 So, here is a photo taken last month of dead fish in the High Her River in Tianjin. There were over 40 chemicals that have been detected at the blast site and many of these are toxic and many of them are found in large quantities. And so, we are always exposed to mixtures of chemicals in our food and our ability to track how they impact our health in a concerted way is actually limited. So, these chemicals can get into food from being there naturally, from agricultural chemical applications from industrial use, as well as introduction in processing and packaging and storage. So, this problem came to our attention when we were studying a chemical that is in a toxic mushroom. And we did very careful mechanistic studies of this chemical and in understanding that molecular process, we were able to predict that a combined exposure to other chemicals from food would actually lead to an increase in toxicity to this chemical. And so, our typical one-at-a-time approach in evaluating toxicity failed us there. Now, around this time, I moved my research team to the ETH Zurich and I met a colleague there who was a pioneer in systems biology. And I can describe his approach with the movie The Matrix. So, if you remember The Matrix, in The Matrix, our exterior world is actually computer programs and we can break down all of our experiences to binary code. So, if my colleague was basically describing a human in the code of The Matrix, what I wanted to do was see how that human changed in The Matrix when it ate contaminated food. Well, basically, this is a complex way to approach even the basic science of toxicology. And so then the idea that we could apply this to risk assessment of chemicals seems like a little bit of a stretch. And it's also important to note that if we do this incorrectly, actually the implications could be in chronic long-term health effects of many of these chemicals. We're really motivated because of the major advances in handling and interpreting large biological data. So, for example, the Human Genome Project, 15 years ago, 100 million to sequence 3 billion bases of the human genome, and now we're under $1,000. And so, what we need to do to start this approach is take the basic concept of dose responsiveness in toxicity. And so, doing a traditional experiment, usually using an animal model, a key point to determine is a threshold, and that's a point at which the test system diverges from having no detectable response to the chemical. And so, now in taking our new systems toxicology approach, we keep this idea central, but we need new biological models, ones that we can test in a very rapid way, like human-cultured cells, for example. In using new biological models, we need to change our measurement strategies, and we need to change our mathematical means of extrapolating from our measured points to our safe dose points. To do this, we use basically high-content analytical technologies, like genomics, proteomics, and metabolomics. And we use that to define the molecular response to active chemicals, and then we can use this to build, basically compute a biological network model that allows us to predict adverse effects. So this idea is not just my idea of using a system's approach to toxicology, but is really a paradigm shift in the field of toxicology. It's very collaborative, hugely multidisciplinary. This is an image of a robot in the US ToxCast platform that's currently screening 8,000 chemicals in in vitro toxicity assays. So a major scientific challenge right now is taking information from different models and different platforms, and using that information to understand an adverse response that is realistic to a human, and doesn't just represent a normal adaptive response. So this is a complicated big thing that we will take into the future. And so how can we use this right now to address something like the problem of mixtures of chemicals in foods? So something we're doing at the ETH Zurich as part of a large European consortium called EuroMix is that we are taking chemicals in foods and classifying them into assessment groups based on their mechanism understood in this manner. And then we can mathematically predict their toxicological relationship. So the take home message here is that we're decoding blueprint for the toxicological effect of active compounds. This is contrasted to the traditional black box animal testing approaches. And the advantages is that it offers a scalability, and it also offers a means of really simulating and predicting adverse responses. So taking a systems toxicology approach, really and applying that into risk assessment allows us to integrate a fundamental understanding of molecular responses to chemicals. And that is enabled by these high content analytical methods and biological modeling and applying that to our practical need to ensure the safety of our food.