Avi Ma'ayan, Ph.D. - US grants

The Information Managements Unit (IMU) will centralize, federate, disseminate and manage all experimental data, models and computational results in a centralized coherent repository named Information Management System (IMS). The unit will also continue to expand current developments of data-mining by integrating and filtering qualitative and quantitative data from relevant external sources such as online databases and legacy experimental literature. The IMS repository will be continually populated with highcontent experimental data including imaging data from internal sources as described above, modeling results, models and software developed from internal sources, as well as data mined from external sources to enrich the content of the IMS repository. The IMU will maintain local copies of processed versions of publicly available external databases for genes, proteins, protein-protein interactions and drug-protein interactions (i.e. PubMed, EntrezGene (1), Swiss-Prot (2), and DrugBank(3)). The data produced by the internal experimental efforts and modeling efforts associated with the proposed projects will be annotated using metadata (data that describes the data). XML schemas(4), Object Oriented and Relational Database schemas (5) will be developed to organize all these data for the purpose of optimal search, visualization, web-access, download retrieval, statistical analysis and modeling, as well as for the purpose of configuration management, auditing, reporting, backup, and security. The computational modeling efforts and experimental efforts would both benefit from standardization of data exchange and storage schemas. The collaborative interdisciplinary nature of the projects requires rapid flow, exchange, and reusability of information and information systems to facilitate and boost interactions among the laboratories. Therefore, effective sedimentation and sharing of research results, models and databases within the Center and with the research community will be achieved through web-based database interfaces and visualization tools, webbased interactive tools (i.e. Wiki), and other web-based/enabled technologies linked to the IMS repository. Additionally, since computational modeling, as well as initial analysis of experimental data often requires onthe- fly development of software utilities to support the ongoing wet and dry experimental research, the IMU will provide ad-hoc software services and solutions, including all aspect of software development life-cycle including: design, development, testing, training and documentation of the software utilities. These software tools will be cross-platform compatible and will be disseminated for free to the research community in hopes they will assist scientists, inside and outside the center, who have similar computational requirements.

DESCRIPTION (provided by applicant): The real-time chemistry of life occurs at the interfaces of proteins and small chemicals present in the cellular environment. For the first time in history, sufficient information on the human proteome and its interacting chemogenome has accumulated in the form of the 3D protein structural database and the NIH Pubchem chemical library. In addition, 'complementarity algorithms', which accurately assess the energetic complementarity, or fit, between any chemical and any protein surface have recently become available, as has the computing power to deploy these tools in a high-throughput manner. Finally, systems biology methods have evolved to the point that the complete network of interactions of the real-time chemistry of life may be visualized and explored once complementarity algorithms have cross-scored all the chemicals to all the proteins. We propose here to build a chemical biology network to unify these data, along with an intuitive web interface for easy use by non-quantitative biomedical investigators. The network and accompanying interface, once deployed, represents a critical infrastructure that could substantially accelerate collaborative, multi- and interdisciplinary basic, translational, and/or clinical research, specifically by enabling new avenues of drug development for personalized medicine and traditional drug development efforts. Personalized medicine seeks to identify the genetic profiles of patients-frequently highly complex probabilistic expression profiles from microarray data--that correlate with optimal therapy responses. Essentially, the proposed chemical biology network may be queried with one of these profiles and a list of biologically matched drug-like chemical compounds would instantly be retrieved. Such a tool would have a profound impact on the rapidity of clinical trial completion and drug approval, as history has shown that clinical trials of drugs matched to biomarkers, such as her2-neu and Herceptin, proceed more rapidly and succeed more often at a much reduced cost. In addition, the proposed chemical biology network may be queried by lead compound and chemically diverse compounds with similar biological activity may be retrieved instantly. This, too, may accelerate drug approval as advancing a diverse portfolio of leads in parallel for a specific therapy is more likely to succeed rapidly than advancing a single chemical class. The challenge addressed by taking on the grand opportunity targeted by this proposal is elegantly articulated by the FDA's Critical Path Initiative (http://www.fda.gov/oc/initiatives/criticalpath/initiative.html), a forward- looking challenge that faces both the NIH and the FDA jointly, for which there are currently few solutions on the horizon. This specific opportunity benefits from the readiness of all the required elements, so that, despite the enormous potential impact, the milestones are highly achievable within the ARRA timeframe of two years. Once built, this network has the combined advantage of low overhead maintenance for future years, and multiple highly applicable funding opportunities for expansion at the local, commercial and governmental level. The infrastructure, once built, is thus highly sustainable and extensible for continued utility in the U.S. biomedical research enterprise. ) PUBLIC HEALTH RELEVANCE: As the RCSB Protein Data Bank of 3D structures nears 60,000 entries (with the human genome estimated to contain approximately 20,000 genes) and the NIH's Pubchem database nears 20 million chemical compounds, the possibility exists that the majority of the 3D structures of life as well as an informative sample of the diversity of chemical space is readily available to us in 2009. The grand opportunity thus presents itself to cross-connect these two elemental bioscientific databases into a single chemical biology network of direct links between genes, protein targets, and potential chemical therapeutics. Using high-throughput computational methods, we will build this network of relationships as a transformative tool for personalized medicine and drug discovery.

DESCRIPTION (provided by applicant): HIV-associated nephropathy (HIVAN) is a leading cause of end stage kidney disease among African-Americans. Despite growing knowledge of the disease mechanism, therapeutic options have been limited. While gene expression microarray has been widely applied for the study of kidney diseases, data analytical tools are quite limited. Here we propose a novel approach to link the microarray data to the upstream signaling kinase activation using computation/systems biology approach. To accomplish this, we have developed several computational programs including Gene2 Network, a system that uses protein-protein interactions and cell signaling networks to build subnetworks based on seed lists of genes, and kinase enrichment analysis (KEA), a tool that links lists of proteins to the kinases most likely regulating their activity. Using this approach to study the kidney disease in a HIV-1 transgenic mouse model (Tg26 mice), which is a well-characterized animal model for human HIVAN, we identified that homeodomain interacting protein kinase 2 (HIPK2) is a novel upstream kinase regulating the transcription factors and genes activated in kidneys of Tg26 mice. HIPK2 is known to be involved in the regulation of p53, TGF-2, Wnt/2-catenin, and Notch pathways, which are known to mediate apoptosis and fibrosis in kidney disease including HIVAN. Our preliminary data suggest that HIPK2 protein expression is increased markedly in the renal tubulo-interstitial compartment of Tg26 mice and human with HIVAN. In addition, HIPK2 mediates HIV-induced apoptosis and epithelial-mesenchymal transition (EMT) of renal tubular epithelia cells, contributing to kidney fibrosis. Based on these preliminary data we hypothesize that HIPK2 is an upstream protein kinase that mediates tubulointersitial injury in HIVAN and in other kidney diseases. To test our hypotheses, we propose the following specific aims: Specific aim 1: Examine the role of HIPK2 in vitro by confirming the role of HIPK2 in apoptosis and EMT of HIV-infected cells and by determining signaling pathways up- and down-stream of HIPK2 that are activated by HIV. Specific aim 2: Confirm the role of HIPK2 in vivo by investigating whether HIPK2 knockout mice are protected from the development of tubulointerstitial injury in the unilateral ureteral obstruction model and by assessing the effect of HIPK2 knockout on the development of tubulointerstitial injury in Tg26. PUBLIC HEALTH RELEVANCE: HIPK2 could be a novel upstream kinase activating multiple downstream pathways leading to tubulo-interstitial injury, which is a final common pathway in the progression of chronic kidney disease. HIPK2 could be a potential new therapeutic target for the treatment of HIV-associated nephropathy as well as other kidney diseases. Our studies will also provide a novel approach to link upstream signaling pathways to data from gene expression microarray, which could help us to identify key upstream kinases responsible for kidney injury in HIVAN, as well as in other diseases.

DESCRIPTION (provided by applicant): Genome-wide mRNA profiling provides a snapshot of the global state of mammalian cells under different experimental conditions such as diseased vs. normal or drug vs. mock treatment cellular states. However, since measurements are in the form of quantitative changes in mRNA levels, such experimental data does not provide direct understanding of the regulatory upstream molecular mechanisms responsible for the observed changes. Identifying potential cell signaling regulatory mechanisms responsible for changes in gene expression under different experimental conditions or in different tissues has been the focus of many computational systems biology efforts. Most popular approaches include gene ontology or pathway enrichment analyses, as well as reverse engineering of networks from mRNA expression data. However, these methods often assume that differentially expressed genes give rise to pathways and functional modules which is not always true in higher eukaryotes. Here we propose an alternative rational approach, called Expression2Kinases, to identify and rank transcription factors, chromatin modifiers, protein complexes, and protein kinases that are likely responsible for observed changes in gene expression. By combining data from ChIP-seq and ChIP-chip experiments, protein-protein interactions reported in publicly available databases, and kinase-protein phosphorylation reactions collected from the literature, we can identify and rank upstream regulators based on genome-wide changes in gene expression. The idea is to infer the transcription-factors and chromatin regulators responsible for changes in gene-expression; then use protein-protein interactions to connect the identified factors to build transcriptional complexes involving the factors; then use kinase-protein phosphorylation reactions to identify and rank candidate protein kinases that most likely regulate the formation of the identified transcriptional complexes. We plan to validate this method with phosphoproteomics data, data from drug perturbations followed by genome-wide gene expression, RNAi screens, as well as through literature-based text-mining approaches. The project will produce several high quality datasets, web-based software, new algorithms, and robust lists of transcription-factors, histone modifiers, and kinase rankings likely responsible fo mammalian cell regulation. The approach will be experimentally tested in several collaborative projects mainly exploring regulation of differentiating stem and iPS cells. The databases, software tools and algorithms developed for this project will advance drug target discovery and help in unraveling drug mechanisms of action.

The two Cores of SBCNY will serve as the incubators for the development and standard usage of tools and technologies that will support the four research themes and the education and outreach activities of the Center. The cores will interact closely with one another, other Center investigators and non-Center researchers on an ongoing basis. Space limitations preclude detailed descriptions. The cores are run in a team fashion by Center investigators with deep technological expertise. Computational Core Center Investigators: Avi Ma'ayan, Marc Birtwistie, David McQueen and Eric Sobie.

The Knowledge Management Center for Illuminating the Druggable Genome (KMC-IDG) at Mount Sinai will assemble, organize and visualize data collected from the under-studied druggable genome from the four families: protein kinases, nuclear receptor, ion channels and GPCRs. The KMC-IDG will also attempt linking such under-studied druggable targets for their potential applications in various diseases. To achieve this we will assemble and abstract data from four domains: proteins/genes/targets, drugs/perturbagens, diseases/phenotypes/side-effects, and data from individual patients. Various pipe-lines and workflow will be established to connect clusters of patients from various diseases to under-studied druggable targets. The Ma'ayan and Dudley Labs are well positioned to carry out successfully this project based on their prior track record of productivity, foundation of source code and data that is already collected and organized, and strong existing user base that can be directed to the newly developed portal. In addition, both labs have a strong track record of collaborations including the computational identification and experimental validation of at least one under-studied protein kinase as a potential important target for attenuating kidney fibrosis. One unique and innovative research component of this project is an investigation into the sources of the literature and experimental biases that exist in the molecular and cellular biology research domains.

DESCRIPTION (provided by applicant): For this project we will establish the Data Coordination and Integration Center (DCIC) for the Library of Integrated Network-based Signatures (LINC) program as part of the Big Data to Knowledge (BD2K) initiative. The Center will have four major components: Integrated Knowledge Environment (IKE), Consortium Coordination and Administration (CCA), Data Science Research (DSR) and Community Training and Outreach (CTO). The Center will construct a high-capacity scalable IKE enabling federated access, intuitive querying and integrative analysis and visualization across all LINCS resources and many additional external data types from other relevant resources. The Center will perform, support, and fund several Internal and external DSR projects, addressing various data integration and intracellular molecular regulatory network challenges. The CTO efforts will establish several educational programs including a LINCS MOOC, summer undergraduate research program, initiate and support diverse collaborative projects leveraging LINCS resources, and systematically disseminate LINCS data and tools via a variety of mechanisms. The organizational structure of the Center will include a strong CCA that will support and manage the Center goals and deliverables, and coordinate activities across the LINCS and BD2K programs. The IKE resources will build on the infrastructure, analysis tools and data that we have already established in the LINCS pilot and transition phases, thus minimizing executional risks. The Center brings together a proven team of computational experts with several years of experience with LINCS data and complementary expertise: Drs. Ma'ayan, Schurer, and Medvedovic will develop and deploy a next generation computational infrastructure, develop novel analysis tools and methods enabling researchers to glean new insights from integrative models of biological systems to link complex diseases/phenotypes with drugs and the pathways that those drugs target in different cells and tissues. The project will play a key role to transform and accelerate the discovery of novel therapeutics and improve diagnostics for significantly advancing human health.

The Data Organization Core (DOC) of the Mount Sinai's KMC-IDG will collect, process, and maintain attributes about the druggable targets for all proposed families: protein kinases, G-protein coupled receptors, nuclear receptors and ion channels. The emphasis will be to focus on those genes/proteins that are understudied and collect unbiased genome-wide profiling datasets. In addition, the DOC will collect, process and maintain data tables and attributes for all other genes/proteins, drugs/small-molecules and other perturbagens, pheontypes/diseases/side-effects, and clinical as well as genomics datasets from cohorts of patients. This will enable us to identify links between and across genes/proteins networks, drugs/small-molecules and other perturbagens networks, pheontypes/diseases/side-effects networks, and clusters of individual patients with similar profiles. For this, the Core will develop and apply clustering and classification algorithms as well as workflows to make predictions about the potential applicability of targeting the understudied proteins for various translational applications in personalized medicine.

The Knowledge Management Center for Illuminating the Druggable Genome (KMC-IDG) at Mount Sinai will assemble, organize and visualize data collected from the under-studied druggable genome from the four families: protein kinases, nuclear receptor, ion channels and GPCRs. The KMC-IDG will also attempt linking such under-studied druggable targets for their potential applications in various diseases. To achieve this we will assemble and abstract data from four domains: proteins/genes/targets, drugs/perturbagens, diseases/phenotypes/side-effects, and data from individual patients. Various pipe-lines and workflow will be established to connect clusters of patients from various diseases to under-studied druggable targets. The Ma'ayan and Dudley Labs are well positioned to carry out successfully this project based on their prior track record of productivity, foundation of source code and data that is already collected and organized, and strong existing user base that can be directed to the newly developed portal. In addition, both labs have a strong track record of collaborations including the computational identification and experimental validation of at least one under-studied protein kinase as a potential important target for attenuating kidney fibrosis. One unique and innovative research component of this project is an investigation into the sources of the literature and experimental biases that exist in the molecular and cellular biology research domains.

The User Interface Portal (UIP) of the Mount Sinai's KMC-IDG will develop a state-of-the-art web-site that would host the presentation and access to the data collected by the DOC. The web-site will have a dedicated page of each under-studied druggable target with different plug-in tools to explore the various aspects of each target including: protein structure, protein-protein interactions, regulation by transcription factors, mouse phenotypes, expression profiles in different tissues and conditions, mouse knockout phenotypes, gene ontology information, post-translational modification and many more. In addition, the portal will enable interactive visualization of various applications that will place the under-studied targets within networks made of cohorts of patients, cell lines, diseases/side-effects/phenotypes, drugs and other genes and proteins. The UIP will have a powerful search engine that would index all entities in the DOC and will learn from user experience. The UIP will enable users to build their own data analysis pipelines based on the user specific needs. In addition, the UIP will be designed in a plug-in architecture to enable the community to contribute data analysis and visualization tools.

SUMMARY The understudied protein targets that are the focus of the implementation phase of the Illuminating the Druggable Genome (IDG) project need to be placed in the contexts of gene-sets/pathways, drugs/small-molecules, diseases/phenotypes, and cells/tissues. By extending our previous methods, we will impute knowledge about the understudied potential target protein kinases, GPCRs, and ion channels listed in the RFA using machine learning strategies. To establish this classification system, we will organize data from many omics- and literature- based resources into attribute tables where genes are the rows and their attributes are the columns. Examples of such attribute tables include gene or protein expression in cancer cell lines (CCLE) or human tissues (GTEx), changes in expression in response to drug perturbations or single-gene knockdowns (LINCS), regulation by transcription factors based on ChIP-seq data (ENCODE), and phenotypes in mice observed when single genes are knocked out (KOMP). In total, we will process and abstract data from over 100 resources. We will then predict target functions, target association with pathways, small-molecules/drugs that modulate the activity and expression of the target, and target relevance to human disease. To further validate such predictions, we will employ text mining to identify knowledge that corroborates with the data mining predictions, perform molecular docking of predicted small molecules using homology modeling, and seek associations between variants and human diseases by mining electronic medical records (EMR) together with genomic profiling of thousands of patients. In addition, we will develop innovative data visualization tools to allow users to interact with all the collected data, and develop social networking software to build communities centered around proteins/genes/targets as well as biological topics including pathways, cell types, drugs/small-molecules, and diseases. Overall, we will develop an invaluable resource that will accelerate target and drug discovery.