful RL techniques more accessible for a growing community of researchers. However, most existing frameworks do not directly address the problem of learning in complex operating environments, such as dense urban settings or defense-related scenarios, that incorporate distributed, heterogeneous teams of agents. To help enable AI research for this important class of applications, we introduce the AI Arena: a scalable framework with flexible abstractions for associating agents with policies and policies with learning algorithms. Our results highlight the strengths of our approach, illustrate the importance of curriculum design, and measure the impact of multi-agent learning paradigms on the emergence of cooperation.

Toward a scalable framework for reproducible processing of volumetric, nanoscale neuroimaging datasets.

BACKGROUND: Emerging neuroimaging datasets (collected with imaging techniques such as electron microscopy, optical microscopy, or X-ray microtomography) describe the location and properties of neurons and their connections at unprecedented scale, promising new ways of understanding the brain. These modern imaging techniques used to interrogate the brain can quickly accumulate gigabytes to petabytes of structural brain imaging data. Unfortunately, many neuroscience laboratories lack the computational resources to work with datasets of this size: computer vision tools are often not portable or scalable, and there is considerable difficulty in reproducing results or extending methods. RESULTS: We developed an ecosystem of neuroimaging data analysis pipelines that use open-source algorithms to create standardized modules and end-to-end optimized approaches. As exemplars we apply our tools to estimate synapse-level connectomes from electron microscopy data and cell distributions from X-ray microtomography data. To facilitate scientific discovery, we propose a generalized processing framework, which connects and extends existing open-source projects to provide large-scale data storage, reproducible algorithms, and workflow execution engines. CONCLUSIONS: Our accessible methods and pipelines demonstrate that approaches across multiple neuroimaging experiments can be standardized and applied to diverse datasets. The techniques developed are demonstrated on neuroimaging datasets but may be applied to similar problems in other domains.

Container-Based Clinical Solutions for Portable and Reproducible Image Analysis.

Medical imaging analysis depends on the reproducibility of complex computation. Linux containers enable the abstraction, installation, and configuration of environments so that software can be both distributed in self-contained images and used repeatably by tool consumers. While several initiatives in neuroimaging have adopted approaches for creating and sharing more reliable scientific methods and findings, Linux containers are not yet mainstream in clinical settings. We explore related technologies and their efficacy in this setting, highlight important shortcomings, demonstrate a simple use-case, and endorse the use of Linux containers for medical image analysis.

Combinatorial metabolic pathway assembly in the yeast genome with RNA-guided Cas9.

The yeast Saccharomyces cerevisiae is an important industrial platform for the production of grain and cellulosic ethanol, isobutanol, butanediol, isoprenoids, and other chemicals. The construction of a successful production strain usually involves multiple gene knockouts and chromosomal integration of expression cassettes to redirect the metabolic fluxes for the conversion of sugars and other feed stocks into the desired product. RNA-guided Cas9 based genome editing has been demonstrated in many prokaryotic and eukaryotic hosts including S. cerevisiae, in which it has been additionally exploited as a tool for metabolic engineering. To extend the utilization of RNA-guided Cas9 as a metabolic pathway building tool, we demonstrated the direct assembly and chromosomal integration of up to 17 overlapping DNA fragments encoding the beta-carotene biosynthetic pathway. Furthermore, we generated a combinatorial strain library for the beta-carotene biosynthetic pathway, directly integrated into the yeast genome to create a diverse library of strains. This enabled the screening of combinatorial libraries in stable chromosomally integrated strains for rapid improvements of product titers. This combinatorial approach for pathway assembly will significantly accelerate the current speed of metabolic engineering for S. cerevisiae as an industrial platform, and increase the number of strains that can be simultaneously evaluated for enzyme screening, expression optimization and protein engineering to achieve the titer, rate and yield necessary for the commercialization of new industrial fermentation products.

Synergy between a collagen IV mimetic peptide and a somatotropin-domain derived peptide as angiogenesis and lymphangiogenesis inhibitors.

Angiogenesis is central to many physiological and pathological processes. Here we show two potent bioinformatically-identified peptides, one derived from collagen IV and translationally optimized, and one from a somatotropin domain-containing protein, synergize in angiogenesis and lymphangiogenesis assays including cell adhesion, migration and in vivo Matrigel plugs. Peptide-peptide combination therapies have recently been applied to diseases such as human immunodeficiency virus (HIV), but remain uncommon thus far in cancer, age-related macular degeneration and other angiogenesis-dependent diseases. Previous work from our group has shown that the collagen IV-derived peptide primarily binds ß1 integrins, while the receptor for the somatotropin-derived peptide remains unknown. We investigate these peptides' mechanisms of action and find both peptides affect the vascular endothelial growth factor (VEGF) pathway as well as focal adhesion kinase (FAK) by changes in phosphorylation level and total protein content. Blocking of FAK both through binding of ß1 integrins and through inhibition of VEGFR2 accounts for the synergy we observe. Since resistance through activation of multiple signaling pathways is a central problem of anti-angiogenic therapies in diseases such as cancer, we suggest that peptide combinations such as these are an approach that should be considered as a means to sustain anti-angiogenic and anti-lymphangiogenic therapy and improve efficacy of treatment.

Constructing the angiome: a global angiogenesis protein interaction network.

Angiogenesis is the formation of new blood vessels from pre-existing microvessels. Excessive and insufficient angiogenesis have been associated with many diseases including cancer, age-related macular degeneration, ischemic heart, brain, and skeletal muscle diseases. A comprehensive understanding of angiogenesis regulatory processes is needed to improve treatment of these diseases. To identify proteins related to angiogenesis, we developed a novel integrative framework for diverse sources of high-throughput data. The system, called GeneHits, was used to expand on known angiogenesis pathways to construct the angiome, a protein-protein interaction network for angiogenesis. The network consists of 478 proteins and 1,488 interactions. The network was validated through cross validation and analysis of five gene expression datasets from in vitro angiogenesis assays. We calculated the topological properties of the angiome. We analyzed the functional enrichment of angiogenesis-annotated and associated proteins. We also constructed an extended angiome with 1,233 proteins and 5,726 interactions to derive a more complete map of protein-protein interactions in angiogenesis. Finally, the extended angiome was used to identify growth factor signaling networks that drive angiogenesis and antiangiogenic signaling networks. The results of this analysis can be used to identify genes and proteins in different disease conditions and putative targets for therapeutic interventions as high-ranked candidates for experimental validation.

Serpin-derived peptides are antiangiogenic and suppress breast tumor xenograft growth.

Angiogenesis is the formation of neovasculature from preexisting microvessels. Several endogenous proteins regulate the balance of vessel formation and regression in the body including pigment epithelium-derived factor (PEDF), which has been shown to be antiangiogenic and to suppress tumor growth. Using sequence homology and bioinformatics, we previously identified several peptide sequences homologous to an active region of PEDF existing in multiple proteins in the human proteome. These short 11-mer peptides are found in a DEAH box helicase protein, CKIP-1 and caspase 10, and show similar activity in altering endothelial cell adhesion, migration and inducing apoptosis.We tested the peptide derived from DEAH box helicase protein in a triple-negative MDA-MB-231 breast orthotopic xenograft model in severe combined immunodeficient mice and show significant tumor suppression.

Sensitive detection of pathway perturbations in cancers.

BACKGROUND: The normal functioning of a living cell is characterized by complex interaction networks involving many different types of molecules. Associations detected between diseases and perturbations in well-defined pathways within such interaction networks have the potential to illuminate the molecular mechanisms underlying disease progression and response to treatment. RESULTS: In this paper, we present a computational method that compares expression profiles of genes in cancer samples to samples from normal tissues in order to detect perturbations of pre-defined pathways in the cancer. In contrast to many previous methods, our scoring function approach explicitly takes into account the interactions between the gene products in a pathway. Moreover, we compute the sub-pathway that has the highest score, as opposed to merely computing the score for the entire pathway. We use a permutation test to assess the statistical significance of the most perturbed sub-pathway. We apply our method to 20 pathways in the Netpath database and to the Global Cancer Map of gene expression in 18 cancers. We demonstrate that our method yields more sensitive results than alternatives that do not consider interactions or measure the perturbation of a pathway as a whole. We perform a sensitivity analysis to show that our approach is robust to modest changes in the input data. Our method confirms numerous well-known connections between pathways and cancers. CONCLUSIONS: Our results indicate that integrating differential gene expression with the interaction structure in a pathway is a powerful approach for detecting links between a cancer and the pathways perturbed in it. Our results also suggest that even well-studied pathways may be perturbed only partially in any given cancer. Further analysis of cancer-specific sub-pathways may shed new light on the similarities and differences between cancers.

Analysis of VEGF--a regulated gene expression in endothelial cells to identify genes linked to angiogenesis.

Angiogenesis is important for many physiological processes, diseases, and also regenerative medicine. Therapies that inhibit the vascular endothelial growth factor (VEGF) pathway have been used in the clinic for cancer and macular degeneration. In cancer applications, these treatments suffer from a "tumor escape phenomenon" where alternative pathways are upregulated and angiogenesis continues. The redundancy of angiogenesis regulation indicates the need for additional studies and new drug targets. We aimed to (i) identify novel and missing angiogenesis annotations and (ii) verify their significance to angiogenesis. To achieve these goals, we integrated the human interactome with known angiogenesis-annotated proteins to identify a set of 202 angiogenesis-associated proteins. Across endothelial cell lines, we found that a significant fraction of these proteins had highly perturbed gene expression during angiogenesis. After treatment with VEGF-A, we found increasing expression of HIF-1α, APP, HIV-1 tat interactive protein 2, and MEF2C, while endoglin, liprin ß1 and HIF-2α had decreasing expression across three endothelial cell lines. The analysis showed differential regulation of HIF-1α and HIF-2α. The data also provided additional evidence for the role of endothelial cells in Alzheimer's disease.

Novel peptide-specific quantitative structure-activity relationship (QSAR) analysis applied to collagen IV peptides with antiangiogenic activity.

Angiogenesis is the growth of new blood vessels from existing vasculature. Excessive vascularization is associated with a number of diseases including cancer. Antiangiogenic therapies have the potential to stunt cancer progression. Peptides derived from type IV collagen are potent inhibitors of angiogenesis. We wanted to gain a better understanding of collagen IV structure-activity relationships using a ligand-based approach. We developed novel peptide-specific QSAR models to study the activity of the peptides in endothelial cell proliferation, migration, and adhesion inhibition assays. We found that the models produced quantitatively accurate predictions of activity and provided insight into collagen IV derived peptide structure-activity relationships.

Angiogenesis-associated crosstalk between collagens, CXC chemokines, and thrombospondin domain-containing proteins.

Excessive vascularization is a hallmark of many diseases including cancer, rheumatoid arthritis, diabetic nephropathy, pathologic obesity, age-related macular degeneration, and asthma. Compounds that inhibit angiogenesis represent potential therapeutics for many diseases. Karagiannis and Popel [Proc. Natl. Acad. Sci. USA 105(37):13775-13780, 2008] used a bioinformatics approach to identify more than 100 peptides with sequence homology to known angiogenesis inhibitors. The peptides could be grouped into families by the conserved domain of the proteins they were derived from. The families included type IV collagen fibrils, CXC chemokine ligands, and type I thrombospondin domain-containing proteins. The relationships between these families have received relatively little attention. To investigate these relationships, we approached the problem by placing the families of proteins in the context of the human interactome including >120,000 physical interactions among proteins, genes, and transcripts. We built on a graph theoretic approach to identify proteins that may represent conduits of crosstalk between protein families. We validated these findings by statistical analysis and analysis of a time series gene expression data set taken during angiogenesis. We identified six proteins at the center of the angiogenesis-associated network including three syndecans, MMP9, CD44, and versican. These findings shed light on the complex signaling networks that govern angiogenesis phenomena.

Anti-angiogenic peptides for cancer therapeutics.

Peptides have emerged as important therapeutics that are being rigorously tested in angiogenesis-dependent diseases due to their low toxicity and high specificity. Since the discovery of endogenous proteins and protein fragments that inhibit microvessel formation (thrombospondin, endostatin) several peptides have shown promise in pre-clinical and clinical studies for cancer. Peptides have been derived from thrombospondin, collagens, chemokines, coagulation cascade proteins, growth factors, and other classes of proteins and target different receptors. Here we survey recent developments for anti-angiogenic peptides with length not exceeding 50 amino acid residues that have shown activity in pre-clinical models of cancer or have been tested in clinical trials; some of the peptides have been modified and optimized, e.g., through L-to-D and non-natural amino acid substitutions. We highlight technological advances in peptide discovery and optimization including computational and bioinformatics tools and novel experimental techniques.

The human-bacterial pathogen protein interaction networks of Bacillus anthracis, Francisella tularensis, and Yersinia pestis.

BACKGROUND: Bacillus anthracis, Francisella tularensis, and Yersinia pestis are bacterial pathogens that can cause anthrax, lethal acute pneumonic disease, and bubonic plague, respectively, and are listed as NIAID Category A priority pathogens for possible use as biological weapons. However, the interactions between human proteins and proteins in these bacteria remain poorly characterized leading to an incomplete understanding of their pathogenesis and mechanisms of immune evasion. METHODOLOGY: In this study, we used a high-throughput yeast two-hybrid assay to identify physical interactions between human proteins and proteins from each of these three pathogens. From more than 250,000 screens performed, we identified 3,073 human-B. anthracis, 1,383 human-F. tularensis, and 4,059 human-Y. pestis protein-protein interactions including interactions involving 304 B. anthracis, 52 F. tularensis, and 330 Y. pestis proteins that are uncharacterized. Computational analysis revealed that pathogen proteins preferentially interact with human proteins that are hubs and bottlenecks in the human PPI network. In addition, we computed modules of human-pathogen PPIs that are conserved amongst the three networks. Functionally, such conserved modules reveal commonalities between how the different pathogens interact with crucial host pathways involved in inflammation and immunity. SIGNIFICANCE: These data constitute the first extensive protein interaction networks constructed for bacterial pathogens and their human hosts. This study provides novel insights into host-pathogen interactions.

NeMo: Network Module identification in Cytoscape.

BACKGROUND: As the size of the known human interactome grows, biologists increasingly rely on computational tools to identify patterns that represent protein complexes and pathways. Previous studies have shown that densely connected network components frequently correspond to community structure and functionally related modules. In this work, we present a novel method to identify densely connected and bipartite network modules based on a log odds score for shared neighbours. RESULTS: To evaluate the performance of our method (NeMo), we compare it to other widely used tools for community detection including kMetis, MCODE, and spectral clustering. We test these methods on a collection of synthetically constructed networks and the set of MIPS human complexes. We apply our method to the CXC chemokine pathway and find a high scoring functional module of 12 disconnected phospholipase isoforms. CONCLUSION: We present a novel method that combines a unique neighbour-sharing score with hierarchical agglomerative clustering to identify diverse network communities. The approach is unique in that we identify both dense network and dense bipartite network structures in a single approach. Our results suggest that the performance of NeMo is better than or competitive with leading approaches on both real and synthetic datasets. We minimize model complexity and generalization error in the Bayesian spirit by integrating out nuisance parameters. An implementation of our method is freely available for download as a plugin to Cytoscape through our website and through Cytoscape itself.

Network legos: building blocks of cellular wiring diagrams.

Publicly available datasets provide detailed and large-scale information on multiple types of molecular interaction networks in a number of model organisms. The wiring diagrams composed of these interaction networks capture a static view of cellular state. An important challenge in systems biology is obtaining a dynamic perspective on these networks by integrating them with gene expression measurements taken under multiple conditions. We present a top-down computational approach to identify building blocks of molecular interaction networks by: (i) integrating gene expression measurements for a particular disease state (e.g., leukemia) or experimental condition (e.g., treatment with growth serum) with molecular interactions to reveal an active network, which is the network of interactions active in the cell in that disease state or condition; and (ii) systematically combining active networks computed for different experimental conditions using set-theoretic formulae to reveal network legos, which are modules of coherently interacting genes and gene products in the wiring diagram. We propose efficient methods to compute active networks, systematically mine candidate legos, assess the statistical significance of these candidates, arrange them in a directed acyclic graph (DAG), and exploit the structure of the DAG to identify true network legos. We describe methods to assess the stability of our computations to changes in the input and to recover active networks by composing network legos. We analyze two human datasets using our method. A comparison of three leukemias demonstrates how a biologist can use our system to identify specific differences between these diseases. A larger-scale analysis of 13 distinct stresses illustrates our ability to compute the building blocks of the interaction networks activated in response to these stresses. Source code implementing our algorithms is available under version 2 of the GNU General Public License at http://bioinformatics.cs.vt.edu/ murali/software/network-lego.

VIRGO: computational prediction of gene functions.

Dramatic advances in sequencing technology and sophisticated experimental assays that interrogate the cell, combined with the public availability of the resulting data, herald the era of systems biology. However, the biological functions of more than 40% of the genes in sequenced genomes are unknown, posing a fundamental barrier to progress in systems biology. The large scale and diversity of available data requires the development of techniques that can automatically utilize these datasets to make quantified and robust predictions of gene function that can be experimentally verified. We present a service called the VIRtual Gene Ontology (VIRGO) that (i) constructs a functional linkage network (FLN) from gene expression and molecular interaction data, (ii) labels genes in the FLN with their functional annotations in the Gene Ontology and (iii) systematically propagates these labels across the FLN in order to precisely predict the functions of unlabelled genes. VIRGO assigns confidence estimates to predicted functions so that a biologist can prioritize predictions for further experimental study. For each prediction, VIRGO also provides an informative 'propagation diagram' that traces the flow of information in the FLN that led to the prediction. VIRGO is available at http://whipple.cs.vt.edu:8080/virgo.