 So, what would you do if you had pole genome sequence? Would you be able to better treat your diabetes? Would you be able to know that you were at risk of early onset Alzheimer's? Would you change your diet or start to exercise to address this risk? No, we actually don't know the answers to any of these questions. Just because we can sequence the genome as well doesn't mean that we understand them. We need to change genomic data into biomedical knowledge. Our bodies are highly complex. They're made up of a number of tissues and organs. They're all made up of cells that have the same genome that encodes the complexity that make us up. So, the genome is the blueprint. But we actually don't know how to read it or interpret it. I'm going to show you a small part of your genome. We know some things about it. We know how to tell a gene. So, that's what a gene looks like. We know how to tell some signals that turn genes on and off. But we don't know most things about your genome. For example, we don't know how to interpret and predict the outcome of most mutations. So, one might be a difference between my curly hair and your straight one. Another one might mean getting cancer or not. A single letter change can cause devastating consequences. The single letter change I'm showing you on chromosome one causes a rare disease called progeria that causes children to rage at highly accelerated pace. The key challenge here is that we don't observe most disease-causing mutations. And multiple mutations can actually lead to the same outcome. All of these people have the same disease, but it might be caused by different mutations or even a combination of multiple mutations. We need a method that can tell us what a mutation will mean whether we've seen it or not. We use deep learning to be able to understand which patterns in DNA are important. Our deep convolutional method called deep C learns associations between different parts of the genome on how they are important for molecules that interact with the DNA and then uses these associations to predict with single nucleotide precision the effect of any mutation. Something that is impossible to do if you have to rely on previously observed mutations. There's hundreds of mutations in any genome. And deep C can predict which ones of those are actually disease-causing and predict specific molecular consequences of each mutation, such as effects on histone binding or transcription factors. So now we can actually start looking at patients' genomes and interpret them. Out of hundreds of mutations that we see an etched genome of a child with autism, for example, we can identify the single letter change from C to A that causes transcription factor to no longer bind and fail to turn on important protein in the brain. Furthermore, things are even more complicated. So DNA is the blueprint and it codes not simply the parts but when to make those parts. And then those parts or genes actually interact with each other in a highly complex and coordinated fashion that makes cells work. We use AI on very large collections of biomedical data to build network models that represent how these genes work with each other to make cells work and to understand what happens in disease. For example, an autism gene shown here in red might interact closely with genes involved in making connections among neurons in blue or genes involved in early development in yellow. Identifying such patterns in the networks using machine learning will enable us and does enable us to make predictions about new genes potentially involved in autism. Furthermore, we can then use networks that are tissue-specific to understand the actual role these genes play in disease. So here one of those predicted autism-associated genes that has since actually been found mutated in a child with autism is put in a specific pathway context that explains how it is actually underlying the disease. We use such machine learning methods and network models to study a number of important biomedical questions. We study a number of diseases from autism to neurodegenerative disease to cancer and we work very closely with experimentalists and clinicians to take these predictions and turn them into actionable clinical hypotheses. We're also very interested in understanding drugs and drug effects in predicting off-target effects and minimizing them and in being able to use network models to repurpose drugs from commonly studied to rare diseases. While we study questions that are deeply rooted in human health, we use information across model organisms such as mouse so that we can enable detailed experimental investigation. I started by asking you what you would do if you had your whole genome. So to really enable the promise of personalized and precision medicine, we need to not only be able to predict which mutations are disease-causing but be able to understand precise effects of each of those mutations, put it in a specific cellular context, then be able to integrate this across genes and other biomolecules such as metabolites and put this together into a picture that involves multiple cell types, multiple tissues, multiple organs and integrate detailed environmental factor models to truly be able to transfer the information about molecular biology to whole organism physiology.