I would like to elaborate a little bit further on the concepts, which were at the basis of the Human Toxome Project. We want to explore how a cell system is impacted by a toxicant. A little bit this is like trying a dinosaur footstep, the perturbation to reconstruct how does the dinosaur look like, which did produce this. So, we have the impact on the cell, and we want to understand what was the footstep. But we have certain helps. The first one is that, we can have our dinosaur step again, and again, and again, because that's the elegant thing in toxicology. Different to other areas of medicine, we have the disease agent in our hand. We can expose the cells to the toxic substance again and again. We don't need disease materials. We can observe what is happening viral it is imprinting. So, we can use a resolution of seconds, of minutes, of hours observe how our system is changing after a toxicant came in contact with our materials. And, we can compare the information from substances, from toxicants, from different types of dinosaurs you could say, and we learn from them, because they are different in some aspects, and they share others, and from this we can deduce general guidelines. And, we have more. We have a lot of additional information. We have for example, a lot of databases understanding about the biochemistry, about the genetics, we have the entire human genome. So, we have things like pathway databases, gene ontologies. So, we are able to add to this type of information to the dinosaur footprint. We started off with a conceptual article. Food for thought article as we attempted on mapping the human toxome. With my co-author Mary McBride from Agilent Technologies, one of the leading companies in providing the technologies we are going to use. And, in this article, we laid out our idea how such a mapping of pathways of toxicity can actually help the implementation of Tox21, a vision and a strategy. And, we describe that our assessment of mechanisms of process of toxicity on a larger scale in a systematic manner, and target that we can actually lead to human toxome database as a key implementation activity. The basis of this is that, we believe that, there is probably only a limited finite number of pathways of toxicity. So, you can imagine that toxicants are more or less using the achilles heels of a cell, the vulnerable spots. And, we believe there is not a tremendous number of these. At least, this is all over in hypothesis, because then we could actually start to create a list of possible toxicity number one, and at number two, and at number three, and at some point we will slow down, and not find more and more of these pathways. And, if we have such a catalogue of pathways, we could start annotating them. We could say, "Yes, a certain hazard, a certain phenotype is obviously linked to certain pathways and these pathways can be redundant. These pathways can be mandatory, they can be synergistic. We will start putting into an order, how these different pathways of toxicity are linked to finally creating a hazard manifestation." We also can look the other way around. What do different classes of toxicants have in common? What are endocrine disrupters doing? What are carcinogens typically activating? What other pathways they are triggering in order to annotate two classes of toxicants? We can also look into which cell type allows a certain pathway to manifest. What are the pathways of hepatotoxicity? What other ones of cardiotoxicity or neurotoxicity? And, we can also look into species annotations. We very often observe in toxicology that certain phenomena are seen in our rats or rabbits, but they're not necessarily taking place in humans. So, it is good to annotate these pathways and say, "Yes, that's a rat specific pathway, but we know it is not relevant in humans, or the other way around." So, this is the big vision of creating the human toxome, a comprehensive list of such pathways. And, this will allow us also at some point to say when we have such a comprehensive list, and we can show that a given substance is not interacting with any of the necessary pathways, that a substance is actually negative. And this is a completely new quality, because testing today in a rat never give us a result, which says this substance is harmless. It could be that we did not dose correctly, that the rat is not sensitive while the human is. It could be another type of interaction which is necessary to deliver, but if there is no talk to the biology, if there's no impact, no perturbation of the necessarily triggered pathways, we could actually say the substance is something we can put on the back burner. This is nothing we have evil to fear a certain hazard to manifest. On this slide, you can see a mobile and I have replaced the figures of this baby mobile with some amino acids, to express that these are homeostatic situations where the different levels of amino acids are controlled under homeostatic conditions, and a perturbation by a toxicant. We could imagine like such an exogenous perturbation, which is changing if it's not collapsing the relative levels of metabolites here, and you could do the same for gene expression. So, we're producing new situation, a new homeostasis under stress. This is what we would like to measure and interpret to understand what has happened to the equilibrium of substances of molecules in the cell. So, we're using for identifying pathways of toxicity. First of all, the new homeostasis, the signature of toxicity which we observe in response to stressor. And, we try to understand the critical cell infrastructures for example, mitochondria. Mitochondria are very often involved in manifestations of toxicity, because they're such a vulnerable part of the machinery of the cell. And, we want to make use of the network knowledge of the connections between metabolites, the gene expression, the networks of genes, and molecules, and their regulation, which we increasingly understand. And, we are suggesting to use reference models, which means cell models which have shown to show relevant results and the reference toxicants. We're not going to study substances with unclear or borderline effects. We are looking into toxicants which are very well understood, very well known in order to start identifying how these substances in the relevant test systems are going to perturb the biology. Now, we have to choose among the different Omex Technologies. And, we did actually choose for the project, Metabolomics and Transcriptomics. If you see the sequence from DNA, to transcription, to RNA, and microRNAs then the translation to proteins and the proteins impacting on the turnover of metabolites, we do have here the more or less the two extremes. The early effect of transcribing genes and the late effect of metabolites. You could say that one is very close to the genotype, so the repertoire of the cell while actual changes in metabolites are real phenotypic changes. Because you have to be clear. A change in expression of a gene, does not necessarily mean that there is more protein, and even if there's more protein, it does not mean that this was short in supply, and did impact by more expression on the biology of the cell. But if the metabolite levels change, you have actually some change. They also have two advantages. Transcript Hemogenomics has the advantage that this is the most standardized of all of these technologies. As this has pioneered the Omex Technologies as we have benefited from the Human Genome Project and other areas of molecular biology. This is the area of highly standardized technology for which quality assurance, for which a lot variety of vendors is available, and which can be used interchangeably. Metabolomics at the same time has the advantage that we're dealing with a relatively small number of substances. It's only something like five to seven, eight thousand metabolites which are relevant in the human metabolome. So, we're dealing here with considerably less endpoints. 27,000 genes, hundred thousands of miRNAs possibly millions of proteins with all of the modifications. And, this makes it much easier to understand, and analyze due to this limited number. So, we've been focusing a bit on Metabolomics and in order to make this technology more accessible to toxicologists, we started with a series of workshops and information days about metabolomics. This is one of our review articles which I recommend for those of you interested in what metabolomics can do. So, the comprehensive measurement of metabolites in toxicology. Another article dealt with the metabolomics in toxicology, preclinical research. And, here you see our transatlantic think tank for toxicology, an organization we created, which is spanning the Atlantic and has carried out more than 30 workshops and more than 22 of them have been published in reports already. Which is trying to demonstrate new concepts, and find expert consensus about how to use the respective technologies. All of these are published in our journal ALTEX and are freely available for those of you interested in going into more detail. Metabolomics is not at the same level as Transcriptomics, with regard to quality assurance and standardization. And, a big problem are the databases. To show you, this is taken from this website, consensus and conflict databases. And this website compared some of the most common web databases on metabolites of human metabolites. And, as you can see here, the consensus so inclusion all of these databases does exist only for 3% of the reactions and 9% of the metabolites. Still only 15% and 21% are found in the majority of these databases. And, a large number is actually unique, found in only one of these databases. And, this shows that you will get dramatically different results depending on which of the databases you choose to refer your data to. Another example. This is what you find in different databases about estrogen. We will come back to estrogen in a minute because the human toxin project is going to study the substance. If you see the different names for Estradiol, you'll find very different information in these databases on with which compartment this metabolite is found, and whether its a deadend metabolite or not. And this is just a little highlight, a glimpse on difficulties in these databases and we need to do much more in order to harmonize such databases, make consensus databases available, quality control them, and also create something which is normal for those people working in the field of transcriptomics genomics. We need something like a gene ontology something which says, this gene is responsible for these functions. We don't have a metabolite ontology which would do something similar for our analysis. And we also, for this topic of quality assurance of metabolomics, held another workshop which was published. So, you can see that we are really starting to base our work on solid quality assurance of the technologies we are applying. Another important concept of the measurement technologies is the concept of biomarkers. What is a biomarker? A biomarker is something we measure, which is giving us a meaningful readout, which is representing a mechanism. Because all of these technologies are very noisy. If you're measuring thousands of genes, if you're measuring thousands of metabolites, there's a lot of false positives and false negatives, reactions. There's a lot of interference with these systems. And it is our job now to identify what are the meaningful signals in this noise. Dividing, signal and noise. It is our belief that a good biomarker has a mechanistic foundation, because as I said earlier, mechanism translates between model systems. So, we are interested in the knots of the network which are representing mechanism, which are helping us. And again, we held a conceptual workshop on the use of biomarkers of toxicity in cellular systems. And I don't need to go into much detail, but we tried in this figure, to illustrate that a chemical insert will lead to changes. Some of them are representing the mechanism of action. That's what we're really interested in. And there's others which are somehow showing the subsequent cascade of events, many of them will actually be the counter-regulation, the repair, the system defending, which is not really leading to damage. And in the longer term they changed the system. The cell which has been exposed to a stronger toxic insert is no longer the same cell. And there's a lot of epiphenomena, non-relevant activities. So, it is about identifying what are the signals which are representing mechanism of action, or the mode of action the biomarkers of toxicity. And we have some problems in these technologies. You see a gene array on the left and on the lower left an example of mass spectrum, the basis for more prevalent type of metabolomics. The one we have been using, which is mass-spectroscopy based metabolomics. So, what are these problems? The problem is we are measuring far too many variables. It's a very small number of repetitions. And we have the noise I have already discussed. So, if you are measuring 27,000 genes and their expression, it is impossible to do this definitively based on a handful of gene areas, but more we can typically not afford, these are expensive technologies. So, we have to overcome this problem. The first thing is to reduce the noise by applying quality control using good models and good measurement technologies. We are also favoring the Omics technologies which are closer to phenotype because we don't want to be misled by gene expressions which are not translating into phenotypic changes. And then, we have to reduce. Instead of 20 genes and their expression, we want to look into a pathway which is activated to understand what is the overall impact on this. And we are trying to acquire and develop novel tools for network analysis because we don't want to stay at the level of, "this gene has changed significantly, this metabolite has gone down significantly", we are interested in what is the overall activity of the network, and a key idea of the project is to use orthogonal Omics technology. What do I mean by orthogonal. I mean if one Omics technology shows us that a certain pathway is perturbed and another technology which is just measuring, not the gene expression but the metabolites, is showing a perturbation in the very same pathway, these two are obviously corresponding and there's a higher likelihood of actually seeing a relevant perturbation. So, they are strengthening each other. There's a cross-talk between these technologies. And this was the basis for forming the consortium of pathways of toxicology for using multi-Omics technologies and we call this project Mapping the Human Toxome by System Toxicology, and our age transformative research grant are funded for five years by the National Institute for Environmental Health Sciences. You see the different groups, two of them here in Hopkins, our center for first animal testing and the group of Jim Yager who gave the introductory talk to the Tox 21 Activities. He was a member of the Tox 21 panel who started a lot of this discussion. So was Martin Stephens in my team. Kim Boekelheide, from Brown University and Mel Andersen from the Hamner Institute. And we included Al Fornace from Georgetown, Agilent Technologies was Michael Rosenberg and the EPA, as an associate partner with their computational group talks cast. And this formed the consortium. And the consortium started off to analyze in vitro models with Omics data generation, the development of software tools to identify processes of toxicity to start the validation of tools, and finally with the goal to create a human toxome database. As a test case, we chose endocrine disruption. Endocrine disruption is of considerable interest. Large scale testing programs are being set up in the U.S. and Europe and other parts of the world because there's concern that chemicals could interfere with our hormonal system and for this reason have impact for example on reproductive, on fertility, on sperm quality, and other qualities linked to reproduction. We did choose this also because endocrine disruption is very much pathway oriented toxicology where we understand that the triggering of certain receptors by chemicals is producing signal cascades which are leading to endocrine disruption. So, the idea was to use Omics now to map pathways of toxicity for endocrine disruption and use this test case to sharpen our tools, to develop the software tools to do an actual identification and develop a process around this and start with the public database. And we were doing so by using Metobolomics and transcriptomics, my group Metabolomics, Jim Yager and Brown University doing the cell-culture work. Georgetown was doing the work on transcriptomics. Agilent was mainly focusing on the tools of interpreting the data and we all jointly worked towards the identification of pathways of toxicity and ultimately we were discussing and starting to build up the human toxome database as a collaboration between Hamner, EPA, and our center. The project has a website. The website is called humantoxome.com and here you can find more information. Also, we published a conceptual article summarizing the project, The Human Toxome Project which is available from this website as well as from the Journal [inaudible] as an open access journal. And we got a lot of interest. This is an example of an article written in the Journal Science, one of the leading journals in the scientific field and they were commenting on and lauding our project writing, "driven both by legislative mandate and scientific need, a new suite of in vitro and cell culture based animal-free methods are gaining a foothold in toxicology labs". This is giving a nice overview about the variety of Tox 21 implementation activities. But we had more, the European Parliament actually held a two hour event about The Human Toxome Project. I have to mention the U.S. and our H Project discussed two hours in European Parliament with 200 experts on what is the impact for Europe of such a project. Nature welcomed it with a blog and it starred researchers to map human toxic pathways, and later covered it again in an article about the Omes Puzzle.