HFSP Grant Proposal


reStructured

==================================================================================== Reprogramming lifespan - in silico simulation of organismal aging and its reversal ====================================================================================

Abstract

Aging is arguably the major biomedical challenge of the 21st century, and elucidating causal mechanisms of aging is one of the outstanding scientific challenges of our time. Longevity pathways, discovered in the past two decades, have yet fully to explain how lifespan is determined and which molecular processes drive aging. In the nematode worm C. elegans, feedback-loop circuits were found to regulate insulin-like signaling, inflammation and nutrient-sensing, and jointly to modulate lifespan over a 10-fold range. We propose to create the first unified model of lifespan-control circuits in worms, to robustly predict responses to perturbations. The initial model, treating lifespan as a function of transcript data, will be refined iteratively to incorporate effects of environment and genotype. By defining molecular circuits that control the rate and state of aging, we can predict means to slow or perhaps even reverse aging. Three highly collaborative and interdisciplinary aims are proposed:

Aim 1. Acquisition, analysis and integration of age- and longevity-related transcript data

Transcripts will be quantified from C. elegans populations of widely varying lifespans due to genetic and dietary factors. The resulting profiles will be statistically analyzed and integrated with prior datasets monitoring expression by age and tissue, and interaction networks including age- and tissue-specific subnetworks will identify hub genes. Clustering and machine-learning algorithms will define profiles that track with age or predict lifespan.

Aim 2. In silico models to simulate organismal aging

Organisms can be described at several levels of complexity, each with its own rules for interactions among components. The entities and relations, based on integrated data, will be represented in simulations via object-orientated programming and scalable parallel computing. Models will be guided by predictive profiles from Aim 1, and will also conform to known constraints (e.g. regulatory feedback loops). Initial simplifying assumptions will be progressively removed, and simulations refined as new data are added.

Aim 3. Predicting and testing interventions

The simulations in Aim 2 will reveal proximal factors limiting lifespan, and mechanisms by which interventions extend lifespan. A reporter panel will be compiled, in which GFP is driven by promoters of genes whose transcript levels are most predictive of age and/or lifespan. With this panel, cell-, tissue- and age-dependent responses of key system components will be compared over diverse genetic and dietary contexts. Lifespan and reporter data, assessed for interventions predicted to extend survival, will be used to test the validity of simulations and to iteratively improve them.

Overall, we aim to create the first unified model of lifespan-control circuitry in an animal (worms), able to predict its behaviour in response to perturbations. The multidisciplinary research team encompasses a biologist/geneticist (Shmookler Reis), a computer scientist (Kestler), a mathematician/statistician (Lee) and is led by an integrative systems biologist (de Magalhaes).

Overview (Section 5A)

Aging is arguably the major biomedical challenge of the 21st century, and elucidating causal mechanisms of aging is one of the outstanding scientific challenges of our time. The project proposed here is a novel, interdisciplinary endeavor to create the first unified model of lifespan-control circuitry in worms, able to predict its behaviour in response to perturbations. Since the first identification of a life-extending mutation two decades ago [1], genetic mechanisms governing longevity have been unveiled at an accelerating pace, particularly in C. elegans [2-8]. Nevertheless, full understanding of the processes underlying aging and longevity remains an elusive goal.

Several feedback-loop circuits, involving multiple tissues and types of signaling, govern a switch between pro-longevity and reproductive states implemented by the C. elegans insulin/IGF-1 signaling (IIS) pathway [9-11]. Pathways directing TOR (target of rapamycin) nutrient-sensing, innate immunity and inflammation, which intersect IIS to modulate lifespan, also involve feedback loops [11-16]. Although integration of these circuits has not hitherto been attempted, it appears both achievable and worthwhile. Current evidence suggests that longevity circuitry is simple enough to be fully represented on a computer of unexceptional power, and yet is sufficiently complex to require computer modeling to predict its behavior in response to perturbations. We propose to use available computational power in an innovative way, to create a unified model of longevity set-point determination and to define its relationship to biologically meaningful outputs - including metabolism [15,17-22], stress responses [2,8,23-25] and longevity [7,15-16,26]. These data, in turn, can lead to refinement of the models in an iterative way, to create higher-generation models of increasing complexity and robustness for prediction of lifespan and the effects of interventions.

The aging process is remarkably plastic in response to a variety of known perturbations: genetic modifications (such as knock-downs), dietary changes (primarily diverse modes of limiting intake), and temperature each typically effecting increases in lifespan by 1.5- to 2-fold, but exceptionally by up to 10-fold [5,7-8,16,25,27-32]. If we can identify and characterize the molecular circuitry that controls the rate and state of organism aging, then in principle it should be possible to reset an animal to a younger state. Reversal of animal aging has never previously been demonstrated, and a systematic effort to achieve it would be informative even if it were to fail, since irreversibility of aging has been widely assumed but never tested as a hypothesis in animals.

A. Especially Novel/Collaborative Aspects

In contrast to traditional biological reductionism, focusing on narrow abstractions of one or more specific subsystems, we will take a holistic view by creating a computer simulation of a multicellular organism as it undergoes aging, via extensive and inclusive integration of biological information acquired at many levels. Data will be gathered and a robust unification method will be applied. Central to this, we will develop a framework with unprecedented power to efficiently integrate collaborative work. Meta-analytical and machine-learning algorithms, coupled to the development of a systematic classification system, will complement human limitations in recognizing complex biological relationships. Contrary to the static nature of conventional network analyses, we will generate dynamic spatio-temporal networks of major biomolecular contributors to aging and the regulation of lifespan. We will apply the object-orientated programming paradigm to model biological systems, and utilize parallel computing to overcome any performance bottlenecks. By extending network biology to real-time simulation of time-dependent processes in a simple metazoan, we hypothesize that current biological knowledge is already sufficient to predict the effects of genetic and environmental alterations on lifespan. As a corollary, we propose that such a model can not only predict novel intervention strategies to extend lifespan, but may also suggest means to reverse aging by resetting key processes to an earlier state. This is an ambitious project but one that we think will be successful via the collaboration of experts from various disciplines and will have a significant impact in the field of research on aging.

B. Essential Collaborative Elements

Among the multitude of changes that occur during aging, it is difficult to distinguish the drivers from the passengers. It is expected that relatively few changes are causal, while many more are their downstream consequences. Novel approaches of data integration, which scale with the ever increasing volume of data, are necessary to take advantage of current knowledge and extrapolate beyond its boundaries. Therefore, we need deep knowledge about mechanisms of aging and longevity expansion [7-9,27,33-39], genomics techniques [8,33,40-42], meta-analytic methods [8,42-45], and network-based inference [33,37,46-48], which can only be achieved by close collaborations between a team encompassing different expertises. JPdM will develop the informatics structure required for efficient collaboration, will oversee the data collection and integration, and will coordinate bioinformatic analyses based on data that are publically available or added by RJSR. HAK will construct mathematical models and simulations as well as will be responsible for elimination of performance-critical bottlenecks by the use of functional and parallel computing paradigms. TL will advise on statistical assessments, will pre-process data to remove systematic biases, and will perform significance measurements based on functional categories or other classification schemas as necessary. RJSR will be responsible for experimental validation/testing of simulation-derived predictions. Both RJSR and JPdM provide knowledge of pathways that modulate longevity in diverse taxa.

C. Different Disciplines Represented Among the Team Members

Our team comprises a diverse set of disciplines, including but not limited to cell, moluecular, and evolutionary biology; genetics; theoretical biochemistry; mathematics and statistics; and bioinformatics, systems and computational biology supported through computer sciences. Backgrounds of team members include the genetics and biology of aging, spanning species from yeast to H. sapiens (JPdM and RJSR), as well as ontogeny, statistics, mathematics, computation and simulation as applied to biological systems (TL and HAK). These areas complement each other in a balanced fashion, while bringing together novel combinations of realms of expertise. JPdM (Portuguese; UK-based), although initially trained as a microbiologist, has exceptional IT skills and extensive experience in data integration and building of informatic databases [49-51]. He has previously combined these to create web resources that encompass the aging process. Additionally, although JPdM is a Young Investigator, he already has a strong track record in next-generation sequencing [40,52] and in promulgating theories to account for observed features of the aging process [36,53-54]. RJSR is a US-born and -based molecular geneticist who made important contributions to our understanding of the instability of DNA structure and methylation during aging, and was the first to map quantitative trait loci (QTLs) governing longevity of a metazoan (e.g. [55-57]), and the first to translate an age-dependent osteoporosis QTL from mice to humans [58-59]. He also has extensive experience in generating -omics data at the transcript, protein and metabolite levels [7-8,25]. TL is a Korean mathematician/statistician and also a Young Investigator, who has made important innovations in the categorical analysis of large-scale datasets [42-43,43-44,60-62]. HAK is a German computer scientist/engineer who has pioneered artificial intelligence (AI)-related tasks [65] such as pattern recognition, signal processing, and their applications in bioinformatics; software development and concurrent computing [64]; modeling and simulation in systems biology; and the use of image processing [65] and visualisation of biological data [48,66-69].

D. Origin of the Idea of Collaboration

The proposed project is based on the research focus of the PI (JPdM) and RJSR, who study the biology and genetics of aging, although with different approaches. The ideas leading to this collaboration originated at an aging conference in Cambridge/UK, from discussions between the labs of RJSR and JPdM about ways to use molecular profiling and computer modeling to gain insights into C. elegans longevity control. These ideas matured at a bioinformatics and systems biology conference (RoSyBA) in Germany and a conference on genetics and bioinformatics of longevity in Moscow/Russia at which RJSR and JPdM chaired sessions. Scientific interactions at these meetings led to RJSRs co-authorship on a recent paper from JPdMs lab [33]. HAK and TL were invited to participate in this proposal based on their respective outstanding contributions to components deemed essential to success: modeling and simulations via parallel processing (HAK), and the harmonization and meta-analysis of large data collections (TL). A key aspect of this collaboration is that each of the investigators has a deep and long-standing interest in its goals, and is an established leader in one or two of its aspects, but has been unable to achieve such an ambitious undertaking without the complementary contributions of the others.

E. Conduct of the Collaboration

Specific tasks are allocated among the team members, although interaction will be essential at all levels of the project. At the center of this collaborative project will be a web framework for data integration, exchange, and curation that all members will routinely use. In addition to extensive exchange of e-mails, data and information, monthly web-video conferences will allow team members to discuss progress and to brainstorm in order to overcome any obstacles encountered. An annual meeting in Liverpool will bring together all the applicants and staff working on this project; relevant outside experts will be invited as well to participate. Collaborative visits will also be established between the participating labs for staff exchange.

Difference From Ongoing Research (Section 5B)

Our proposed project builds upon the skills and expertise available in the applicants' laboratories, yet the aims of the project can only be achieved via a collaborative, multidisciplinary approach involving all the applicants. JPdM has established the leading information repositories on aging research, TL introduced several innovative improvements that dramatically increase the sensitivity and specificity of meta-analysis for gene expression profiles, and RSJR has profiled transcripts that change with genetic extension of C. elegans lifespan, but none of these applicants has attempted to make predictive models. HAK has pioneered machine learning, parallel computing and data visualization but had not yet combined this with predictive model building and meta-signatures to unravel aging of a whole animal. Therefore, while the applicants have the skills to carry out individual tasks, none is working on building predictive models of aging in an animal. Individually none of the applicants can accomplish such as ambitious goal and only by working together can we achieve the aims we propose. For example, for HAK to construct models and simulations, the data generated by RJSR, processed and analyzed by TL and integrated by JPdM will be essential. Refinement of the predictive models by HAK also depends on experimental validation by RJSR and integration with prior knowledge by JPdM and previous datasets, meta-analyzed by TL. Only working as a team can we achieve the goals of the project.

Proposed Research

Background

The greying of the population is arguably the major biological and biomedical challenge of the 21st century [49]. Therefore, research on the biology of aging has an unparalleled potential to improve quality of life and health [71-72]. Longevity pathways, discovered in the past two decades, have yet to fully explain how lifespan is determined and which molecular processes drive aging, have yet to explain how lifespan is determined. In the nematode worrm C. elegans, feedback-loop circuits were found to regulate insulin-like signaling, inflammation and nutrient-sensing, and jointly to modulate lifespan over a 10-fold range. In spite of this recent progress, mechanisms of lifespan extension interweave with multiple genetic and regulatory pathways and are thus not fully understood.

In the post-genomic era, the generation of genome-wide data provides increasing opportunities to dissect the mechanisms of aging and of anti-aging interventions with computational and systems-biology approaches. Challenges remain, however, in the integration of diverse data types, their interpretation in the context of aging, and in development of models with predictive capacity for age and longevity. Recent progress in this broad area has been impressive [73-74] and recently the first full model of a cell has been developed [81]. However, new approaches are needed to tackle the intrinsic complexity of biological systems and especially of aging systems.

Objective

The goal of this project is to create the first unified model of lifespan-control circuitry in an animal (worms), able to predict its behaviour in response to perturbations.

Detailed description of the research project

This project will focus on worms (C. elegans), one the simplest and best characterized metazoan species with regard to aging. To achieve our ambitious goal, three highly collaborative and interdisciplinary aims are proposed (see Figure 1 for an overview).

Aim 1. Acquisition, Analysis and Integration of Age- and Longevity-Related Transcript Data ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Transcript-abundance data will be generated from C. elegans populations with longevities that vary widely due to genetic and dietary factors [RJSR]. Data will be preprocessed to remove systematic biases [8,42-43] and to define categories that differ significantly among longevity groups with high specificity and sensitivity [8,43-44] [TL]. Profiles and signatures generated will be integrated with extant datasets that provide similar data across the lifespan [82], and for specific tissues and cells [77] [TL/JPdM].

1.1 Transcriptomics of isogenic strains varying by more than 10-fold in longevity [RJSR] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The RJSR laboratory has constructed an isogenic set of 12 longevity mutants representing diverse pathways (e.g. [7-8]), which will be assessed at several adult ages using the same transcriptomics platform: either via sequencing of mRNAs and miRNAs (Il) or by microarrays (Affymetrix). We have experience with both procedures (e.g. [8,25,38,40,52]), and present both alternatives because sequencing methods are now substantially more expensive than arrays, and yet we expect their costs to fall below array pricing by the time of funding. If this occurs, we would favor a sequencing platform due to its greater information content and sensitivity, and lower error rate.

1.2 Data compilation and integration [JPdM] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Because of the rapid pace at which data complexity grows, assembly and maintenance of a comprehensive compilation of the current state of knowledge in any field such as aging is not a trivial task. JPdM previously developed the GenAge database with hundreds of aging-related genes inferred from the literature [49]. GenAge has been recently revamped and contains now over 1,500 genes (>500 of which are from worms), with new annotations indicating whether they are pro- or anti-longevity (i.e., whether they accelerate or retard aging when disrupted), and providing lifespan summary data (median, mean and maximum lifespan, as available) for a variety of gene-expression manipulations in C. elegans. We now propose a complementary combination of relational and non-relational databases to integrate low- and high-throughput data from C. elegans and develop a platform we call WormAge. Integration of existing resources (including GenAge, WormBase, modEncode, Ensembl, NCBI and UniProt), and the development of innovative ways of data mining and handling, will be primarily coordinated by JPdM, whose lab presently maintains both GenAge and the Digital Ageing Atlas (DAA) the most widely used databases of age-related genes and changes [49]. To coordinate the data generation by RJSR and to facilitate analyses by TL and modeling by HAK, we will develop WormAge using the techniques and methods employed for GenAge and the DAA. WormAge will include information from public databases (e.g., microarray data from GEO and ArrayExpress) as well as data generated within this project by RJSR (1.1).

1.3 Data pre-processing, classification and meta-analysis [TL/JPdM] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The interpretation of large amounts of transcriptomic data (as well as proteomic and metabolite data), and their translation into biologically meaningful information, must precede model building. TL developed a procedure to normalize large datasets and filter out "noise", as well as a method for functional-annotation meta-analysis to determine the overall significance of differential expression for genes in a Gene Ontology (GO) category while correcting for transcript-level correlations among those genes. These two operations combined to greatly improve both sensitivity and specificity for detection of categories that altered between experimental groups [8,43]. The same procedures can be applied to other omics-related data as well as to pathways and interaction networks that vary with age or lifespan, to classify important functional categories comprising traditional omics or extending to interactomics data. Pre-processing and meta-analysis of high-throughput data to define significant subclasses, and their component genes and transcripts, will be coordinated by TL, who recently developed powerful tools for data normalization, and subsequent meta-analysis of differential expression data within functional-annotation categories. This will be combined with complementary methods developed by JPdM [50-51].

TLs meta-analysis procedure, to determine the overall significance of a treatment effect within a gene category (e.g. a pathway or GO annotation), combines the p-values that reflect the by-gene significance of a contrast between two treatments, across all members of that category [8,43]. We propose to expand this method to handle time-series data, which represent the continuous change of gene expression as a function of age, lifespan, or time of intervention (e.g., DR). Such an extension is non-trivial because the correlation structures among genes in time-series data may include a delay or response time t. Comparable tools have proven quite useful to intercorrelate single-gene expression data, but have not been available for meta-analysis of category-aggregated data.

Data from RJSR 1.1 will be uploaded to the databases developed by JPdM (1.2), which will then be provided to TL for processing and analyses and fed back into the databases (1.4).

1.4 Data integration [JPdM] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ In order for our analyses to be merged with previous knowledge, three distinct approaches will be applied to integrate information within a unified schema (Table 1). Initially, all interaction and annotation databases that pass evaluation criteria will be integrated. Individual parsers will be created for each external database, allowing information to be combined under a unified format. Two modes of data integration and mapping will be utilized: 1.) Implicit: All data inputs will be mapped as either an entity, or a relationship between two entities (stored in just three relational tables: Entity, Relationship, and Attribute). 2.) Explicit: All entities will be assigned to defined database abstraction models such as genes, transcripts, proteins, molecules, metabolites, chromosomes (= string), genome (= a large assembly of strings), cells (container), subcompartments (e.g. nucleus, mitochondria, etc.), tissues, organs, and organ system, as well as relationships such as interactions (of diverse types), reactions, homologies, and associations. Information is then mapped onto this schema and stored in databases as appropriate.

1.5 Molecular signatures [JPdM/RJSR/TL/HAK] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Transcript data describing changes at the molecular level will be assembled as molecular signatures. As described above for other entity annotation and relationship information (1.4), individual profiles will be converted into a unified format in which each profile is annotated with adult age, genotype, treatment, etc. Profiles will then be contrasted to yield meaningful comparisons (e.g. old vs. young, or aging time-series, long-lived mutant vs. wild-type, diverse modes of dietary-restriction vs. ad libitum fed, etc.) and used to annotate transcripts for their differential expression levels. Multiple comparisons will be utilized to derive a common signature of a given process. Intersections of signatures may identify the underlying processes driving multiple trends; they will be used to evaluate significance of overlap, and to annotate entities in interaction networks and simulations. For example, co-regulation of specific processes or pathways, across multiple lifespan-extending interventions, may provide insights into common underlying mechanism (e.g. shared by IIS and dietary restriction). The compendium of all comparisons can be used to create a coexpression network of genes and their pairwise extents of coregulation.

Boolean representation will test the robustness of the results of involving diverse datasets (HAK), and meta-analytic statistical methods (developed by TL) will be used to assess the significance of meta-signature associations with longevity [42-43]. Interaction networks will be used to identify hub genes [33], incorporating age- and cell-type-specific subnetworks based on transcript levels per class [JPdM]. Clustering and machine-learning algorithms will define profiles that track with age and predict lifespan [JPdM/TL/HAK], while pathways driving the lifespan effects of interventions will be discerned with the help of pattern-recognition algorithms [HAK].

The results generated, assembled and analyzed will serve as input for Aim 2 to refine the models, although model generation will begin from the outset of the project based on data already available.

Aim 2. in silico Simulation of Aging and Longevity Control in an Animal Model ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ To create simulations, C. elegans will be described at sequential levels of complexity (organs/tissues/cells/molecules), each with its own rules governing interactions among components. The entities and relations, based on integrated data, will be represented in simulations via object-orientated programming and parallel computing scaled to available resources [HAK]. Models will be constrained by data accrued and analyzed in Aim 1, and by previously defined system properties (e.g., regulatory feedback loops) [RJSR/JPdM]. Simulations will be refined as new data are added and as simplifying assumptions are progressively removed.

We will take advantage of the fully categorized anatomy of C. elegans to map, as dictated by available data from Aim 1, gene expression data to the corresponding adult age and tissue localisation as well as genotype and diet. Simulations can embody complex relationships, which may allow us to answer important questions that do not readily yield to reductionist strategies, such as: Which age-related changes are most crucial to limit lifespan? What are the mechanisms underlying known life-extension interventions, including dietary restriction (which may span a 3-fold range of longevities), or gene mutations that extend lifespan up to 10-fold [27,78]? The goal of this aim is to generate one or more simulations capable of predicting highly effective life-extension interventions via transcript data integrated with protein-interaction networks.

2.1 High-performance parallel computing [HAK] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ An unusual feature of this proposal is the extensive use of parallel computing paradigms. The integration of data, meta-analysis, network generation, and in particular simulation, are perfectly suited to concurrent computing. In those paradigms, a large and complex task is spread across multiple processors. Many scientific problems can be decomposed either by isolating separate tasks to be performed independently and simultaneously by multiple processors, and/or by partitioning the space (or time) coordinates of the system being modeled so that values for each sub-region (or time interval) can be calculated simultaneously. Simulation of biological systems requires high performance computing which can be achieved with the use of compute clusters or general-purpose scientific computing on graphics processing units (GPGPU). HAK showed that by by combining functional programming and parallelism via Software Transactional Memory (STM) on multiple processing cores it is possible to dramatically accelerate the operation of clustering algorithms even for massively large datasets [64]. Also, a grid-based solution for parallel cross-validation enables the realization of large-scale optimization experiments [Mussel, et al. 2012]. GPU-streaming programming fits especially well with biological parallelism [79] and has already proven its enormous potential in traditional network dissections. This approach, applied on large networks consisting of 325K nodes and 1.46M edges, reduced the analysis time from hours to a fraction of a second. Such an "embarrassingly high parallelism will enable us to construct full featured real-time applications for exploring biological networks and processes interactively in a simulation [85]. For this purpose we choose to use OpenCL (Open Computing Language) and write the required kernels. We intend to utilize both CPU- and GPU-based methods, and map different processes to different architectures to accelerate computation. Aging will present a challenge to the power of modern computer modeling and simulation, high performance computing, and computer visualization, but one we think is solvable.

2.2 Object-oriented biological modeling [HAK/JPdM] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ All entities and relations will be described as objects. This concept is very powerful as it allows extensive code reuse, as well as native and intuitive mapping of biological structures and hierarchies. A protein, for instance, inherits the attributes from a transcript which itself inherits from a gene. Genes are positioned and annotated slices of a chromosome object which itself is part of the genome. The genome is represented by strings which can be accessed efficiently from the sequence data storage. Relationships such as interactions refer to just two instances of objects which can be any kind of entity. More complex relationships can be built from sets of two-object interactions. Another advantage of the object-oriented approach is the possibility to modularize the system, mapping the functional units to independent modules. This also allows for a multi-scale representation of large regulatory systems: a closer view of a regulatory subsystem reveals detailed interactions and dynamic behavior. A high-level abstraction establishes a global view of the behavior of the complete system, treating the subsystems as black boxes.

2.3 Comparative dissection of heterologous interaction networks [JPdM/HAK/TL] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ The assembled relation information will link diverse types of entities (genes, transcripts, proteins, metabolites, processes, etc.). Although we do not propose to incorporate data on proteins and metabolites for this proposal, databases and data annotations will have the capability to add them at a later time. These data annotations will allow us to generate heterologous interaction networks which we will explore to uncover new insights into aging and longevity mechanisms. For example, the set of genes classified as aging-related can be compared to a random gene-set with respect to network properties, or to identify novel candidate genes predicted to affect aging and lifespan in a specified direction (i.e. as pro-aging or longevity-assurance genes). In some instances, directionality of interactions (such as activating or inhibiting post-translational modifications, or transcriptional regulation) allows for a clear inference as to which genes are more likely to be upstream controllers or downstream effectors of aging, and whether increasing or decreasing function is expected to extend lifespan. Molecular signatures will be integrated with the networks to incorporate expression-level changes during aging or upon life-extending interventions. JPdM recently developed a method to compare molecular signatures at the network level in yeast [33], by restricting interaction networks to one specific compartment and then testing the validity of predictions. Unexpectedly, this procedure immediately yielded novel insights into the life-extending mechanisms of dietary restriction [33]. We will identify the shortest paths between connected components, and also calculate network topological properties (node degree and network connectivity, etc.). Networks generated from defined signatures will be merged or intersected to reveal their effects at the level of the interactome.

2.4 Simulations [HAK/JPdM] ^^^^^^^^^^^^^^^^^^^^^^^^^^ The initial simulation we propose to develop will be extremely simple in order to obtain a working prototype, and will then be scaled up substantially to include more information and complexity. Several simplifying assumption will be made at the beginning. Boolean networks are parameter-free, time-discrete and qualitative models of regulatory processes and therefore provide a good starting point [46-48,68]. In this modeling approach, regulatory components are represented by simple switches (ON or OFF). The high degree of abstraction in Boolean models allows for the inclusion of various types of knowledge from the database and literature. Model building starts at a fine-grained level by simulating a small regulatory subsystem, e.g. a single cell. Combining several subsystems allows for the gradual construction of high-level models, e.g. for signal transduction pathways. At this higher level, calculation time can be significantly reduced by condensing the subsystems to functionally identical models of lower detail. Furthermore, we will simulate all molecular entities with Brownian (random-walk) motion in either 3D space (a cell or organelle) or 2D space (a membrane surface), so that interactions can only occur between entities that are spatiotemporally coincident (expressed at the same place [tissue, cell, compartment] and time [developmental stage]) (Table 2). Some relations such as physical interactions may result in the formation of temporary or permanent association if two entities, known to interact, collide. An interaction by collision can also result in a chemical reaction in which two or more coincident molecules react to form one or more products. Each molecule also has a half-life value, which will be initially estimated naively but can later be corrected when measured values are available. Classes that need to be defined in a simulation are: Substance, Reactor, System (compartments in the cell). Properties, such as position, velocity, and color, will be handled as vectors in suitable arrays for parallel computing. The overall logic is written in a high-level language for rapid prototyping and employment, while data-intensive functions are written in a low language (such as shaders or kernels), digested in parallel and called by the simulation logic which is on a high-level.

2.5 Visualization: a 4D atlas of C. elegans aging [JPdM/HAK/RJSR] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ OpenGL (Open Graphical Language) will be used to visualize the simulation. Manual curation is possible in the simulation as the integrated and meta-analyzed data are stored on a single server. Curation can easily be accomplished by changing the attributes of entities, entity interactions, or entity groups treated in bulk. Changes will be marked and can be backtracked to time and user. To enhance reliability and reduce human error, only confirmed changes will be merged with the other data. Visual representations of the simulation are absolutely necessary and also provide additional benefits such as quality-control (QC) feedback for curators and instructional / educational demonstrations. Simulations thus serve multiple functions: they offer a visually appealing curation interface (in addition to web-framework administration), which in turn can be used to modify information in the databases; they are a logical next step to follow from network biology; they provide robust and testable models of biological systems on which many types of information can be mapped; and they can reveal relationships that are in other ways difficult to grasp, such as positional and temporal aspects of interactions and processes, and net effects of complex interactions. OpenCL will be used extensively for physical characterization of simulations, while GLSL (OpenGL Shading Language) will utilize well-established shaders (programs running on the GPU) to create and modulate the graphics.

A major advantage of our model organism is that it is transparent and the developmental lineage and position of each cell can be followed. Video atlases have been constructed in 3D, wherein each cell is named by lineage and tracked over time [http://caltech.wormbase.org/virtualworm/]. The adult hermaphrodite comprises 959 somatic cells. Machine learning can improve on the automated 3D-image annotation of single-cell gene expression [81] or the identification of labeled single molecules [82] and can assist interpretation of molecular changes by recording statistics of the simulation. After establishing an initial ensemble of components, molecular signatures of aging will be mapped to these entities to provide the simulation with quantitative measures of time-dependent changes. Once this point has been reached, we will explore means to efficiently reverse observed age-related changes, by altering the quantity, localization or activity of one or more entities.

Aim 3. Prediction and Testing of Interventions to Extend or Reset Lifespan ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Through simulations developed in Aim 2 we will seek proximal factors limiting lifespan, and mechanisms by which interventions can extend or reset lifespan. To refine our models and to experimentally test the predictions, a reporter panel will be selected and assembled, comprising genes (or their promoters) whose transcript levels are most predictive of age and/or lifespan, fused to GFP. This panel will allow tissue- and age-dependent responses of key system components to be compared over a variety of genetic and dietary contexts [RJSR]. Lifespan and reporter data assessed for interventions predicted to extend survival and/or reverse age-related changes will be used to test the validity of simulations [RJSR], and to iteratively improve them [HK/JPdM].

Although many theories have been proposed to account for aging, there is as yet no theoretical framework to unify all observations into a single model. Initial models will rely on available tissue mapping of gene expression [http://www.wormbase.org], but for key genes implicated in the model, promoter/enhancer fusions to GFP will be acquired from the Orfeome and Promoterome projects, and digital images of these reporters will be analyzed [65] to generate detailed information on tissue-specific expression and its dependence on age, diet, or other interventions. Simulations incorporating these data will be used to formulate hypotheses of interventions that should increase lifespan, and these predictions will then be tested in C. elegans. Specifically, levels of key transcripts will be defined that are characteristic of youthful physiology and that (when measured in a young adult) predict maximal longevity. Simulations will then be used to define feasible interventions that are predicted in silico to reinstate the same molecular parameters associated with youth or extended lifespan in vivo. To test those predictions, RNA interference will be used to knock down expression of specific genes, either globally or in particular tissues, and the effects on lifespan will be tested. To a more limited extent, transgenes may be introduced to overexpress normal or modified genes either globally or in a targeted tissue. More ambitious interventions are beyond the scope of this proposal, but could include (a.) modulation of protein activities via targeted gain-of-function or dominant-negative transgenes; (b.) modification of specific signaling kinases by codon substitution to block or emulate their phosphorylation; and (c.) alteration of ambient temperature, osmolarity or specific nutrients. New data resulting from such experiments will be integrated into the WormAge database (Fig. 1) in order to enhance the predictive power of simulations. Data integration, meta-analysis, simulation and validation will be subsequently repeated in an iterative fashion, as necessary. The profiling and testing for intervention efficacy (with a potential reversal of aging) will be primarily undertaken and coordinated by RJSR.

3.1 Defining features predictive of lifespan or life-extending interventions [JPdM/HK/TL] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Having established an initial feature set of interventions, genes, signatures and changes, a variety of predictive methods (including paradigms from artificial-intelligence research) will be applied to determine which features are the strongest predictors for pro-aging and pro-longevity genes [HK/JPdM]. Moreover, the same tools will be used to define differentially expressed gene-sets that best predict age, lifespan, or the longevity effect of interventions. From this it will be possible to predict processes associated with certain categories of genes/molecules as well as candidate entities [41]. Identification of the strongest predictive features will allow us to estimate confidence for effects of a particular genotype/manipulation (i.e., to calculate a 95% confidence interval for its longevity effect) and to predict the effects of combinations of interventions. Unsupervised learning will cluster data into previously undefined classes, and group interventions based on similarity of changes they elicit in associated reporter data. Having data assembled within a unified schema allows us to employ black-box methods such as neural networks (e.g., PyBrain) to seek predictive features implicit in data assemblies [HK/JPdM].

Predictions will come from many different forms of analysis (1.5, 2.3, 2.4, 2.5), ranging from candidate drivers of aging and longevity-assurance inferred directly from the genome annotated with molecular signatures, to implicating controllers of aging from directed interaction networks and models, to complex feature sets emerging from machine-learning analyses of the integrated data assembly and simulations. First, the most significant implications of each analysis will be validated with the potential to provide important insights of its own. Ultimately, however, all experimental results will be interpreted in the context of prior studies on aging of C. elegans and other animal models [JPdM/RJSR]. Patterns that emerge from several different analyses (in C. elegans and/or from other model species) are likely to constitute the most promising candidates. Those selected for validation will be tested with a view to finding interventions that maximally extend lifespan or hold the potential to reverse aging (3.3 below). From the collective evaluation of simulation data, we anticipate that some specific hypotheses will emerge on which there is broad agreement and will have the highest priority for testing.

3.2 Lifespan assays and molecular profiling to test longevity interventions [RJSR] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Our primary data for testing the efficacy of interventions predicted to extend lifespan will come from semi-automated, moderate-throughput assays using the wMicrotracker in 96-well format to monitor survival. Worms will be subjected to each intervention beginning either in the preceding generation (i.e., the parents of the tested worms) or at the L4/adult molt, because we have found that the efficacy of RNA interference often varies with timing [7-9,83]. They are manually separated from progeny every second day [27,84], or maintained in 2 M FUdR to retard maturation of progeny without affecting lifespan [85] until worms are postgravid (typically day 5 or 6 of adulthood). Worms are then transferred, at 10 per well to 96-well plates containing agar medium seeded with OP50 bacteria, at 10 per well, for automated monitoring of activity with manual confirmation of deaths. For DR studies, worms will be maintained in liquid medium [83] or on an agar surface with varying concentrations of killed bacteria [86].

Because our models require quantitative measures to assess changes in levels of major biological entities, we will use high-throughput methods complemented with more painstaking validations of data that are critical to the most intriguing hypotheses. Further, in view of our objective to test whether age-related changes can be reversed, we will compare interventions begun at varying ages to determine whether any simulation-derived interventions can restore youthful levels to age-dependent parameters. If this proves to be feasible, it would herald a new category of high-throughput intervention screens (of dsRNAs or drugs), that can be performed rapidly on aged animals using (for example) multiple fluorescent-protein reporters for a small panel of age-biomarker genes.

In refining the simulations, such reporters will be imaged at various ages, in diverse longevity mutants, and under different environmental and dietary conditions, to provide tissue- and even cell-specific profiles of protein expression. These data have the potential to greatly improve the accuracy of simulations and to eliminate errors that arise from averaging gene activity across multiple tissues and cell types. Simultaneous acquisition of both gene activities and metabolite levels (derived from bioactive dyes such as DHFC, MitoTracker, redox reporters, etc.) will furnish data that help to bridge the gap between genotype and phenotype in determining lifespan.

The allocation of the research to each applicant is defined explicitly (Table 3).

References

1.    Kulkarni M, Deopujari JY, Purohit HJ (2003) *Synergistic effect of ayurvedic pearl preparation on enhancing effectiveness of antibiotics.* Indian J Exp Biol 40: 831-4.
2.    Gana ML, Kebbouche-Gana S, Touzi A, Zorgani MA, Pauss A, Lounici H, Mameri N (2010) *Antagonistic activity of Bacillus sp. obtained from an Algerian oilfield and chemical biocide THPS against sulfate-reducing bacteria consortium inducing corrosion in the oil industry.* J Ind Microbiol Biotechnol 38: 391-404.
3.    Cande J, Andolfatto P, Prud'homme B, Stern DL, Gompel N (2012) *Evolution of Multiple Additive Loci Caused Divergence between Drosophila yakuba and D. santomea in Wing Rowing during Male Courtship.* PLoS One 7: e43888.
4.    M., Sheldon (2006) *Simulation, Fourth Edition (Statistical Modeling and Decision Science)*.: Academic Press
5.    Orrell, David (2008) *The Future of Everything: The Science of Prediction*.: Basic Books
6.    Kimura A, Sogawa K, Satoh M, Kodera Y, Yokosuka O, Tomonaga T, Nomura F (2012) *The application of a three-step serum proteome analysis for the discovery and identification of novel biomarkers of hepatocellular carcinoma.* Int J Proteomics 2012: 623190.
7.    Shmookler Reis RJ, Xu L, Lee H, Chae M, Thaden JJ, Bharill P, Tazearslan C, Siegel E, Alla R, Zimniak P, Ayyadevara S (2011) *Modulation of lipid biosynthesis contributes to stress resistance and longevity of C. elegans mutants.* Aging (Albany NY) 3: 125-47.
8.    Shmookler Reis RJ, Ayyadevara S, Crow WA, Lee T, Delongchamp RR (2011) *Gene categories differentially expressed in C. elegans age-1 mutants of extraordinary longevity: new insights from novel data-mining procedures.* J Gerontol A Biol Sci Med Sci 67: 366-75.
9.    Tazearslan C, Ayyadevara S, Bharill P, Shmookler Reis RJ (2009) *Positive feedback between transcriptional and kinase suppression in nematodes with extraordinary longevity and stress resistance.* PLoS Genet 5: e1000452.
10.    Murphy CT, Lee SJ, Kenyon C (2007) *Tissue entrainment by feedback regulation of insulin gene expression in the endoderm of Caenorhabditis elegans.* Proc Natl Acad Sci U S A 104: 19046-50.
11.    Cornils A, Gloeck M, Chen Z, Zhang Y, Alcedo J (2011) *Specific insulin-like peptides encode sensory information to regulate distinct developmental processes.* Development 138: 1183-93.
12.    Gudkov AV, Gurova KV, Komarova EA (2011) *Inflammation and p53: A Tale of Two Stresses.* Genes Cancer 2: 503-16.
13.    Iliopoulos D, Jaeger SA, Hirsch HA, Bulyk ML, Struhl K (2010) *STAT3 activation of miR-21 and miR-181b-1 via PTEN and CYLD are part of the epigenetic switch linking inflammation to cancer.* Mol Cell 39: 493-506.
14.    Hatziapostolou M, Polytarchou C, Aggelidou E, Drakaki A, Poultsides GA, Jaeger SA, Ogata H, Karin M, Struhl K, Hadzopoulou-Cladaras M, Iliopoulos D (2011) *An HNF4alpha-miRNA inflammatory feedback circuit regulates hepatocellular oncogenesis.* Cell 147: 1233-47.
15.    Yen K, Le TT, Bansal A, Narasimhan SD, Cheng JX, Tissenbaum HA (2010) *A comparative study of fat storage quantitation in nematode Caenorhabditis elegans using label and label-free methods.* PLoS One.
16.    Leiser SF, Begun A, Kaeberlein M (2011) *HIF-1 modulates longevity and healthspan in a temperature-dependent manner.* Aging Cell 10: 318-26.
17.    Lee J, Kim KY, Lee J, Paik YK (2009) *Regulation of Dauer formation by O-GlcNAcylation in Caenorhabditis elegans.* J Biol Chem 285: 2930-9.
18.    Lee EY, Jeong PY, Kim SY, Shim YH, Chitwood DJ, Paik YK (2009) *Effects of sterols on the development and aging of Caenorhabditis elegans.* Methods Mol Biol 462: 167-79.
19.    Bartholome A, Kampkotter A, Tanner S, Sies H, Klotz LO (2010) *Epigallocatechin gallate-induced modulation of FoxO signaling in mammalian cells and C. elegans: FoxO stimulation is masked via PI3K/Akt activation by hydrogen peroxide formed in cell culture.* Arch Biochem Biophys 501: 58-64.
20.    Perez CL, Van Gilst MR (2008) *A 13C isotope labeling strategy reveals the influence of insulin signaling on lipogenesis in C. elegans.* Cell Metab 8: 266-74.
21.    Gallo M, Park D, Riddle DL (2011) *Increased longevity of some C. elegans mitochondrial mutants explained by activation of an alternative energy-producing pathway.* Mech Ageing Dev 132: 515-8.
22.    Gallo M, Riddle DL (2010) *Regulation of metabolism in Caenorhabditis elegans longevity.* J Biol 9: 7.
23.    Rogers AN, Chen D, McColl G, Czerwieniec G, Felkey K, Gibson BW, Hubbard A, Melov S, Lithgow GJ, Kapahi P (2011) *Life span extension via eIF4G inhibition is mediated by posttranscriptional remodeling of stress response gene expression in C. elegans.* Cell Metab 14: 55-66.
24.    Love DC, Ghosh S, Mondoux MA, Fukushige T, Wang P, Wilson MA, Iser WB, Wolkow CA, Krause MW, Hanover JA (2010) *Dynamic O-GlcNAc cycling at promoters of Caenorhabditis elegans genes regulating longevity, stress, and immunity.* Proc Natl Acad Sci U S A 107: 7413-8.
25.    Ayyadevara S, Tazearslan C, Bharill P, Alla R, Siegel E, Shmookler Reis RJ (2009) *Caenorhabditis elegans PI3K mutants reveal novel genes underlying exceptional stress resistance and lifespan.* Aging Cell 8: 706-25.
26.    Houtkooper RH, Williams RW, Auwerx J (2010) *Metabolic networks of longevity.* Cell 142: 9-14.
27.    Ayyadevara S, Alla R, Thaden JJ, Shmookler Reis RJ (2007) *Remarkable longevity and stress resistance of nematode PI3K-null mutants.* Aging Cell 7: 13-22.
28.    Honjoh S, Nishida E (2011) *Two sides of lifespan regulating genes: pro-longevity or anti-longevity?* J Biochem 149: 381-8.
29.    Brooks KK, Liang B, Watts JL (2009) *The influence of bacterial diet on fat storage in C. elegans.* PLoS One 4: e7545.
30.    Greer EL, Brunet A (2008) *FOXO transcription factors in ageing and cancer.* Acta Physiol (Oxf) 192: 19-28.
31.    Smith ED, Kaeberlein TL, Lydum BT, Sager J, Welton KL, Kennedy BK, Kaeberlein M (2008) *Age- and calorie-independent life span extension from dietary restriction by bacterial deprivation in Caenorhabditis elegans.* BMC Dev Biol 8: 49.
32.    Smith ED, Tsuchiya M, Fox LA, Dang N, Hu D, Kerr EO, Johnston ED, Tchao BN, Pak DN, Welton KL, Promislow DE, Thomas JH, Kaeberlein M, Kennedy BK (2008) *Quantitative evidence for conserved longevity pathways between divergent eukaryotic species.* Genome Res 18: 564-70.
33.    Wuttke D, Connor R, Vora C, Craig T, Li Y, Wood S, Vasieva O, Shmookler Reis R, Tang F, de Magalhaes JP (2012) *Dissecting the gene network of dietary restriction to identify evolutionarily conserved pathways and new functional genes.* PLoS Genet 8: e1002834.
34.    Nikitin AG, Shmookler Reis RJ (1997) *Role of transposable elements in age-related genomic instability.* Genet Res 69: 183-95.
35.    Nicholas A, de Magalhaes JP, Kraytsberg Y, Richfield EK, Levanon EY, Khrapko K (2010) *Age-related gene-specific changes of A-to-I mRNA editing in the human brain.* Mech Ageing Dev 131: 445-7.
36.    Finch CE, Morgan TE, Longo VD, de Magalhaes JP (2010) *Cell resilience in species life spans: a link to inflammation?* Aging Cell 9: 519-26.
37.    Cevenini E, Bellavista E, Tieri P, Castellani G, Lescai F, Francesconi M, Mishto M, Santoro A, Valensin S, Salvioli S, Capri M, Zaikin A, Monti D, de Magalhaes JP, Franceschi C (2010) *Systems biology and longevity: an emerging approach to identify innovative anti-aging targets and strategies.* Curr Pharm Des 16: 802-13.
38.    de Magalhaes JP, Wuttke D, Wood SH, Plank M, Vora C (2011) *Genome-environment interactions that modulate aging: powerful targets for drug discovery.* Pharmacol Rev 64: 88-101.
39.    Li Y, de Magalhaes JP (2011) *Accelerated protein evolution analysis reveals genes and pathways associated with the evolution of mammalian longevity.* Age (Dordr).
40.    Wood SH, Craig T, Li Y, Merry B, de Magalhaes JP (2012) *Whole transcriptome sequencing of the aging rat brain reveals dynamic RNA changes in the dark matter of the genome.* Age (Dordr).
41.    Freitas AA, Vasieva O, de Magalhaes JP (2011) *A data mining approach for classifying DNA repair genes into ageing-related or non-ageing-related.* BMC Genomics 12: 27.
42.    Gupta R, Rao N, Kumar V (2011) *Discovery of error-tolerant biclusters from noisy gene expression data.* BMC Bioinformatics 12 Suppl 12: S1.
43.    Lee T, Desai VG, Velasco C, Reis RJ, Delongchamp RR (2008) *Testing for treatment effects on gene ontology.* BMC Bioinformatics 9 Suppl 9: S20.
44.    Delongchamp R, Lee T, Velasco C (2006) *A method for computing the overall statistical significance of a treatment effect among a group of genes.* BMC Bioinformatics 7 Suppl 2: S11.
45.    Delongchamp RR, Bowyer JF, Chen JJ, Kodell RL (2004) *Multiple-testing strategy for analyzing cDNA array data on gene expression.* Biometrics 60: 774-82.
46.    Maucher M, Kracher B, Kuhl M, Kestler HA (2011) *Inferring Boolean network structure via correlation.* Bioinformatics 27: 1529-36.
47.    Hopfensitz M, Mussel C, Wawra C, Maucher M, Kuhl M, Neumann H, Kestler HA (2011) *Multiscale Binarization of Gene Expression Data for Reconstructing Boolean Networks.* IEEE/ACM Trans Comput Biol Bioinform.
48.    Kestler HA, Wawra C, Kracher B, Kuhl M (2008) *Network modeling of signal transduction: establishing the global view.* Bioessays 30: 1110-25.
49.    de Magalhaes JP, Budovsky A, Lehmann G, Costa J, Li Y, Fraifeld V, Church GM (2008) *The Human Ageing Genomic Resources: online databases and tools for biogerontologists.* Aging Cell 8: 65-72.
50.    Plank M, Wuttke D, van Dam S, Clarke SA, de Magalhaes JP (2012) *A meta-analysis of caloric restriction gene expression profiles to infer common signatures and regulatory mechanisms.* Mol Biosyst 8: 1339-49.
51.    de Magalhaes JP, Curado J, Church GM (2009) *Meta-analysis of age-related gene expression profiles identifies common signatures of aging.* Bioinformatics 25: 875-81.
52.    de Magalhaes JP, Finch CE, Janssens G (2009) *Next-generation sequencing in aging research: emerging applications, problems, pitfalls and possible solutions.* Ageing Res Rev 9: 315-23.
53.    de Magalhaes JP, Costa J, Church GM (2007) *An analysis of the relationship between metabolism, developmental schedules, and longevity using phylogenetic independent contrasts.* J Gerontol A Biol Sci Med Sci 62: 149-60.
54.    de Magalhaes JP, Church GM (2005) *Genomes optimize reproduction: aging as a consequence of the developmental program.* Physiology (Bethesda) 20: 252-9.
55.    Ayyadevara S, Ayyadevara R, Vertino A, Galecki A, Thaden JJ, Shmookler Reis RJ (2003) *Genetic loci modulating fitness and life span in Caenorhabditis elegans: categorical trait interval mapping in CL2a x Bergerac-BO recombinant-inbred worms.* Genetics 163: 557-70.
56.    Ayyadevara S, Ayyadevara R, Hou S, Thaden JJ, Shmookler Reis RJ (2001) *Genetic mapping of quantitative trait loci governing longevity of Caenorhabditis elegans in recombinant-inbred progeny of a Bergerac-BO x RC301 interstrain cross.* Genetics 157: 655-66.
57.    Ebert RH 2nd, Cherkasova VA, Dennis RA, Wu JH, Ruggles S, Perrin TE, Shmookler Reis RJ (1993) *Longevity-determining genes in Caenorhabditis elegans: chromosomal mapping of multiple noninteractive loci.* Genetics 135: 1003-10.
58.    Szumska D, Benes H, Kang P, Weinstein RS, Jilka RL, Manolagas SC, Shmookler Reis RJ (2006) *A novel locus on the X chromosome regulates post-maturity bone density changes in mice.* Bone 40: 758-66.
59.    Parsons CA, Mroczkowski HJ, McGuigan FE, Albagha OM, Manolagas S, Reid DM, Ralston SH, Shmookler Reis RJ (2005) *Interspecies synteny mapping identifies a quantitative trait locus for bone mineral density on human chromosome Xp22.* Hum Mol Genet 14: 3141-8.
60.    Tan AC, Emmerton L, Hattingh HL (2012) *A review of the medication pathway in rural Queensland, Australia.* Int J Pharm Pract 20: 324-39.
61.    Desai VG, Lee T, Delongchamp RR, Leakey JE, Lewis SM, Lee F, Moland CL, Branham WS, Fuscoe JC (2008) *Nucleoside reverse transcriptase inhibitors (NRTIs)-induced expression profile of mitochondria-related genes in the mouse liver.* Mitochondrion 8: 181-95.
62.    Delongchamp RR, Velasco C, Desai VG, Lee T, Fuscoe JC (2008) *Designing toxicogenomics studies that use DNA array technology.* Bioinform Biol Insights 2: 317-28.
63.    Mattfeldt T, Trijic D, Gottfried HW, Kestler HA (2004) *Classification of incidental carcinoma of the prostate using learning vector quantization and support vector machines.* Cell Oncol 26: 45-55.
64.    Kraus JM, Kestler HA (2010) *A highly efficient multi-core algorithm for clustering extremely large datasets.* BMC Bioinformatics 11: 169.
65.    Huth J, Buchholz M, Kraus JM, Schmucker M, von Wichert G, Krndija D, Seufferlein T, Gress TM, Kestler HA (2010) *Significantly improved precision of cell migration analysis in time-lapse video microscopy through use of a fully automated tracking system.* BMC Cell Biol 11: 24.
66.    Muller A, Holzmann K, Kestler HA (2006) *Visualization of genomic aberrations using Affymetrix SNP arrays.* Bioinformatics 23: 496-7.
67.    Kestler HA, Muller A, Gress TM, Buchholz M (2004) *Generalized Venn diagrams: a new method of visualizing complex genetic set relations.* Bioinformatics 21: 1592-5.
68.    Mussel C, Hopfensitz M, Kestler HA (2010) *BoolNet--an R package for generation, reconstruction and analysis of Boolean networks.* Bioinformatics 26: 1378-80.
69.    Kestler HA, Muller A, Kraus JM, Buchholz M, Gress TM, Liu H, Kane DW, Zeeberg BR, Weinstein JN (2008) *VennMaster: area-proportional Euler diagrams for functional GO analysis of microarrays.* BMC Bioinformatics 9: 67.
70.    Weiss G (2003) *Europe wakes up to aging.* Sci Aging Knowledge Environ 2002: ns10.
71.    Kirkwood TB (2003) *The most pressing problem of our age.* BMJ 326: 1297-9.
72.    Butler RN, Miller RA, Perry D, Carnes BA, Williams TF, Cassel C, Brody J, Bernard MA, Partridge L, Kirkwood T, Martin GM, Olshansky SJ (2008) *New model of health promotion and disease prevention for the 21st century.* BMJ 337: a399.
73.    Janes KA, Albeck JG, Gaudet S, Sorger PK, Lauffenburger DA, Yaffe MB (2005) *A systems model of signaling identifies a molecular basis set for cytokine-induced apoptosis.* Science 310: 1646-53.
74.    Heavner BD, Smallbone K, Barker B, Mendes P, Walker LP (2012) *Yeast 5 - an expanded reconstruction of the Saccharomyces cerevisiae metabolic network.* BMC Syst Biol 6: 55.
75.    Karr JR, Sanghvi JC, Macklin DN, Gutschow MV, Jacobs JM, Bolival B Jr, Assad-Garcia N, Glass JI, Covert MW (2012) *A whole-cell computational model predicts phenotype from genotype.* Cell 150: 389-401.
76.    Lund J, Tedesco P, Duke K, Wang J, Kim SK, Johnson TE (2002) *Transcriptional profile of aging in C. elegans.* Curr Biol 12: 1566-73.
77.    Spencer WC, Zeller G, Watson JD, Henz SR, Watkins KL, McWhirter RD, Petersen S, Sreedharan VT, Widmer C, Jo J, Reinke V, Petrella L, Strome S, Von Stetina SE, Katz M, Shaham S, Ratsch G, Miller DM 3rd (2010) *A spatial and temporal map of C. elegans gene expression.* Genome Res 21: 325-41.
78.    Shmookler Reis RJ, Bharill P, Tazearslan C, Ayyadevara S (2009) *Extreme-longevity mutations orchestrate silencing of multiple signaling pathways.* Biochim Biophys Acta 1790: 1075-83.
79.    Dematte L, Prandi D (2010) *GPU computing for systems biology.* Brief Bioinform 11: 323-33.
80.    Foster, Ian (1995) *Designing and Building Parallel Programs: Concepts and Tools for Parallel Software Engineering*.: Addison Wesley
81.    Deguchi T, Itoh M, Urawa H, Matsumoto T, Nakayama S, Kawasaki T, Kitano T, Oda S, Mitani H, Takahashi T, Todo T, Sato J, Okada K, Hatta K, Yuba S, Kamei Y (2009) *Infrared laser-mediated local gene induction in medaka, zebrafish and Arabidopsis thaliana.* Dev Growth Differ 51: 769-75.
82.    Rifkin SA (2011) *Identifying fluorescently labeled single molecules in image stacks using machine learning.* Methods Mol Biol 772: 329-48.
83.    Turner G, Partington M, Kerr B, Mangelsdorf M, Gecz J (2002) *Variable expression of mental retardation, autism, seizures, and dystonic hand movements in two families with an identical ARX gene mutation.* Am J Med Genet 112: 405-11.
84.    Ayyadevara S, Engle MR, Singh SP, Dandapat A, Lichti CF, Benes H, Shmookler Reis RJ, Liebau E, Zimniak P (2005) *Lifespan and stress resistance of Caenorhabditis elegans are increased by expression of glutathione transferases capable of metabolizing the lipid peroxidation product 4-hydroxynonenal.* Aging Cell 4: 257-71.
85.    Van Raamsdonk JM, Hekimi S (2011) *FUdR causes a twofold increase in the lifespan of the mitochondrial mutant gas-1.* Mech Ageing Dev 132: 519-21.
86.    Sutphin GL, Kaeberlein M (2009) *Measuring Caenorhabditis elegans life span on solid media.* J Vis Exp.

Public Abstract

The fundamental basis for organismal aging remains, as Peter Medawar phrased it 60 years ago, an unsolved problem of biology. Our goal is to solve this problem in the roundworm (C. elegans), one the simplest and best characterized animal species with regard to aging. Through a combination of novel genetic, computational and systems biology approaches. The aging process is remarkably plastic; it can be accelerated and delayed by various environmental and genetic perturbations. Whether or not its reversal is possible will be tested in this proposal. Hundreds of genes and many more processes have been associated with lifespan control, yet the mechanisms of aging and anti-aging remain poorly understood. In order to cope with the enormous data in a systematic fashion we will create the first in silico simulation of aging in a multicellular organism which integrates information from the molecular level to whole physiology. Such a framework will guide experimental investigations directed at solving the complex nature of aging and thereby advance our integrated understanding of organismal aging.

Tables and Figures

Tables

Table 1: Data flow ~~~~~~~~~~~~~~~~~~ 1. Data collection & unification + External database parsing + Automated text-mining + Expert knowledge input 2. Integration into adequate database schemas 3. Meta-analysis + Common signatures + Classification + Machine learning 4. Simulation + Refinement + Expert curation 5. Prediction + Unguided (unbiased) + Guided by experts 6. Validation + Lifespan assays + Profiling

Table 2: Interaction determination ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1. Expression at a particular age and/or condition 2. Coexpression in the same tissue 3. Expression in the same cellular compartment

Table 3: Allocation of the research ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - JPM

  • Development of WormAge, a web framework for data integration, exchange, and curation.

  • Integration of relevant data under a unified format

  • Network generation and analysis

  • Interpretation and optimization of models in context with current knowledge

  • RJSR

  • Perform transcriptome analyses for an isogenic panel of longevity mutants (Year 1)

  • Assemble a reporter panel of wild-type worms (outcrossed to a constant background) expressing genes that are the strongest predictors of age and/or lifespan (early in Year 2)

  • Collect fluorescent-reporter images of these strains at varying ages and under varying dietary conditions, transmitted to HAK for image analysis and database deposition (Years 2 - 3)

  • Test simulation-derived predictions by implicated combinations of RNAi knockdown and DR or other environmental manipulation (Year 3)

Table 4: Epistasis of longevity ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1. Synergistic (enhancing the effectiveness) 2. Antagonistic (reducing the activity) 3. No epistasis

.. figure:: http://dgallery.s3.amazonaws.com/wormage_concept.jpg :width: 500 :height: 500

**Figure 1: Concept.** Information parsed from external databases, texted-mined literature and expert input will be gathered into a unified schema within a web framework nominated as WormBase. This resource manages the data integration and curation process and the definition of necessary object-orientated classes. Gene expression profiles which are either publicly available, obtained through collaboration, or generated within this project are meta-analysed to derive molecular signatures. These signatures will be mapped onto respective entities in the heterogenous interaction networks as well as in the simulation which is defined and driven by the integrated data. Algorithms will be trained to machine learn on these data concepts and derive testable hypothesis. Candidates and hypothesis of each analysis will investigated with reporter genes and tested for the effect to reverse the aging process.
Tags: article, proposal


Edit post (Admin)

Comment on This Data Unit