PSC Investigator Profile: Ravishankar Iyer, Ph.D.
In the early 1980s, Ravi Iyer, Ph.D., helped usher in a new application of artificial intelligence, one marked by the ability to predict when individual machines would fail. At the time, computer scientists had developed systems allowing machines to perform tasks normally requiring human intelligence, such as speech recognition and decision making. However, the technology remained unreliable because no one was able to predict system failures that halted all work.
That’s when Iyer, a visiting scholar at Stanford University, applied his strong expertise in systems and statistical methods to develop a new way to manage and analyze massive amounts of data. He invented an algorithm to predict when an individual machine would fail, so researchers could reassign resources in advance to recover the machine.
“That core algorithm actually stuck, that’s why our group became notorious,” says Prof. Iyer, who is now the George and Ann Fisher Distinguished Professor of Engineering and leads the DEPEND group at the Coordinated Science Laboratory at University Illinois Urbana Champaign (UIUC).
Today, Prof. Iyer is developing related algorithms to predict patient, rather than machine, outcomes. The Mayo Clinic PSC research team led by Konstantinos Lazaridis, M.D., recently joined forces with Prof. Iyer’s group.
“What’s exciting is bringing our experience to bear in a much more thoughtful way using analytical and machine learning tools that give us the best chance to impact PSC patients’ health,” says Prof. Iyer.
Such research designed to combine artificial intelligence, engineering, and medicine leads the field. As it’s very hard for machines to outdo a physician’s instinct, Prof. Iyer says using machines to augment a physician’s knowledge and instinct is the real key.
With support provided by a recently awarded NIDDK RC2 grant, the team will uncover the bigger-picture cellular networks and gene-environment interactions driving PSC by integrating multiple layers of –omics data. To identify disease signatures unique to PSC patients, the team will evaluate each –omics layer, which provides a measure of potential drivers of disease. For example, exposomics captures the sum of environmental exposures and toxins, while metabolomics measures metabolism-related chemicals, and metagenomics assesses gut bacteria.
Since each layer of –omics data represents very different things, artificial intelligence tools are required to keep the meaning of the data intact as it is merged together. “What we have to do as computer scientists is to maintain the meaning of the data as well as its size and complexity,” says Prof. Iyer. “Probably our group’s forte is how we handle these complexities and the algorithms we’ve invented to support this. I’m very hopeful that working with Dr. Lazaridis and his team we can meet this challenge.”
Such studies are a natural outgrowth of the Mayo Clinic and Illinois Strategic Alliance for Technology-Based Healthcare collaboration. Since 2010, Prof. Iyer and his graduate students have worked with Mayo Clinic physicians across department and campuses to apply analytical tools to improve patient care. For example, they’ve collaborated with the Center for Individualized Medicine (CIM) to uncover why a diabetes drug reduces triple-negative breast cancer tumors and with neurologists to predict where the onset of seizures in a patient’s brain will occur.
“The real magic has been working with – I know my colleagues at Mayo Clinic get a little embarrassed when I praise them so highly – but they are these amazing physicians who also have time to work with us, so it’s a combination of their medical and clinical insight plus their ability or their desire to make a difference that in my experience is rare,” says Prof. Iyer. “And this is why we’ve been successful; it’s not just putting random engineers and physicians together.”
More recently, Prof. Iyer and his then Ph.D. student Arjun Athreya teamed up with the Department of Psychiatry and CIM to predict a patient’s response to a specific anti-depressant drug. The question he sought to answer: How can the individual patient’s chemistry be used to understand how effective a drug will be.
“The drug is a given, I can’t change its chemistry, but I can use the patient’s record – where does the patient live, what kinds of things he or she is exposed to, how does his or her body metabolize the drug, what aspect of his or her genome impacts how the body metabolizes the drug,” says Prof. Iyer. “The more you can bring these together the more you can understand and predict if a certain drug will work for one person and not work for another person.”
His group’s success applying machine intelligence in other diseases to help physicians make better and better decisions serves as the foundation for the PSC studies. Still, it’s a computational challenge, one that Prof. Iyer says attracts his best students to Mayo Clinic, even during harsh, Rochester winters.
“What drives my interest, and a lot of our best engineering students, is the fact that they can perceptively see making a difference to the patient outcome,” says Prof. Iyer. “The work that we do with data being gathered for the PSC project now, the hope is that we will make a difference perhaps in the way Dr. Lazaridis and his team design treatments in the future.”
Prof. Iyer’s team is beginning to get their hands dirty working with data collected from a couple hundred PSC patients. Although the project is in its early stages, he has seen glimpses of data that would allow for better understanding and hopefully treatment of the disease. “What I see gives me hope,” he says.
Join others discussing PSC