CHAPTER 9 Integrative Genomics of Aging João Pedro de Magalhães and Robi Tacutu Integrative Genomics of Ageing Group, Institute of Integrative Biology, University of Liverpool, Liverpool, UK

OUTLINE

Introduction 263 Finding Needles in Haystacks: Network Approaches and Multi-Dimensional Post-Genome Technologies and Data Integration 272 Biogerontology 264 Construction of Longevity Networks 273 Genome-Wide Approaches and the Topological Features 274 Genetics of Aging and Longevity 264 Network Modularity 276 Surveying the Aging Phenotype on a Multi-Dimensional Data Integration 276 Grand Scale 267 Predictive Methods and Models 278 Challenges in Data Analysis 270 Concluding Remarks 279 Data Integration 271 Acknowledgments 280 Data and Databases 271 References 280

INTRODUCTION processes are complex in the sense that they involve the interplay of multiple genes and The sequencing of genomes has revolution- proteins with each other and with the environ- ized biological and biomedical research. Thanks ment, surveying systems as a whole is impera- to various technologies and approaches that tive to fully comprehending them, and more take advantage of genome sequence knowl- accurately pinpointing how to intervene in edge, researchers can now focus on whole bio- them. Recent breakthroughs in developing logical systems rather than being limited to cheaper and quicker sequencing technolo- studying isolated parts. Because most biological gies have given further power to our capacity

M. Kaeberlein & G.M. Martin (Eds) DOI: http://dx.doi.org/10.1016/B978-0-12-411596-5.00009-5 Handbook of the Biology of Aging, Eighth edition. 263 © 2016 Elsevier Inc. All rights reserved. 264 9. Integrative Genomics of Aging to survey biological systems in a holistic way sources, as this is one of the major challenges with multiple applications in aging research of the post-genome era, and also one of the (reviewed in de Magalhães et al., 2010). In addi- most promising. Various sources of data and tion to genomics, other omics approaches like approaches are discussed in this context. transcriptomics, proteomics, and epigenomics have allowed for a systematic profiling of bio- logical processes and disease states. POST-GENOME TECHNOLOGIES Aging is widely acknowledged as a complex AND BIOGERONTOLOGY process involving changes at various biological levels, interactions between them and feedback There are many open questions in biogeron- regulatory circuits. The underlying mechanis- tology, but arguably most researchers focus tic causes of aging remain a subject of debate, on two key questions (de Magalhães and and it is likely that multiple degenerative pro- Toussaint, 2004b): (i) What are the genetic cesses are involved, including organ-specific determinants of aging, both in terms of longev- processes but also interacting cell- and organ- ity differences between individuals and spe- level communications (Cevenini et al., 2010; de cies differences in aging? (ii) Which changes Magalhães, 2011; Lopez-Otin et al., 2013). While occur across the lifetime to increase vulnerabil- there are simple triggers to complex biological ity, for example, in a person from age 30 to age processes, such as telomere shortening trigger- 70 to increase the chance of dying by roughly ing replicative senescence in human fibroblasts 30-fold? Post-genome technologies may help us (de Magalhães, 2004), most researchers would answer them both. agree that organismal aging involves multiple processes and possibly the interplay between Genome-Wide Approaches and the various causal mechanisms. Likewise, hun- Genetics of Aging and Longevity dreds of genes have been associated with aging in model organisms (Tacutu et al., 2013), and Understanding human phenotypic variation yet the pathways involved are complex and in aging and longevity has been a long-term often interact in nonlinear ways (de Magalhães research goal. Studies in twins have shown that et al., 2012). One hypothesis is that aging and longevity in humans has a genetic component, longevity cannot be fully understood by stud- and the heritability of longevity has been esti- ying individual components and processes mated at approximately 25% (Christensen et al., (Cevenini et al., 2010). To understand aging we 2006). If we could identify genetic variants asso- must then account for the intrinsic complexity ciated with exceptional human longevity, these of biological systems. would likely be suitable for drug discovery (de Our goal in this chapter is to review poten- Magalhães et al., 2012). In 1994, APOE was asso- tial large-scale technologies in the context of ciated with longevity in a French population aging and longevity research and how data (Schachter et al., 1994). The sequencing of the can be analyzed and integrated to advance our human genome in 2001 allowed for much more understanding of these complex processes. We powerful whole-genome genotyping platforms first review the major technologies available capable of surveying hundreds of thousands for researchers to survey biological systems of genetic variants in a cost-effective way (de in a systematic fashion and their applications Magalhães, 2009). In spite of these recent tech- to advance the biology and genetics of aging, nological advances, the genetics of human lon- discuss issues in data analysis and statistics, gevity remains largely misunderstood. Several and discuss data integration between different genome-wide association studies (GWAS) have

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS Post-Genome Technologies and Biogerontology 265

FIGURE 9.1 Exponential growth in sequencing capacity as reflected in the dropping costs of sequencing from 2001 to 2013. Source: NHGRI (http://www.genome.gov/sequencingcosts/). been performed with thousands of individuals, with confidence with longevity, our understand- with largely disappointing results. For exam- ing of the genetics of longevity lags behind ple, one recent landmark study involving sev- our understanding of the genetics of complex eral European populations with a total of over age-related diseases, in itself made difficult by 2000 nonagenarian sibling pairs identified only numerous factors like multiple genes with small APOE as associated with longevity (Beekman effects. Intrinsic difficulties in longevity studies et al., 2013); and although APOE has been con- (e.g., lack of appropriate controls) or because sistently associated with longevity, it only longevity is a more complex trait may explain modestly explains the heritability of longevity. why our understanding of the heritability of GWAS focused on complex diseases and pro- longevity is still poor (de Magalhães, 2014). cesses have been on many occasions equally An even greater source of variation in aging disappointing to date, suggesting that common and longevity than that observed between genetic variants have a modest contribution to humans is observed across species. We know longevity and complex diseases (Manolio et al., that mice, for example, age 25–30 times faster 2009). than human beings, even under the best envi- The falling costs of DNA sequencing (Figure ronmental conditions (Finch, 1990). Even when 9.1) means that sequencing a human genome compared to chimpanzees, our closest living is rapidly becoming affordable. Therefore, in relative whose genome is about 95% similar the coming years researchers will move from to our own, aging is significantly retarded in genotyping platforms based on known genetic humans (de Magalhães, 2006). Therefore, there variants to genome sequencing of thousands of must be a genomic basis for species differ- individuals. It is possible that this will reveal ences in aging, and again the dropping costs rare variants with strong effects on longevity, of sequencing have permitted much more as has been predicted to be the case for ­complex affordable de novo sequencing of genomes diseases (Manolio et al., 2009). Nonetheless, con- (de Magalhães et al., 2010). For example, the sidering that only APOE has been ­associated sequencing of long-lived species, such as the

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS 266 9. Integrative Genomics of Aging naked mole-rat and bats (Keane et al., 2014; Aging is a particularly difficult process to Kim et al., 2011; Zhang et al., 2013b) can pro- unravel because it is much harder to study in vide candidate genes for selection in long- humans than most other processes and diseases. lived species, and it is interesting to observe Observational studies have been conducted but that genes involved in DNA damage responses are extremely time-consuming, and clinical tri- and repair have emerged from such studies (de als for longevity itself are nearly impossible, Magalhães and Keane, 2013). even though they can be performed for specific In addition to the analysis of genomes from age-related pathologies (de Magalhães et al., long-lived species, comparative analyses of 2012). Therefore most biogerontologists rely on genomes from species with different lifespans model systems: human cells; unicellular organ- are also beginning to provide further candi- isms such as the yeast Saccharomyces cerevisiae; date genes for a role in aging. We developed a the roundworm Caenorhabditis elegans; the fruit method to identify candidate genes involved in fly Drosophila melanogaster; rodents and in par- species differences in aging based on detecting ticular mice (Mus musculus) and rats (Rattus nor- proteins with accelerated evolution in multiple vegicus). The small size and short life cycles of lineages where longevity is increased (Li and de these organisms—even mice do not commonly Magalhães, 2013). Our results revealed approxi- live more than 4 years—make them inexpen- mately 100 genes and functional groups that sive subjects for aging studies, and the ability are candidate targets of selection when longev- to genetically manipulate them gives research- ity evolves (Li and de Magalhães, 2013). These ers ample opportunities to test their theories include DNA damage response genes and the and unravel molecular and genetic mechanisms ubiquitin pathway and thus provide evidence of aging. that at least some repair systems were selected The aforementioned traditional biomedi- for, and arguably optimized, in long-lived spe- cal model organisms are widely used in other cies. Other labs have developed methods aimed fields and not surprisingly a variety of tools are at discovering genes associated with longevity available to study them, and recently many of either by focusing on genes showing a stronger these powerful tools have taken advantage of conservation in long-lived species (Jobson et al., omics approaches. While the genetics of aging 2010) or by searching for protein residues that was initially unraveled using traditional genetic are conserved in long-lived species but not in approaches (reviewed in Johnson, 2002), large- short-lived ones (Semeiks and Grishin, 2012). scale forward genetic screening approaches now Because all these methods are conceptually dif- allow for hundreds of genes to be tested simul- ferent from each other, little overlap has been taneously for phenotypes of interest, including observed in the results. Nonetheless, it seems longevity and age-related traits. Genome-wide that genetic alterations contributing to the screens for longevity have been performed evolution of longevity in mammals have com- (McCormick and Kennedy, 2012), in particular mon patterns (or signatures) that are detect- in worms (Hamilton et al., 2005; Hansen et al., able using cross-species genome comparisons, 2005; Samuelson et al., 2007). Hundreds of genes though much work remains in order to improve have been associated with life extension in this the signal-to-noise ratio of these methods. One way, although the overlap between these studies caveat of these studies is the lack of experimen- has been smaller than expected. Another obser- tal validation, and thus all of these genes must vation from these screens is that it seems that be seen as candidates. Given the declining costs the most important pathways that modulate of sequencing we can expect many more such lifespan when disrupted in worms (and pos- studies in the near future. sibly in model organisms) have been identified

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS Post-Genome Technologies and Biogerontology 267 by now, even though there is still ample room One of the goals of biogerontology is to to identify individual components. Although develop interventions that postpone degen- usually more laborious, screens for genes affect- eration, preserve health and extend life (de ing lifespan have also been performed in yeast, Magalhães, 2014). Large-scale drug screening is including for replicative lifespan (Smith et al., now widespread in the pharmaceutical industry 2008), chronological lifespan (Fabrizio et al., (Macarron et al., 2011). While life-extension is 2010) and using pooled screen approaches harder and more expensive to assay than targets (Matecic et al., 2010). Technical limitations in in high-throughput screening, systematic screens flies impede screens at the genome-wide level, for life-extending compounds are now a dis- but lifespan screens have been performed using tinct possibility. Petrascheck et al. assayed 88,000 the P-element modular-misexpression system chemicals for the ability to extend worm lifespan (Paik et al., 2012) and Gene Search misexpres- (Petrascheck et al., 2007); while the success of sion vector system (Funakoshi et al., 2011). Costs this approach was modest (only 115 compounds and lack of mutant libraries prevent large-scale significantly extended lifespan and only 13 more screens in mice, although large-scale knock- than 30%), it provides proof-of-concept for large- out mouse repositories are being established scale screens in the context of life-extending like the Knockout Mouse Project (https:// drugs. Further investigations of drug-mediated www.komp.org/) and the International worm longevity, using a similar protocol, even Mouse Phenotype Consortium (https://www. if with a smaller compounds library of known mousephenotype.org/); a large-scale profil- or suspected mammalian targets (many already ing of mouse mutants for aging-related pheno- approved for use in humans), revealed 60 prom- types is also being conducted in the Harwell ising drugs, which might provide beneficial Aging Screen (http://www.har.mrc.ac.uk/ effects on aging in mammals (Ye et al., 2014). research/large-scale-functional-genomics/ harwell-ageing-screen). Surveying the Aging Phenotype on a A variety of genome-wide screens have Grand Scale also been performed in vitro, in particular using RNAi-based technologies (Echeverri In addition to understanding the genetic and Perrimon, 2006; Moffat and Sabatini, basis for phenotypic variation in aging and 2006; Mohr et al., 2010). These include screens longevity, it is also crucial to elucidate the focused on traits of interest for aging and changes that contribute to age-related degen- longevity. For example, screens for cell lifes- eration. Several age-related changes have been pan have been performed in human fibro- described and historically this focused on blasts revealing that senescent cells activate broad physiological and morphological aspects a self-amplifying secretory network involv- and the molecular and biochemical changes ing CXCR2-binding chemokines (Acosta et al., for which assays existed. Thanks to genome- 2008). A variety of readouts can be employed wide approaches we can now survey the aging to assay for specific traits. For instance, screens phenotype with unprecedented detail (de have been performed for genes modulating Magalhães, 2009; Valdes et al., 2013). In particu- resistance to oxidative stress in mammalian lar, advances in transcriptomics have allowed cells (Nagaoka-Yasuda et al., 2007; Plank et al., researchers to survey the expression levels of all 2013) and antioxidant responses (Liu et al., genes in the genome in a single, relatively inex- 2007). The possibilities are immense and pro- pensive, experiment. vide another large-scale tool for deciphering One major breakthrough in transcriptomics biological processes. was the development of the microarray, which

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS 268 9. Integrative Genomics of Aging allows for the quantification of all annotated unknown genes, including non-coding genes genes simultaneously. Briefly, this led in the and genes not yet annotated in genome data- past 15 years to a large number of gene expres- bases (Wood et al., 2013). Although it is excit- sion profiling studies of aging (de Magalhães, ing that many changes were observed in the 2009; Glass et al., 2013; Lee et al., 1999; Zahn so-called “dark matter” transcripts, because et al., 2007). In a sense, however, these have most of these are not annotated or have lit- been disappointing in that relatively few genes tle information, follow-up is complicated; this are differentially expressed with age in most emphasizes the need to study the new genomic tissues and few insights have emerged. As elements that may be phenotypically impor- an exception, Zahn et al. observed a degree of tant. In this context, large-scale efforts, such as coordination in age-related changes in gene ENCODE which aims to identify all functional expression. In mice different tissues age in a elements in the human genome (Dunham et al., coordinated fashion so that a given mouse may 2012), are crucial to annotate and elucidate the exhibit rapid aging while another ages slowly function of all genomic elements. across multiple tissues (Zahn et al., 2007). In A number of studies have also focused addition, our 2009 meta-analysis of aging gene on profiling gene expression changes in life- expression studies revealed a conserved molec- extending interventions or in long-lived strains ular signature of mammalian aging across (de Magalhães, 2009; Lee et al., 1999), as well organs and species consisting of a clear activa- as in short-lived and/or progeroid animals. tion of inflammatory pathways accompanied For example, a large number of studies have by a disruption of collagen and mitochondrial focused on caloric restriction (CR) to iden- genes (de Magalhães et al., 2009). This molecu- tify specific genes and processes associated lar signature of aging maps well into estab- both with CR and whose age-related change lished hallmarks of aging (Lopez-Otin et al., is ameliorated in CR. In contrast to studies of 2013). It should be noted, however, that tran- aging, CR studies have revealed substantial scriptional changes during aging may represent gene expression changes, some of which can responses to aging rather than underlying caus- be associated with specific pathways and pro- ative mechanisms and thus their interpretation cesses (Lee et al., 1999; Tsuchiya et al., 2004). is not straightforward. A meta-analysis of gene expression studies of The dropping costs of sequencing have CR revealed a number of conserved processes also allowed for gene expression profil- associated with CR effects like growth hormone ing approaches that are digital in nature, signaling, lipid metabolism, immune response, as opposed to microarrays that are analog. and detoxification pathways (Plank et al., 2012). Sequencing the transcriptome, usually referred In another study, midlife gene expression pro- to as RNA-seq, allows for unprecedented accu- filing of mice of different lifespans due to dif- racy and power. A number of recent reviews ferent dietary conditions revealed a possible have focused on the advantages of RNA-seq as contribution of peroxisome to aging, which compared to microarrays (de Magalhães et al., was then tested experimentally in invertebrates 2010; Mortazavi et al., 2008; Wang et al., 2009b), (Zhou et al., 2012). Arguably, gene expression and it is very clear that RNA-seq has a supe- profiling of manipulations of aging has been rior dynamic range and provides more data more successful in providing insights than pro- than microarrays. Our lab performed one of the filing of aging per se. first RNA-seq profiling experiments in the con- Technological and methodological advances text of aging, which revealed gene expression promise to allow even more powerful sur- changes in the rat brain in various previously veys of the molecular state of cells. Ribosome

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS Post-Genome Technologies and Biogerontology 269 profiling is one recent approach, also based on methylation and ChIP-Seq for studying histone next-generation sequencing platforms, it con- modifications, provide the tools for researchers sists of sequencing ribosome-protected mRNA to study the epigenetics of aging (de Magalhães fragments. Compared to RNA-seq using total et al., 2010). As such, the epigenome is yet mRNA, ribosome profiling has the advantage another layer of genomic regulation that can be that it is surveying active ribosomes and thus studied in a high-throughput fashion across the can be used to quantify the rate of protein syn- lifespan and in manipulations of longevity. thesis, which is thought to be a better predic- For all the success of transcriptomics, pro- tor of protein abundance (Ingolia et al., 2009). teins are of course the actual machines of life Advances in sequencing technology have also and the correlation between transcripts and allowed for quantitative surveys of changes at protein levels is not perfect. Transcriptomics the DNA level, including quantifying muta- provides a snapshot of transcriptional responses tion accumulation with age in the genome but in the context of aging we need proteom- and at the level of the mitochondrial genome ics to really assay what changes occur with (reviewed in de Magalhães et al., 2010). One age. Proteomics approaches are still limited, recent study found an age-related increase in however, in that they do not allow a compre- human somatic mitochondrial mutations incon- hensive survey of the proteome in a single sistent with oxidative damage (Kennedy et al., experiment (de Magalhães, 2009). There have 2013). Another study in aging mice found no been some advances, though the number of increase in mitochondrial DNA point mutations proteins surveyed is often small compared to or deletions, questioning whether these play a transcriptional profiling. For example, pro- role in aging (Ameur et al., 2011). tein profiling of aging has been performed in Another level of changes during the life the mouse heart, revealing 8 and 36 protein course comes from epigenetics. These are her- spots whose expression was, respectively, up- itable changes that are not caused by changes regulated and downregulated due to aging in the DNA sequence. Large-scale profiling of (Chakravarti et al., 2008); comparable results in epigenetic changes with age is now becoming terms of number of proteins were also found in more common, and with the dropping costs of the mouse brain (Yang et al., 2008). Insights can sequencing will no doubt become even more be gained, however, and, for instance, proteome widespread. It is clear that epigenetic changes, profiling of aging in mice kidney revealed func- like methylation, are associated with age as tional categories associated with aging related well as with age-related diseases (Johnson et al., to metabolism, transport, and stress response 2012). For example, two recent studies found (Chakravarti et al., 2009). epigenetic (methylation) marks highly predic- Another emerging approach to profile tive of chronological age in humans (Hannum age-related changes involves surveying the et al., 2013; Horvath, 2013). There is still debate metabolome. One study compared metabolic concerning the causality of epigenetic changes, parameters of young and old mice, which and whether they are causes or effects of age- was then integrated with gene expression related degeneration; despite their predictive and biochemical data to derive a metabolic value as biomarkers, epigenetic signatures have footprint of aging (Houtkooper et al., 2011). thus far been uninformative concerning causal Another study determined the sera metabo- mechanisms of aging. Modern approaches lite profile of mice of different ages and differ- that allow the epigenome to be surveyed on a ent genetic backgrounds to derive a metabolic genome-wide scale, however, such as methyl- signature that predicts biological age in mice DNA immunoprecipitation for surveying DNA (Tomas-Loba et al., 2013). In humans, a panel of

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS 270 9. Integrative Genomics of Aging

22 metabolites was found to significantly corre- sequencing approaches there is still no gold late with age and with age-related clinical con- standard for the bioinformatics and statisti- ditions independent of age (Menni et al., 2013). cal analysis. Modest alterations in statistical Taken together, these large-scale approaches parameters, for which there is no established provide researchers with tools to survey biolog- standard, can result in significant changes in ical systems in great detail. results. For example, it is important to mention that microarray platforms for gene expression profiling are at present much quicker in terms CHALLENGES IN DATA ANALYSIS of data analysis than approaches based on next- generation sequencing; because microarrays Although large-scale omics approaches facili- have been used for longer, standard methods tate a broad array of studies and provide an are available for them and this is not yet the incredible amount of data, the sheer volume of case for RNA-seq. Researchers planning experi- data generated creates challenges in turning the ments need to carefully balance the advantages data into meaningful results and novel insights. of the latest next-generation sequencing plat- From a statistical perspective, the large-scale forms with the price and simpler bioinformatics approaches also increase the chances of false and statistics of array-based platforms. hits that need to be accounted for when ana- One major and long-recognized problem of lyzing and interpreting the results. The uncer- large-scale approaches is multiple hypothesis tainties concerning potential false results in testing. Even a low-density microarray plat- large-scale approaches emphasize the need for form with a few hundred genes is testing for further experimental validation using a differ- effects a few hundred times, which by chance ent, usually low-scale, approach. In gene expres- will generate false positives. Modern genomic sion studies, qPCR validation is usually used as approaches, for example in GWAS that survey the gold standard (Derveaux et al., 2010). Some millions of SNPs, must adequately cope with types of studies, like genetic association stud- this problem to generate biologically relevant ies of longevity, are not simple to validate, and results. A standard way of dealing with multi- often depend on further studies in other popula- ple hypothesis testing is the Bonferroni correc- tions, which may or may not be feasible. tion, in which the p-value cutoff (typically 0.05) In a sense, the bottleneck in research using is divided by the number of hypotheses being post-genome technologies is moving away tested (e.g., for an array with 20,000 genes use from generating data toward interpreting 0.05/20,000 as cutoff). Bonferroni correction data. As an example, a single 11-day run from can be deemed as too stringent, and alternative an Illumina HiSeq platform generates up to methods for correcting for false positives have 600 Gb of data, which must be stored, pro- been developed (Storey and Tibshirani, 2003). cessed, quality-controlled, and analyzed. This Benjamini correction is also widely used, and means that the standard experiment using next- is less stringent and equally straightforward to generation sequencing platforms must account calculate (Benjamini and Hochberg, 1995). False for a substantial amount of time for the bioin- discovery rates estimates based on simulations formatics and statistical processing of the data. and scrambling of data have also been widely Although several software tools exist now for used, including by our lab (de Magalhães et al., this dry lab work, labs not experienced with 2009; Plank et al., 2012, 2013), and although it bioinformatics might struggle to develop a suit- requires some customization to the specific able pipeline and have to rely on core facilities, experiments, it provides an estimate of false collaborators, or commercial services. Another positives based on real data captured from the problem is that for many next-generation experiment.

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS Data Integration 271 DATA INTEGRATION and can help in a number of analyses. Although the field of biogerontology has seen a slower As mentioned previously, the recent shift increase in integrative systems biology, a series in biological research toward large-scale of resources specific to aging have also been approaches has resulted in the capacity to gen- created in recent years (Table 9.1), in particular erate huge amounts of data, much of which is in the context of our Human Ageing Genomic publicly available. These data, however, are Resources, which are arguably the benchmark in most cases heterogeneous and obtained in the field (Tacutu et al., 2013). at different timescales and biological levels. It should also be noted that although each of Moreover, differences also often exist due to these databases acts as a stand-alone resource, platform and methodology diversity. Still, if focusing on certain facets of aging, in many cases our aim is to obtain a global picture of complex they also show common patterns. For example, processes, such as aging and most age-related in the Human Ageing Genomic Resources, there diseases, we have to develop the computational are many genes that can be found in two or more methods and tools that allow us to integrate and databases (Figure 9.2), hence also increasing the analyze these diverse data. In this section, we confidence of their association to aging. give an overview of the online resources cur- Similarly, a number of databases for age- rently available for aging research and discuss related diseases have been developed, though some of the studies that aim to integrate and the quality and type of data varies greatly. For analyze various types of data. This can be used example, there are many very good databases on its own using public databases or in combi- for cancer, while the number of database for nation with data from one’s own experiment(s). heart diseases is still limited. Below, a non- exhaustive list of databases for age-related dis- Data and Databases eases is provided (Table 9.2). While some of the resources presented above Before diving into aging-specific resources, and below integrate data related to more than it should be mentioned that one important one facet of aging and/or age-related diseases, prerequisite step for data integration, the the concept of multi-dimensional data integra- existence of databases, has already seen a tre- tion, at least at the level of aging- and disease- mendous expansion in recent years and contin- specific databases, is still in its infancy and the ues to develop at increasing speeds. Currently, task is usually left to the researchers perform- there are a number of databases, for humans ing integrative analyses. Some large resources, and model organisms, which host a plethora however, like NCBI and Ensembl, integrate of information available in a standardized, different types of data and are of course major computational-retrievable and usable form resources for data integration. (in many cases these data are even manu- One other aspect that should be kept in mind ally curated to improve quality). These data- is that sometimes even the amount of high- bases provide access both to a wide range of throughput information for only one type of -omes (including genomes, transcriptomes, data may pose computational challenges, both in proteomes, epigenomes, interactomes, reac- terms of handling and analyzing. Consequently, tomes, etc.) and to a multitude of functional integrating and analyzing data from multiple data (including biological processes, molecu- sources will result in an even bigger challenge, lar functions, appurtenance to molecular path- the complexity increasing in most cases in a ways, etc.). Obviously, integrating this type of nonlinear fashion, and as such data integration information with aging-specific data leads to a comes at a cost: the haystack in which the nee- more holistic perspective of the aging process dles have to be found increases exponentially.

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS 272 9. Integrative Genomics of Aging

TABLE 9.1 List of Major Online Databases and Resources Related to Aging

Name (Citation) Web address Short description

AnAge (Tacutu et al., 2013) http://genomics.senescence. Aging, longevity and life history information in animals info/species/ Comparative Cellular and http://genomics.brocku.ca/ Database with cellular and molecular traits from Molecular Biology of Longevity ccmbl/ vertebrate species collected to identify traits correlated Database (Stuart et al., 2013) with longevity GenAge (Tacutu et al., 2013) http://genomics.senescence. Genes associated with longevity and/or aging in model info/genes/ organisms and candidate aging-related human genes GenDR (Wuttke et al., 2012) http://genomics.senescence. Genes associated with dietary restriction both from info/diet/ mutations and gene expression profiling AgeFactDB (Huhne et al., 2014) http://agefactdb.jenage.de/ Observations on the effect of aging factors on lifespan and/or aging phenotype Lifespan Observations DB http://lifespandb.sageweb.org/ Data on the lifespan effects of interventions by genetic (Olsen and Kaeberlein, 2014) engineering, chemical compounds, and environmental effects NetAge (Tacutu et al., 2010a) http://netage-project.org/ Networks (protein–protein interactions and miRNA regulation) for longevity, age-related diseases, and associated processes AGEMAP (Zahn et al., 2007) http://cmgm.stanford. Gene expression database for aging in mice edu/~kimlab/aging_mouse Digital Ageing Atlas http://ageing-map.org/ Database of molecular, physiological, and pathological (Craig et al., 2015) age-related changes GiSAO.db (Hofer et al., 2011) https://igbbelenus.tugraz.at/ Genes associated with cellular senescence, apoptosis, gisao_web/ and oxidative stress LongevityMap http://genomics.senescence. Database of human genetic variants associated with (Budovsky et al., 2013) info/longevity/ longevity

Finding Needles in Haystacks: Network could be relatively small, the authors argued Approaches and Multi-Dimensional Data that the integrative contribution of defec- Integration tive mitochondria, aberrant proteins, and free radicals, taken together, could explain many With the expansion of large-scale approaches, of the major changes that occur during aging. and the inevitable increase in age-related data Although, since it was first proposed, many available, new hypotheses of aging trying to other types of aging factors have been taken integrate multi-dimensional information have into consideration (and the strife to integrate been developed. More than 15 years ago, the more will probably continue), the “Network idea that aging was caused not simply by the theory of aging” might have been the onset of failure of individual components, but rather studying aging in a holistic way. by a network of parallel and gradual dysregu- One emerging discipline, network biology, lations, was proposed (Kirkwood and Kowald, provides a conceptual framework to study the 1997). While the effects of each individual event complex interactions between the multiple

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS Data Integration 273

2011; de Magalhães et al., 2012; Ideker and Sharan, 2008; Soltow et al., 2010). Common topics addressed by these approaches include network construction for aging/longevity or various related conditions, analysis of topologi- cal features, finding functional submodules, etc.

Construction of Longevity Networks Initial attempts to construct longevity net- works date back more than 10 years. As a first step toward the construction of a human aging network, we used genes previously associ- ated with aging and their interacting partners, in a “guilt-by-association” methodology, to FIGURE 9.2 Venn diagram for the genomic databases in construct networks related to DNA metabo- the Human Ageing Genomic Resources. lism and the GH/IGF-1 pathway. We further suggested that among the interacting partners of genes associated with aging there could components of biological systems (Barabasi also be other genes that are involved in aging. et al., 2011). In network biology, a network is Additionally, functional analysis of the net- defined as a set of nodes (as a mathematical work revealed that many of the genes which model for genes, proteins, metabolites, etc.) are important during development might also with some node pairs being connected through regulate the rate of aging (de Magalhães and directed/asymmetric or undirected/symmetric Toussaint, 2004a). edges (as a model for physical interactions, co- One central question in aging research is expression relationships, metabolic reactions, whether genes and pathways associated with etc.). Depending on the type of components aging and longevity are evolutionarily con- and the nature of the interactions that are ana- served. For example, above is a schematic rep- lyzed, there is currently a large variety of net- resentation of longevity protein interaction work types that can be constructed; perhaps the networks across model organisms (Figure 9.3). most used being protein interaction networks, However, the question of relevance arises: are gene regulatory networks, co-expression net- aging-related data in one species also relevant in works, and metabolic networks. another species? This is an important issue since With regard to aging research, the idea of at times the data available in different species analyzing many longevity/aging determinants could be used complementarily. Results so far at the same time has been pushed forward, suggest that genes whose manipulation results mostly due to the accumulating knowledge in a lifespan effect tend to be highly evolution- about the genetic determinants of aging (de arily conserved across divergent eukaryotic Magalhães and Toussaint, 2004a; Tacutu et al., species (Budovsky et al., 2007; de Magalhães 2013), but also by the development of bioinfor- and Church, 2007; Smith et al., 2008). Moreover, matics tools pertaining to network biology like while not universal, some empirical data sug- Cytoscape (Saito et al., 2012). Not surprisingly, gest that the effect on longevity of many of network-based approaches have been increas- these genes is also conserved (Smith et al., 2008). ingly used to study aging and age-related dis- As such, it is not completely senseless to inte- eases (for recent reviews see Barabasi et al., grate longevity-associated genes (LAGs) from

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS 274 9. Integrative Genomics of Aging

TABLE 9.2 selected Online Databases and Resources Related to Age-Related Diseases

Name (Citation) Web address Short description

OMIM (OMIM, 2014) http://www.omim.org/ Online Mendelian Inheritance in Man (part of NCBI) Catalog of Published Genome- https://www.genome.gov/26525384 NHGRI catalog of GWAS Wide Association Studies (Hindorff et al., 2009)

Genetic Association Database http://geneticassociationdb.nih.gov/ Archive of human genetic association (Zhang et al., 2010) studies of complex diseases and disorders AlzGene (Bertram, et al., 2007) http://www.alzgene.org/ Database with genetic resources for Alzheimer’s disease T2D-Db (Agrawal et al., 2008) http://t2ddb.ibab.ac.in/ Database of molecular factors involved in the pathogenesis of type 2 diabetes CADgene (Liu et al., 2011) http://www.bioguo.org/CADgene/ Gene resource for coronary artery disease The Cancer Gene Census http://cancer.sanger.ac.uk/cancergenome/ Catalog of genes for which mutations have (Futreal et al., 2004) projects/census/ been causally implicated in cancer The Cancer Genome Atlas http://cancergenome.nih.gov/ Portal providing access to cancer-related large-scale data from the NCI and NHGRI TSGene (Zhao et al., 2013) http://bioinfo.mc.vanderbilt.edu/TSGene/ Tumor Suppressor Gene Database Progenetix (Cai et al., 2014) http://www.progenetix.org/cgi-bin/ Copy number abnormalities in human pgHome.cgi cancer from comparative genomic hybridization experiments MethyCancer (He et al., 2008) http://methycancer.psych.ac.cn/ Human DNA methylation and cancer PubMeth (Ongenaert et al., 2008) http://www.pubmeth.org/ Cancer methylation database LncRNADisease (Chen et al., http://www.cuilab.cn/lncrnadisease Long non-coding RNA and disease 2013) associations HMDD 2.0 (Li et al., 2014) http://cmbi.bjmu.edu.cn/cui/ Experimentally supported human microRNA and disease associations

multiple species. Using this premise, it was then characteristics (i.e., studying the way in shown that the human orthologs of LAGs from which the nodes and edges of a network are model organisms, together with their interact- arranged). Particular focus has been on scale- ing partners, could act in a cooperative manner free networks, a very common type of net- and form a continuous protein–protein interac- work among social and biological networks. tion network, termed the Human Longevity The scale-free topology means that the nodes Network (Budovsky et al., 2007). in the network have a connectivity distribution p(k) given by a power-law function k−γ, where Topological Features p(k) is the probability that a certain node has One important aspect in network biol- exactly k edges, and γ is the degree exponent, a ogy is the analysis of a network’s topological parameter value which for most of the studied

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS Data Integration 275

FIGURE 9.3 Cross-species view on longevity networks. (A) Worm longevity network. (B) Fly longevity network. (C) Mouse longevity network. (A)–(C) Networks include LAGs from the GenAge database (build 17) and their interacting part- ners. Known protein–protein interactions were retrieved from the BioGRID database, release 3.2.105 (Stark et al., 2006). Dark/light colors depict LAGs and LAG-interacting partners. The number of nodes in each network is summarized in the table below. Species LAGs in GenAge LAGs with interactions Longevity network LAGs in the network

Worm 741 332 1359 314 Fly 140 116 1220 113 Mouse 112 78 763 72

networks is usually between 2 and 3 (Barabasi networks is beyond the scope of this chapter, and Albert, 1999). The aforementioned Human it should be mentioned that the properties of Longevity Network has a scale-free topology, scale-free networks confer some net advantages with a high contribution of hubs (highly con- in solving cellular tasks. For example, this type nected genes) to the overall connectivity of the of architecture permits an efficient local dissipa- network. Interestingly, almost all of the hubs in tion of external perturbations, while at the same the longevity network had been reported previ- time reliably transmitting signals (and dis- ously to be involved in at least one age-related criminating against noise) between distant ele- pathology (Budovsky et al., 2007), suggesting a ments of the network (Csermely and Soti, 2006). link between diseases and the mechanisms reg- Additionally, the scale-free property offers an ulating longevity. unexpected degree of robustness, maintaining The scale-free design can be found in a wide the ability of nodes to communicate even under range of molecular and cellular systems, largely extremely high fault rates, by minimizing the governing their internal organization (Barabasi effect of random failures on the entire network and Oltvai, 2004), and it appears to have been (Albert et al., 2000; Wagner, 2000). also favored by evolution (Oikonomou and On one hand, analyzing the aging/longev- Cluzel, 2006). Although a more detailed dis- ity networks can provide a framework for the cussion about the evolvability of complex conceptualization of the aging process and

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS 276 9. Integrative Genomics of Aging may reveal fundamental traits and constraints found that aging is associated with a limited of biological systems. On the other hand, net- number of modules which are interlinked works can help in assessing the importance of through genes more likely to affect aging/lon- genes in a certain process. For example, dif- gevity (Xue et al., 2007). ferentiating between the hubs of a longevity network and all other nodes is often a very Multi-Dimensional Data Integration attractive way of reducing a candidate list. This Age-related changes can be found at many approach comes as no surprise, as some com- levels (expression changes, post-translational ponents of a cellular network are more impor- modifications, cross-linking or alterations in tant than others with regard to aging. It was protein interactions), yet integration of multi- previously shown that LAGs in model organ- dimensional data is still in its early stages. isms have a higher average connectivity, with Attempts to integrate protein–protein interac- many being network hubs (Budovsky et al., tion networks with transcriptional data have 2007; Ferrarini et al., 2005; Promislow, 2004). already been made with relative success. As Moreover, it has been established that there is partly mentioned previously, a new analytic a positive correlation between a protein’s con- method permitting the integration of both tran- nectivity and its degree of pleiotropy, an ele- scriptome and interactome information has vated degree being common among proteins been employed to study network modularity associated with senescence (Promislow, 2004). in aging (Xue et al., 2007). In another study, a As such, it makes sense in choosing highly con- human protein interaction network for longev- nected longevity candidates. Still, it should also ity was used in conjunction with transcriptional be kept in mind that other topological measures data from muscle aging in humans for the pre- besides degree also exist (e.g., closeness, eigen- diction of new longevity candidates (Bell et al., vector centrality, betweenness, and bridging 2009). centrality) and their usage could result in a dif- Our meta-analysis of CR microarray stud- ferent sorting order. Ultimately, no matter what ies in mammals integrated co-expression data, the selection criteria are, experimental valida- information on genetic mutants, and analysis tion is warranted. of transcription factor binding sites to reveal promising candidate regulators, providing Network Modularity a comprehensive picture of the changes that Focusing on entire categories of genes or occur during CR. In addition to the several on network modules, and on the cross-talk processes previously associated with CR men- between these modules, could provide valu- tioned above, we also found novel associations, able and unique hints regarding the system’s such as strong indications of the effect that CR susceptibility to failure. In relation to this, has on circadian rhythms (Plank et al., 2012). Xue et al. examined the modular structure of Addressing another crucial aim in gerontol- protein–protein interaction networks during ogy, the need to have reliable biomarkers of brain aging in flies and humans. Interestingly, aging can also be done by using network-based they found two large modules of co-regulated approaches, and the integration of networks genes, both associated with the proliferation– with gene expression data to create modu- differentiation switch, displaying opposite lar biomarkers of aging has been carried out age-related expression changes. A few other (Fortney et al., 2010). modules found to be associated with the oxi- Functional classification analysis, using for dative-reductive metabolic switch were found, example, web tools like DAVID (Huang da but only during fly aging. Overall, the authors et al., 2009) which analyze Gene Ontology and

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS Data Integration 277 pathway annotations can also generate useful complexity assessment of co-occurring tran- information regarding the nature of LAGs. For scription factor binding sites can identify example, several studies have already shown cis-regulatory variants and elucidate their that inhibition of translation can be an effec- mechanistic role in disease. This has been tive modulator of lifespan extension (Curran recently carried out for type 2 diabetes, suc- and Ruvkun, 2007; Hansen et al., 2007; Pan cessfully linking genetic association signals et al., 2007). The integration of large-scale lists to disease-related molecular mechanisms of genes with gene annotation data is there- (Claussnitzer et al., 2014). For Parkinson’s fore common in analyzing omics experiments disease, integrative analyses of gene expres- and can provide insights concerning mecha- sion and GWAS data have also provided key nisms, processes, and pathways (reviewed in insights into the genetic etiology of the dis- de Magalhães et al., 2010). ease (Edwards et al., 2011). Lastly, constructing The study of aging is strongly linked to that molecular networks based on whole-genome of major age-related diseases. This becomes gene expression profiling and genotyping data, obvious when looking at the overlap between together with the use of Bayesian inference, has the genes associated with age-related diseases helped to identify key causal regulators in late- (including atherosclerosis, cancer, type 2 dia- onset Alzheimer’s disease (Zhang et al., 2013a). betes, and Alzheimer’s disease) and the genes The above examples are only a selected few involved in lifespan regulation (Budovsky et al., since many different types of network analyses 2007, 2009; Tacutu et al., 2011; Wolfson et al., and data integration can be performed. It is not 2009), as well as when analyzing the many surprising that integrative approaches are start- direct and indirect molecular interactions which ing to be used to combine disease and aging- exist between them (Simko et al., 2009; Tacutu related data. While some studies have focused et al., 2011). Networks have been extensively on a particular disease and its links to aging/ used for the study of diseases (Goh and Choi, longevity (Budovsky et al., 2009; Miller et al., 2012; Ideker and Sharan, 2008). Recently, exam- 2008), others have attempted in a broader way ples of analyses of multi-dimensional data for to look at the common signatures of aging/lon- age-related diseases have also started to amass. gevity and major age-related diseases (Wang For example, based on a network of genes and et al., 2009a; Wolfson et al., 2009). diseases created by Goh et al. (2007), structural In order to better understand the gene facets of proteins, such as the intrinsic disorder expression and protein-level changes that occur content, and epigenetic aspects as alternative with age, other genomic and epigenetic layers splicing, have been studied (Midic et al., 2009). should be considered. For example, age-related Models of diseases–genes–drugs have also changes in miRNA expression profiles can have been constructed, and new insights have been a significant impact on protein levels. In terms found about the usage of drugs (Yildirim et al., of data integration, some initial attempts have 2007). Although outside the scope of this chap- been done to combine experimentally validated ter, gene–drug interaction data are thus another miRNA data with a protein–protein interac- type of data that can be used. In fact, a network- tion network, with an emphasis on longevity based view of drug discovery and biomarkers is and age-related disease networks (Tacutu et al., starting to emerge to also account for the com- 2010a). These data have also been used to ana- plexity of human biology (de Magalhães et al., lyze the strong molecular links between aging, 2012; Erler and Linding, 2010). longevity, and age-related diseases, and to sug- In the context of GWAS, combining gest the potential role for miRNAs in target- GWAS with phylogenetic conservation and a ing certain genes with features of antagonistic

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS 278 9. Integrative Genomics of Aging pleiotropy, implying thus a preferability to initi- drugs are considered candidates due to their ate longevity-promoting interventions in adult relation with genes that are already known to life (Tacutu et al., 2010b). In another study, be associated with aging or longevity. Though interpreting the methylation patterns in cancer this premise is common to many strategies, and aging has been done using an integrative they usually differ in the type of associations system. By developing a novel epigenome–­ that are considered. For example, based on the interactome approach with differential meth- finding that hubs and centrally located nodes ylation data, tissue-independent age-associated have a higher likelihood to be associated with methylation hotspots targeting stem-cell differ- aging/longevity, Witten and Bonchev used entiation pathways have been recently discov- a C. elegans network to predict new LAGs ered (West et al., 2013a). (Witten and Bonchev, 2007). Likewise, other One important aspect about data integration topological measures have been employed for is that integrating multiple data sources will similar goals. Using a proximity measure in a significantly expand our view of the aging pro- yeast network (the shortest path to an already- cess, and it is possible that some of the current known gene reported to be associated with an well-accepted hypotheses will even be chal- increased lifespan), Managbanag et al. identi- lenged. For example, although at the network fied a set of single-gene deletions predicted level of protein–protein interactions it seems to affect lifespan. Testing this experimentally, that hub genes are of utmost importance for the their validation showed that the predicted set robustness of the entire network, when look- was enriched for mutations conferring either ing at an epigenetic level it has been suggested increased or decreased replicative lifespan that the age-associated drift in DNA methyla- (Managbanag et al., 2008). In another example, tion occurs preferentially in genes that occupy using machine learning and classification tech- peripheral network positions of exceptionally niques, Freitas et al. devised a predictive model low connectivity (West et al., 2013b). Only by to discriminate between aging-related and non- having a complete, multi-layered picture of the aging-related DNA repair genes. In this analy- aging process can we hope to fully understand sis, they found that gene connectivity together its intricacies. with specific Gene Ontology terms, having interaction with the XRCC5 protein, and a high Predictive Methods and Models expression in T lymphocytes are good predic- tors of aging-association for human DNA repair Given the intrinsic costs of performing ani- genes (Freitas et al., 2011). mal aging studies, particularly in mammals, In C. elegans, various properties of longevity developing predictive computational tools genes have been analyzed and then used to ver- is of utmost importance. Indeed, to identify ify the prediction of new longevity regulators (Li suitable drug targets with anti-aging proper- et al., 2010). In one study, the authors found that ties, methods for prioritizing them are neces- longer genomic sequences, co-expression with sary (de Magalhães et al., 2012). Fortunately, other genes during the transition from dauer to many computational tools are already avail- non-dauer state, enrichment in certain functions able for prioritizing candidates (Moreau and and RNAi phenotypes, higher sequence conser- Tranchevent, 2012), and could be of great use to vation, and a higher connectivity in a functional biogerontologists. interaction network, are all predictors of an asso- One of the main assumptions for many ciation with longevity. While the validation of predictive methods is based on the “guilt-by- the prediction was only computational, based association” principle, in which new genes or on the precision calculated with a tenfold cross

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS Concluding Remarks 279 method for a set of known positive and negative features and biochemical/physicochemical LAGs, the authors found in the scientific litera- features, a two-layer deletion network model ture that a few of the predicted genes had been has been developed and used for predict- in the meantime experimentally validated (Li ing the epistatic effects of double deletions on et al., 2010). yeast longevity. Results showed that the func- We have recently used a combined approach, tional features (such as mitochondrial function first reasoning that the interaction partners of and chromatin silencing), the network features LAGs are more likely to modulate longevity, (such as the edge density and edge weight den- and second narrowing down the candidate list sity of the deletion network), and the local cen- based on features of antagonistic pleiotropy. trality of deletion gene are important predictors Although by the time of this study several for the deletion effects on longevity (Huang genome-wide longevity assays had been per- et al., 2012). formed in C. elegans, our prediction method, Candidate gene prioritization methods, such followed by experimental validation, resulted as the ones described in this section, have been in the discovery of new longevity regulators instrumental in guiding various experiments at a frequency much higher than previously that provided important insights into aging achieved (Tacutu et al., 2012). mechanisms (Lorenz et al., 2009; Wuttke et al., Combining a network-based approach with 2012; Xue et al., 2007). The accuracy and speci- transcriptional data from human aging has also ficity of these in silico predictive methods is been used as a method of prediction. Using a still limited, however. Similarly, while compu- human longevity network constructed based tational methods have been developed for pre- on homologs from invertebrate species, and dicting candidate drugs from gene expression comparing the result with age-related transcrip- data (Iorio et al., 2010; Lamb et al., 2006; Sirota tional data from human muscle aging, Bell et al. et al., 2011), these have not been fully imple- determined a set of human interaction part- mented in the context of aging, in spite of their ners potentially involved in aging. Testing the widespread interest. homologs of these genes in C. elegans, revealed that 33% of the candidates extended lifespan when knocked-down (Bell et al., 2009). CONCLUDING REMARKS Focusing on CR, network and systems biol- ogy approaches have also been used to predict Biological and medical research has often genes necessary for the life-extending effects of failed to capture the whole picture of the dis- CR. By looking at genes that are more connected ease or process under study. Researchers have to already-known CR-related genes, Wuttke traditionally focused on a limited number of et al. successfully predicted a set of novel genes players that either had the greatest impact or, mediating the life-extending effects of CR. Nine by chance, happened to be associated with the novel genes related to CR were validated exper- given disease or process. For some diseases imentally in yeast. This revealed three novel CR (e.g., antibiotics were developed as therapies mimetic genes (Wuttke et al., 2012). with only modest understanding of the mecha- While in the last few decades many studies nisms involved) and processes (e.g., overcoming in model organisms have successfully identi- replicative senescence with ectopic telomerase fied genetic factors that affect lifespan, the effect expression) a limited mechanistic understand- of combined interventions (epistasis), whether ing may suffice to develop interventions, but synergistic or antagonistic, has been evaluated for many others our understanding and models to a much more limited degree. Using network are imperfect at best and possibly even flawed.

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS 280 9. Integrative Genomics of Aging

Researchers do not see the forest for the trees BB/H008497/1 and BB/K016741/1), the Wellcome Trust and for complex processes like aging, and age- (WT094386MA) and the Ellison Medical Foundation for supporting his lab’s development of integrative genomic related diseases like cancer and neurodegenera- resources and methods described in this paper. RT is sup- tive diseases, this impedes the development of ported by a Marie Curie FP7-PEOPLE-2011-IEF Fellowship successful interventions. Not surprisingly, the within the 7th European Community Framework overall rate of success of clinical trials is only Programme. We apologize to those whose relevant papers about 20% (DiMasi et al., 2010), and to date are not cited owing to lack of space. there is no established approach to retard, even if slightly, human aging (with the exception, per- References haps, of exercise, see Chapter 19). While seren- dipitous discoveries, like antibiotics, are always Acosta, J.C., O’Loghlen, A., Banito, A., Guijarro, M.V., possible, it is widely acknowledged that the Augert, A., Raguz, S., et al., 2008. Chemokine signaling via the CXCR2 receptor reinforces senescence. Cell 133 study of complex processes like aging stands a (6), 1006–1018. better chance of developing clinical interven- Agrawal, S., Dimitrova, N., Nathan, P., Udayakumar, K., tions based on a broad biological understanding Lakshmi, S.S., Sriram, S., et al., 2008. T2D-Db: an inte- (de Magalhães, 2014). Besides, the discoveries in grated platform to study the molecular basis of type 2 the genetics of aging and technological advances diabetes. BMC Genomics 9, 320. Albert, R., Jeong, H., Barabasi, A.L., 2000. Error and attack in large-scale methodologies, like high-through- tolerance of complex networks. Nature 406 (6794), put profiling and screening, mean that it is vital 378–382. now to cope with the growing amount of data Ameur, A., Stewart, J.B., Freyer, C., Hagstrom, E., Ingman, in the context of drug discovery (de Magalhães M., Larsson, N.G., et al., 2011. Ultra-deep sequencing et al., 2012). As new layers of genomic regula- of mouse mitochondrial DNA: mutational patterns and their origins. PLoS Genet. 7 (3), e1002028. tion are uncovered (e.g., non-coding RNAs) this Barabasi, A.L., Albert, R., 1999. Emergence of scaling in ran- raises new challenges and further emphasizes dom networks. Science 286 (5439), 509–512. the need to study biological systems in a com- Barabasi, A.L., Oltvai, Z.N., 2004. Network biology: under- prehensive fashion to capture and decipher their standing the cell’s functional organization. Nat. Rev. intrinsic complexity. Genet. 5 (2), 101–113. Barabasi, A.L., Gulbahce, N., Loscalzo, J., 2011. Network Overall, our belief is that the combination medicine: a network-based approach to human disease. of large-scale approaches to unravel both age- Nat. Rev. Genet. 12 (1), 56–68. related changes as well as identify the causes Beekman, M., Blanche, H., Perola, M., Hervonen, A., for variability across individuals and species Bezrukov, V., Sikora, E., et al., 2013. Genome-wide link- will drive the field forward. These require, age analysis for human longevity: genetics of Healthy Aging Study. Aging Cell 12 (2), 184–193. however, adequate data and statistical analy- Bell, R., Hubbard, A., Chettier, R., Chen, D., Miller, J.P., sis to avoid biases and false results. The inte- Kapahi, P., et al., 2009. A human protein interaction net- gration of different types of data provides work shows conservation of aging processes between opportunities for synergy and discovery that human and invertebrate species. PLoS Genet. 5 (3), we believe will result in a much deeper under- e1000414. Benjamini, Y., Hochberg, Y., 1995. Controlling the false dis- standing of aging and the development of inter- covery rate: a practical and powerful approach to mul- ventions to extend lifespan and preserve health. tiple testing. J. R. Stat. Soc. B 57 (1), 289–300. Bertram, L., McQueen, M.B., Mullin, K., Blacker, D., Tanzi, R.E., 2007. Systematic meta-analyses of Alzheimer dis- Acknowledgments ease genetic association studies: the AlzGene database. Nat. Genet. 39 (1), 17–23. We are thankful to Thomas Craig for helping make Budovsky, A., Abramovich, A., Cohen, R., Chalifa-Caspi, V., Figure 9.2. JPM is grateful to the UK Biotechnology and Fraifeld, V., 2007. Longevity network: construction and Biological Sciences Resource Council (BB/G024774/1, implications. Mech. Ageing Dev. 128 (1), 117–124.

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS REFERENCES 281

Budovsky, A., Tacutu, R., Yanai, H., Abramovich, A., de Magalhães, J.P., 2009. Aging research in the post- Wolfson, M., Fraifeld, V., 2009. Common gene signature genome era: new technologies for an old problem. In: of cancer and longevity. Mech. Ageing Dev. 130 (1–2), Foyer, C.H., Faragher, R., Thornalley, P.J. (Eds.), Redox 33–39. Metabolism and Longevity Relationships in Animals Budovsky, A., Craig, T., Wang, J., Tacutu, R., Csordas, A., and Plants Taylor & Francis, New York and Abingdon, Lourenco, J., et al., 2013. LongevityMap: a database pp. 99–115. of human genetic variants associated with longevity. de Magalhães, J.P., 2011. The biology of ageing: a primer. Trends Genet. 29 (10), 559–560. In: Stuart-Hamilton, I. (Ed.), An Introduction to Cai, H., Kumar, N., Ai, N., Gupta, S., Rath, P., Baudis, M., Gerontology Cambridge University Press, Cambridge, 2014. Progenetix: 12 years of oncogenomic data cura- pp. 21–47. tion. Nucleic Acids Res. 42 (1), D1055–1062. de Magalhães, J.P., 2014. The scientific quest for lasting Cevenini, E., Bellavista, E., Tieri, P., Castellani, G., Lescai, F., youth: prospects for curing aging. Rejuvenation Res. 17 Francesconi, M., et al., 2010. Systems biology and lon- (5), 458–467. gevity: an emerging approach to identify innovative de Magalhães, J.P., 2014. Why genes extending lifespan in anti-aging targets and strategies. Curr. Pharm. Des. 16 model organisms have not been consistently associated (7), 802–813. with human longevity and what it means to translation Chakravarti, B., Oseguera, M., Dalal, N., Fathy, P., Mallik, research. Cell Cycle 13 (17), 2671–2673. B., Raval, A., et al., 2008. Proteomic profiling of aging de Magalhães, J.P., Church, G.M., 2007. Analyses of human- in the mouse heart: altered expression of mitochondrial chimpanzee orthologous gene pairs to explore evolu- proteins. Arch. Biochem. Biophys. 474 (1), 22–31. tionary hypotheses of aging. Mech. Ageing Dev. 128 Chakravarti, B., Seshi, B., Ratanaprayul, W., Dalal, N., Lin, (5–6), 355–364. L., Raval, A., et al., 2009. Proteome profiling of aging de Magalhães, J.P., Keane, M., 2013. Endless paces of in mouse models: differential expression of proteins ­degeneration—applying comparative genomics to involved in metabolism, transport, and stress response study evolution’s moulding of longevity. EMBO Rep. 14 in kidney. Proteomics 9 (3), 580–597. (8), 661–662. Chen, G., Wang, Z., Wang, D., Qiu, C., Liu, M., Chen, X., de Magalhães, J.P., Toussaint, O., 2004a. GenAge: a genomic et al., 2013. LncRNADisease: a database for long-non- and proteomic network map of human ageing. FEBS coding RNA-associated diseases. Nucleic Acids Res. 41 Lett. 571 (1–3), 243–247. (Database issue), D983–986. de Magalhães, J.P., Toussaint, O., 2004b. How bioinformat- Christensen, K., Johnson, T.E., Vaupel, J.W., 2006. The quest ics can help reverse engineer human aging. Ageing Res. for genetic determinants of human longevity: chal- Rev. 3 (2), 125–141. lenges and insights. Nat. Rev. Genet. 7 (6), 436–448. de Magalhães, J.P., Curado, J., Church, G.M., 2009. Meta- Claussnitzer, M., Dankel, Simon N., Klocke, B., Grallert, H., analysis of age-related gene expression profiles identi- Glunk, V., Berulava, T., et al., 2014. Leveraging cross- fies common signatures of aging. Bioinformatics 25 (7), species transcription factor binding site patterns: from 875–881. diabetes risk loci to disease mechanisms. Cell 156 (1–2), de Magalhães, J.P., Finch, C.E., Janssens, G., 2010. Next- 343–358. generation sequencing in aging research: emerging Craig, T., Smelick, C., Tacutu, R., Wuttke, D., Wood, S.H., applications, problems, pitfalls and possible solutions. Stanley, H., et al., 2015. The digital ageing atlas: inte- Ageing Res. Rev. 9 (3), 315–323. grating the diversity of age-related changes into a uni- de Magalhães, J.P., Wuttke, D., Wood, S.H., Plank, M., fied resource. Nucleic Acids Res. 43 (Database issue), Vora, C., 2012. Genome-environment interactions that D873–D878. modulate aging: powerful targets for drug discovery. Csermely, P., Soti, C., 2006. Cellular networks and the aging Pharmacol. Rev. 64 (1), 88–101. process. Arch. Physiol. Biochem. 112 (2), 60–64. Derveaux, S., Vandesompele, J., Hellemans, J., 2010. How to Curran, S.P., Ruvkun, G., 2007. Lifespan regulation by evo- do successful gene expression analysis using real-time lutionarily conserved genes essential for viability. PLoS PCR. Methods 50 (4), 227–230. Genet. 3 (4), e56. DiMasi, J.A., Feldman, L., Seckler, A., Wilson, A., 2010. de Magalhães, J.P., 2004. From cells to ageing: a review of Trends in risks associated with new drug development: models and mechanisms of cellular senescence and their success rates for investigational drugs. Clin. Pharmacol. impact on human ageing. Exp. Cell. Res. 300 (1), 1–10. Ther. 87 (3), 272–277. de Magalhães, J.P., 2006. Species selection in comparative Dunham, I., Kundaje, A., Aldred, S.F., Collins, P.J., Davis, studies of aging and antiaging research. In: Conn, P.M. C.A., Doyle, F., et al., 2012. An integrated encyclopedia (Ed.), Handbook of Models for Human Aging Elsevier of DNA elements in the human genome. Nature 489 Academic Press, Burlington, MA, pp. 9–20. (7414), 57–74.

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS 282 9. Integrative Genomics of Aging

Echeverri, C.J., Perrimon, N., 2006. High-throughput RNAi profiles reveal quantitative views of human aging rates. screening in cultured cells: a user’s guide. Nat. Rev. Mol. Cell 49 (2), 359–367. Genet. 7 (5), 373–384. Hansen, M., Hsu, A.L., Dillin, A., Kenyon, C., 2005. New Edwards, Y.J., Beecham, G.W., Scott, W.K., Khuri, S., genes tied to endocrine, metabolic, and dietary regula- Bademci, G., Tekin, D., et al., 2011. Identifying consen- tion of lifespan from a Caenorhabditis elegans genomic sus disease pathways in Parkinson’s disease using an RNAi screen. PLoS Genet. 1 (1), 119–128. integrative systems biology approach. PLoS One 6 (2), Hansen, M., Taubert, S., Crawford, D., Libina, N., Lee, S.J., e16917. Kenyon, C., 2007. Lifespan extension by conditions that Erler, J.T., Linding, R., 2010. Network-based drugs and bio- inhibit translation in Caenorhabditis elegans. Aging Cell 6 markers. J. Pathol. 220 (2), 290–296. (1), 95–110. Fabrizio, P., Hoon, S., Shamalnasab, M., Galbani, A., Wei, He, X., Chang, S., Zhang, J., Zhao, Q., Xiang, H., M., Giaever, G., et al., 2010. Genome-wide screen in Kusonmano, K., et al., 2008. MethyCancer: the database Saccharomyces cerevisiae identifies vacuolar protein sort- of human DNA methylation and cancer. Nucleic Acids ing, autophagy, biosynthetic, and tRNA methylation Res. 36 (Database issue), D836–841. genes involved in life span regulation. PLoS Genet. 6 Hindorff, L.A., Sethupathy, P., Junkins, H.A., Ramos, E.M., (7), e1001024. Mehta, J.P., Collins, F.S., et al., 2009. Potential etiologic Ferrarini, L., Bertelli, L., Feala, J., McCulloch, A.D., and functional implications of genome-wide association Paternostro, G., 2005. A more efficient search strategy loci for human diseases and traits. Proc. Natl. Acad. Sci. for aging genes based on connectivity. Bioinformatics USA 106 (23), 9362–9367. 21 (3), 338–348. Hofer, E., Laschober, G.T., Hackl, M., Thallinger, G.G., Finch, C.E., 1990. Longevity, Senescence, and the Genome. Lepperdinger, G., Grillari, J., et al., 2011. GiSAO.db: a The University of Chicago Press, Chicago and London. database for ageing research. BMC Genomics 12, 262. Fortney, K., Kotlyar, M., Jurisica, I., 2010. Inferring the func- Horvath, S., 2013. DNA methylation age of human tissues tions of longevity genes with modular subnetwork bio- and cell types. Genome Biol. 14 (10), R115. markers of Caenorhabditis elegans aging. Genome Biol. 11 Houtkooper, R.H., Argmann, C., Houten, S.M., Canto, C., (2), R13. Jeninga, E.H., Andreux, P.A., et al., 2011. The metabolic Freitas, A.A., Vasieva, O., de Magalhães, J.P., 2011. A data footprint of aging in mice. Sci. Rep. 1, 134. mining approach for classifying DNA repair genes into Huang da, W., Sherman, B.T., Lempicki, R.A., 2009. ageing-related or non-ageing-related. BMC Genomics Systematic and integrative analysis of large gene lists 12, 27. using DAVID bioinformatics resources. Nat. Protoc. 4 Funakoshi, M., Tsuda, M., Muramatsu, K., Hatsuda, H., (1), 44–57. Morishita, S., Aigaki, T., 2011. A gain-of-function screen Huang, T., Zhang, J., Xu, Z.P., Hu, L.L., Chen, L., Shao, J.L., identifies wdb and lkb1 as lifespan-extending genes in et al., 2012. Deciphering the effects of gene deletion on Drosophila. Biochem. Biophys. Res. Commun. 405 (4), yeast longevity using network and machine learning 667–672. approaches. Biochimie 94 (4), 1017–1025. Futreal, P.A., Coin, L., Marshall, M., Down, T., Hubbard, Huhne, R., Thalheim, T., Suhnel, J., 2014. AgeFactDB—the T., Wooster, R., et al., 2004. A census of human cancer JenAge Ageing Factor Database—towards data inte- genes. Nat. Rev. Cancer 4 (3), 177–183. gration in ageing research. Nucleic Acids Res. 42 (1), Glass, D., Vinuela, A., Davies, M.N., Ramasamy, A., Parts, D892–896. L., Knowles, D., et al., 2013. Gene expression changes Ideker, T., Sharan, R., 2008. Protein networks in disease. with age in skin, adipose tissue, blood and brain. Genome Res. 18 (4), 644–652. Genome Biol. 14 (7), R75. Ingolia, N.T., Ghaemmaghami, S., Newman, J.R., Weissman, Goh, K.I., Choi, I.G., 2012. Exploring the human diseasome: J.S., 2009. Genome-wide analysis in vivo of translation the human disease network. Brief Funct. Genomics 11 with nucleotide resolution using ribosome profiling. (6), 533–542. Science 324 (5924), 218–223. Goh, K.I., Cusick, M.E., Valle, D., Childs, B., Vidal, M., Iorio, F., Bosotti, R., Scacheri, E., Belcastro, V., Mithbaokar, Barabasi, A.L., 2007. The human disease network. Proc. P., Ferriero, R., et al., 2010. Discovery of drug mode Natl. Acad. Sci. U. S. A. 104 (21), 8685–8690. of action and drug repositioning from transcrip- Hamilton, B., Dong, Y., Shindo, M., Liu, W., Odell, I., Ruvkun, tional responses. Proc. Natl. Acad. Sci. USA 107 (33), G., et al., 2005. A systematic RNAi screen for longevity 14621–14626. genes in C. elegans. Genes Dev. 19 (13), 1544–1555. Jobson, R.W., Nabholz, B., Galtier, N., 2010. An evolution- Hannum, G., Guinney, J., Zhao, L., Zhang, L., Hughes, ary genome scan for longevity-related natural selection G., Sadda, S., et al., 2013. Genome-wide methylation in mammals. Mol. Biol. Evol. 27 (4), 840–847.

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS REFERENCES 283

Johnson, A.A., Akman, K., Calimport, S.R., Wuttke, D., Macarron, R., Banks, M.N., Bojanic, D., Burns, D.J., Cirovic, Stolzing, A., de Magalhães, J.P., 2012. The role of DNA D.A., Garyantes, T., et al., 2011. Impact of high-through- methylation in aging, rejuvenation, and age-related dis- put screening in biomedical research. Nat. Rev. Drug ease. Rejuvenation Res. 15 (5), 483–494. Discov. 10 (3), 188–195. Johnson, T.E., 2002. A personal retrospective on the genetics Managbanag, J.R., Witten, T.M., Bonchev, D., Fox, L.A., of aging. Biogerontology 3 (1–2), 7–12. Tsuchiya, M., Kennedy, B.K., et al., 2008. Shortest-path Keane, K., Craig, T., Alföldi, J., Berlin, A.M., Johnson, J., network analysis is a useful approach toward identify- Seluanov, A., et al., 2014. The Naked Mole Rat Genome ing genetic determinants of longevity. PLoS One 3 (11), Resource: facilitating analyses of cancer and longevity- e3802. related adaptations. Bioinformatics 30 (24), 3558–3560. Manolio, T.A., Collins, F.S., Cox, N.J., Goldstein, D.B., Kennedy, S.R., Salk, J.J., Schmitt, M.W., Loeb, L.A., 2013. Hindorff, L.A., Hunter, D.J., et al., 2009. Finding the Ultra-sensitive sequencing reveals an age-related missing heritability of complex diseases. Nature 461 increase in somatic mitochondrial mutations that are (7265), 747–753. inconsistent with oxidative damage. PLoS Genet. 9 (9), Matecic, M., Smith, D.L., Pan, X., Maqani, N., Bekiranov, e1003794. S., Boeke, J.D., et al., 2010. A microarray-based genetic Kim, E.B., Fang, X., Fushan, A.A., Huang, Z., Lobanov, screen for yeast chronological aging factors. PLoS A.V., Han, L., et al., 2011. Genome sequencing reveals Genet. 6 (4), e1000921. insights into physiology and longevity of the naked McCormick, M.A., Kennedy, B.K., 2012. Genome-scale mole rat. Nature 479 (7372), 223–227. studies of aging: challenges and opportunities. Curr. Kirkwood, T.B., Kowald, A., 1997. Network theory of aging. Genomics 13 (7), 500–507. Exp. Gerontol. 32 (4–5), 395–399. Menni, C., Kastenmuller, G., Petersen, A.K., Bell, J.T., Lamb, J., Crawford, E.D., Peck, D., Modell, J.W., Blat, I.C., Psatha, M., Tsai, P.C., et al., 2013. Metabolomic mark- Wrobel, M.J., et al., 2006. The Connectivity Map: using ers reveal novel pathways of ageing and early develop- gene-expression signatures to connect small molecules, ment in human populations. Int. J. Epidemiol. 42 (4), genes, and disease. Science 313 (5795), 1929–1935. 1111–1119. Lee, C.K., Klopp, R.G., Weindruch, R., Prolla, T.A., 1999. Midic, U., Oldfield, C.J., Dunker, A.K., Obradovic, Z., Gene expression profile of aging and its retardation by Uversky, V.N., 2009. Protein disorder in the human dis- caloric restriction. Science 285 (5432), 1390–1393. easome: unfoldomics of human genetic diseases. BMC Li, Y., de Magalhães, J.P., 2013. Accelerated protein evo- Genomics 10 (Suppl 1), S12. lution analysis reveals genes and pathways associ- Miller, J.A., Oldham, M.C., Geschwind, D.H., 2008. A ated with the evolution of mammalian longevity. Age systems level analysis of transcriptional changes in (Dordr) 35 (2), 301–314. Alzheimer’s disease and normal aging. J. Neurosci. 28 Li, Y., Qiu, C., Tu, J., Geng, B., Yang, J., Jiang, T., et al., 2014. (6), 1410–1420. HMDD v2.0: a database for experimentally supported Moffat, J., Sabatini, D.M., 2006. Building mammalian signal- human microRNA and disease associations. Nucleic ling pathways with RNAi screens. Nat. Rev. Mol. Cell. Acids Res. 42 (1), D1070–1074. Biol. 7 (3), 177–187. Li, Y.H., Dong, M.Q., Guo, Z., 2010. Systematic analysis and Mohr, S., Bakal, C., Perrimon, N., 2010. Genomic screen- prediction of longevity genes in Caenorhabditis elegans. ing with RNAi: results and challenges. Annu. Rev. Mech. Ageing Dev. 131 (11–12), 700–709. Biochem. 79, 37–64. Liu, H., Liu, W., Liao, Y., Cheng, L., Liu, Q., Ren, X., et al., Moreau, Y., Tranchevent, L.C., 2012. Computational tools 2011. CADgene: a comprehensive database for coronary for prioritizing candidate genes: boosting disease gene artery disease genes. Nucleic Acids Res. 39 (Database discovery. Nat. Rev. Genet. 13 (8), 523–536. issue), D991–996. Mortazavi, A., Williams, B.A., McCue, K., Schaeffer, L., Liu, Y., Kern, J.T., Walker, J.R., Johnson, J.A., Schultz, P.G., Wold, B., 2008. Mapping and quantifying mamma- Luesch, H., 2007. A genomic screen for activators of the lian transcriptomes by RNA-Seq. Nat. Methods 5 (7), antioxidant response element. Proc. Natl. Acad. Sci. 621–628. USA 104 (12), 5205–5210. Nagaoka-Yasuda, R., Matsuo, N., Perkins, B., Limbaeck- Lopez-Otin, C., Blasco, M.A., Partridge, L., Serrano, M., Stokin, K., Mayford, M., 2007. An RNAi-based genetic Kroemer, G., 2013. The hallmarks of aging. Cell 153 (6), screen for oxidative stress resistance reveals retinol sat- 1194–1217. urase as a mediator of stress resistance. Free Radic. Biol. Lorenz, D.R., Cantor, C.R., Collins, J.J., 2009. A network Med. 43 (5), 781–788. biology approach to aging in yeast. Proc. Natl. Acad. Oikonomou, P., Cluzel, P., 2006. Effects of topology on net- Sci. U. S. A. 106 (4), 1145–1150. work evolution. Nat. Phys. 2, 532–536.

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS 284 9. Integrative Genomics of Aging

Olsen, B., Kaeberlein, M., 2014. Lifespan Observations Sirota, M., Dudley, J.T., Kim, J., Chiang, A.P., Morgan, A.A., Database. . Sweet-Cordero, A., et al., 2011. Discovery and preclinical OMIM, 2014. Online Mendelian validation of drug indications using compendia of pub- Inheritance in Man, OMIM®. McKusick-Nathans lic gene expression data. Sci. Transl. Med. 3 (96), 96ra77. Institute of Genetic Medicine, Johns Hopkins Smith, E.D., Tsuchiya, M., Fox, L.A., Dang, N., Hu, D., Kerr, University, Baltimore, MD. E.O., et al., 2008. Quantitative evidence for conserved Ongenaert, M., Van Neste, L., De Meyer, T., Menschaert, longevity pathways between divergent eukaryotic spe- G., Bekaert, S., Van Criekinge, W., 2008. PubMeth: a cies. Genome Res. 18 (4), 564–570. cancer methylation database combining text-mining Soltow, Q.A., Jones, D.P., Promislow, D.E., 2010. A network and expert annotation. Nucleic Acids Res. 36 (Database perspective on metabolism and aging. Integr. Comp. issue), D842–846. Biol. 50 (5), 844–854. Paik, D., Jang, Y.G., Lee, Y.E., Lee, Y.N., Yamamoto, R., Gee, Stark, C., Breitkreutz, B.J., Reguly, T., Boucher, L., H.Y., et al., 2012. Misexpression screen delineates novel Breitkreutz, A., Tyers, M., 2006. BioGRID: a general genes controlling Drosophila lifespan. Mech. Ageing repository for interaction datasets. Nucleic Acids Res. Dev. 133 (5), 234–245. 34 (Database issue), D535–539. Pan, F., Chiu, C.H., Pulapura, S., Mehan, M.R., Nunez- Storey, J.D., Tibshirani, R., 2003. Statistical significance for Iglesias, J., Zhang, K., et al., 2007. Gene Aging Nexus: a genomewide studies. Proc. Natl. Acad. Sci. USA 100 web database and data mining platform for microarray (16), 9440–9445. data on aging. Nucleic Acids Res. 35 (Database issue), Stuart, J.A., Liang, P., Luo, X., Page, M.M., Gallagher, E.J., D756–759. Christoff, C.A., et al., 2013. A comparative cellular and Petrascheck, M., Ye, X., Buck, L.B., 2007. An antidepressant molecular biology of longevity database. Age (Dordr) that extends lifespan in adult Caenorhabditis elegans. 35 (5), 1937–1947. Nature 450 (7169), 553–556. Tacutu, R., Budovsky, A., Fraifeld, V.E., 2010a. The NetAge Plank, M., Wuttke, D., van Dam, S., Clarke, S.A., de database: a compendium of networks for longev- Magalhães, J.P., 2012. A meta-analysis of caloric restric- ity, age-related diseases and associated processes. tion gene expression profiles to infer common signa- Biogerontology 11 (4), 513–522. tures and regulatory mechanisms. Mol. Biosyst. 8 (4), Tacutu, R., Budovsky, A., Wolfson, M., Fraifeld, V.E., 2010b. 1339–1349. MicroRNA-regulated protein-protein interaction net- Plank, M., Hu, G., Silva, A.S., Wood, S.H., Hesketh, E.E., works: how could they help in searching for pro-lon- Janssens, G., et al., 2013. An analysis and validation gevity targets? Rejuvenation Res. 13 (2–3), 373–377. pipeline for large-scale RNAi-based screens. Sci. Rep. 3, Tacutu, R., Budovsky, A., Yanai, H., Fraifeld, V.E., 2011. 1076. Molecular links between cellular senescence, longevity Promislow, D.E., 2004. Protein networks, pleiotropy and and age-related diseases-a systems biology perspective. the evolution of senescence. Proc. Biol. Sci. 271 (1545), Aging (Albany NY) 3 (12), 1178–1191. 1225–1234. Tacutu, R., Shore, D.E., Budovsky, A., de Magalhães, J.P., Saito, R., Smoot, M.E., Ono, K., Ruscheinski, J., Wang, Ruvkun, G., Fraifeld, V.E., et al., 2012. Prediction of C. P.L., Lotia, S., et al., 2012. A travel guide to Cytoscape elegans longevity genes by human and worm longevity plugins. Nat. Methods 9 (11), 1069–1076. networks. PLoS One 7 (10), e48282. Samuelson, A.V., Carr, C.E., Ruvkun, G., 2007. Gene Tacutu, R., Craig, T., Budovsky, A., Wuttke, D., Lehmann, activities that mediate increased life span of C. elegans G., Taranukha, D., et al., 2013. Human Ageing Genomic insulin-like signaling mutants. Genes Dev. 21 (22), Resources: integrated databases and tools for the biol- 2976–2994. ogy and genetics of ageing. Nucleic Acids Res. 41 Schachter, F., Faure-Delanef, L., Guenot, F., Rouger, H., (Database issue), D1027–1033. Froguel, P., Lesueur-Ginot, L., et al., 1994. Genetic asso- Tomas-Loba, A., Bernardes de Jesus, B., Mato, J.M., Blasco, ciations with human longevity at the APOE and ACE M.A., 2013. A metabolic signature predicts biological loci. Nat. Genet. 6 (1), 29–32. age in mice. Aging Cell 12 (1), 93–101. Semeiks, J., Grishin, N.V., 2012. A method to find longevity- Tsuchiya, T., Dhahbi, J.M., Cui, X., Mote, P.L., Bartke, A., selected positions in the mammalian proteome. PLoS Spindler, S.R., 2004. Additive regulation of hepatic gene One 7 (6), e38595. expression by dwarfism and caloric restriction. Physiol. Simko, G.I., Gyurko, D., Veres, D.V., Nanasi, T., Csermely, Genomics 17 (3), 307–315. P., 2009. Network strategies to understand the aging Valdes, A.M., Glass, D., Spector, T.D., 2013. Omics technolo- process and help age-related drug design. Genome gies and the study of human ageing. Nat. Rev. Genet. 14 Med. 1 (9), 90. (9), 601–607.

I. BASIC ME­ ­C­HANISMS OF AGING: MODELS AND SYSTEMS REFERENCES 285

Wagner, A., 2000. Robustness against mutations in genetic Yang, S., Liu, T., Li, S., Zhang, X., Ding, Q., Que, H., et al., networks of yeast. Nat. Genet. 24 (4), 355–361. 2008. Comparative proteomic analysis of brains of natu- Wang, J., Zhang, S., Wang, Y., Chen, L., Zhang, X.S., 2009a. rally aging mice. Neuroscience 154 (3), 1107–1120. Disease-aging network reveals significant roles of aging Ye, X., Linton, J.M., Schork, N.J., Buck, L.B., Petrascheck, genes in connecting genetic diseases. PLoS Comput. M., 2014. A pharmacological network for lifespan Biol. 5 (9), e1000521. extension in Caenorhabditis elegans. Aging Cell 13 (2), Wang, Z., Gerstein, M., Snyder, M., 2009b. RNA-Seq: a revo- 206–215. lutionary tool for transcriptomics. Nat. Rev. Genet. 10 Yildirim, M.A., Goh, K.I., Cusick, M.E., Barabasi, A.L., (1), 57–63. Vidal, M., 2007. Drug-target network. Nat. Biotechnol. West, J., Beck, S., Wang, X., Teschendorff, A.E., 2013a. An 25 (10), 1119–1126. integrative network algorithm identifies age-associated Zahn, J.M., Poosala, S., Owen, A.B., Ingram, D.K., Lustig, differential methylation interactome hotspots targeting A., Carter, A., et al., 2007. AGEMAP: a gene expression stem-cell differentiation pathways. Sci. Rep. 3, 1630. database for aging in mice. PLoS Genet. 3 (11), e201. West, J., Widschwendter, M., Teschendorff, A.E., 2013b. Zhang, B., Gaiteri, C., Bodea, L.G., Wang, Z., McElwee, J., Distinctive topology of age-associated epigenetic drift Podtelezhnikov, A.A., et al., 2013a. Integrated systems in the human interactome. Proc. Natl. Acad. Sci. USA approach identifies genetic nodes and networks in late- 110 (35), 14138–14143. onset Alzheimer’s disease. Cell 153 (3), 707–720. Witten, T.M., Bonchev, D., 2007. Predicting aging/longev- Zhang, G., Cowled, C., Shi, Z., Huang, Z., Bishop-Lilly, ity-related genes in the nematode Caenorhabditis elegans. K.A., Fang, X., et al., 2013b. Comparative analysis of bat Chem. Biodivers. 4 (11), 2639–2655. genomes provides insight into the evolution of flight Wolfson, M., Budovsky, A., Tacutu, R., Fraifeld, V., 2009. and immunity. Science 339 (6118), 456–460. The signaling hubs at the crossroad of longevity and Zhang, Y., De, S., Garner, J.R., Smith, K., Wang, S.A., Becker, age-related disease networks. Int. J. Biochem. Cell Biol. K.G., 2010. Systematic analysis, comparison, and inte- 41 (3), 516–520. gration of disease based human genetic association data Wood, S.H., Craig, T., Li, Y., Merry, B., de Magalhães, J.P., and mouse genetic phenotypic information. BMC Med. 2013. Whole transcriptome sequencing of the aging rat Genomics 3, 1. brain reveals dynamic RNA changes in the dark matter Zhao, M., Sun, J., Zhao, Z., 2013. TSGene: a web resource of the genome. Age (Dordr) 35, 763–776. for tumor suppressor genes. Nucleic Acids Res. 41 Wuttke, D., Connor, R., Vora, C., Craig, T., Li, Y., Wood, S., (Database issue), D970–976. et al., 2012. Dissecting the gene network of dietary Zhou, B., Yang, L., Li, S., Huang, J., Chen, H., Hou, L., et al., restriction to identify evolutionarily conserved pathways 2012. Midlife gene expressions identify modulators of and new functional genes. PLoS Genet. 8 (8), e1002834. aging through dietary interventions. Proc. Natl. Acad. Xue, H., Xian, B., Dong, D., Xia, K., Zhu, S., Zhang, Z., et al., Sci. USA 109 (19), E1201–1209. 2007. A modular network model of aging. Mol. Syst. Biol. 3, 147.

I. BASIC M­E­C­HANISMS OF AGING: MODELS AND SYSTEMS