Just four months after Elsa Violante was born, she began to have seizures. The spasms lasted from 30 seconds to four minutes, and they came every 15 minutes to half an hour. Her parents, Alexis and Christiane Violante, rushed her to her pediatrician near their home in suburban Toronto. He sent them to the emergency department of a downtown hospital.
The diagnosis was early infantile epileptic encephalopathy, a rare and serious condition that can have debilitating lifelong effects. “After they told us what it really was, it became terrifying,” says Alexis.
Physicians at the Hospital for Sick Children (SickKids) in Toronto tried one treatment after another to address Elsa’s seizures, but they couldn’t identify the underlying cause. So they sent the family’s blood tests to Yale Medicine’s Pediatric Genomics Discovery Program (PGDP). Using gene sequencing machines, high-performance computers, and CRISPR-Cas9 genome editing technologies, the PGDP team discovered that a rare variant of the gene NEUROD2 on chromosome 17 is responsible for Elsa’s condition.
Ultimately, a high-fat low-carbohydrate diet addressed Elsa’s seizures, but physicians can’t repair her damaged nervous system. At age 5, she is unable to walk or even sit up, and she has limited control of her arms and hands. Yet her parents hope that medical science will produce advances that will make Elsa’s life better—and that the discovery of the NEUROD2 variant will help other kids and families.
Elsa’s story illustrates a powerful melding of technologies that’s helping to solve the mysteries of life and to usher in the era of precision medicine. Gene sequencing, computer data analysis, and the simulation of scenarios using experimental genome editing techniques enable researchers to identify and understand the differences among people on the genetic level. As a result, we can better diagnose diseases and choose—or design—treatments tailored to help individual patients. “From the standpoint of medicine and health, this can be transformative—and in ways that we can’t even appreciate now,” says Saquib Lakhani, MD, assistant professor of pediatrics and clinical director of the PGDP program.
Yet this combination of tools is just one of many applications of technology to medicine that are changing the ways medical researchers do their work and physicians treat patients. For decades, technology has played a critical role in medicine. From the iron lung in 1929 through the cardiac pacemaker, the artificial heart, and robotic surgery in more recent times, engineers and medical scientists have long collaborated to produce innovations at the intersections of their fields.
Yale faculty members have been in the vanguard. Scientists and clinicians here were pioneers in the use of magnetic resonance imaging (MRI) machines. William V. Tamborlane, MD, FW ’77, professor of pediatrics, and Robert S. Sherwin, MD, FW ’74, the C.N.H. Long Professor Emeritus of Internal Medicine (Endocrinology), developed the insulin infusion pump for people with diabetes. Stuart Weinzimer, MD, professor of pediatrics, led a team that produced an artificial pancreas. And Jonathan Rothberg, PhD ’91, professor (adjunct) of genetics, developed high-speed “Next-Gen” DNA sequencing.
Perhaps the most important advance in computer-related genetics research at Yale came in 2009, when a research team led by Richard Lifton, MD, PhD, who was then chair of the Department of Genetics and is now president of Rockefeller University, used new technologies to discover the genes and biochemical mechanisms that cause hypertension. Lifton’s team developed a technique for performing whole-exome sequencing, enabling researchers to examine the protein-coding region of the human genome (about 1% of the whole) at much lower cost than sequencing the full genome.
Medical technology innovation accelerated after the Institute of Medicine in 1999 released a report titled To Err is Human: Building a Safer Health System, according to Peter Schulam, MD, PhD, chair of the Department of Urology and chief innovation and transformation officer for the Yale New Haven Health System. The report called for technology-based advances aimed at improving the quality of health care. Since then, a host of new devices and computer-based tools have been introduced that enable physicians to do their jobs better. “Technology has the potential to minimize human variability and improve patient outcomes,” Schulam says.
Yale is newly committed to playing an even more prominent role in bringing technology to bear on humanity’s most critical problems—with medicine as a top priority. The University Science Strategy Committee (USSC) last year unveiled a plan to accelerate discoveries, often through multidisciplinary collaborations. Focus areas in medicine include neuroscience, inflammation science, conquering cancer, regenerative medicine, and precision medicine. “A large fraction of the modern frontiers of scientific and medicine research is interdisciplinary,” says Peter Schiffer, PhD, vice provost for research. “You’ve got data people working with fundamental scientists, with clinicians, and with the engineers designing devices.”
Because data analysis plays an important role in most of these cross-disciplinary endeavors, Schiffer is working with leaders of schools and departments to provide researchers and students with the computational resources they need. Until recently, most faculty members and programs at Yale purchased and operated their own research computers. But in 2015, the university opened Yale Center for Research Computing (YCRC), which has five clusters of computers made available to researchers as shared resources.
The YCRC’s computers are housed in a nondescript former manufacturing building on Yale West Campus. One of the clusters is dedicated to genetic research. Named Ruddle (for Francis Ruddle, PhD, a Yale pioneer in genetic engineering), the cluster is made up of two parallel rows of metal racks containing computers, storage devices, and the networking equipment to allow them to communicate. Between the racks, it’s hot and noisy.
Ruddle has roughly the same computational power as you would get by stringing together 2,000 or so high-end laptop computers. In addition, it possesses three petabytes (three quadrillion bytes) of digital storage. That’s a lot. If you were to fill Ruddle’s storage devices with MP3-encoded songs, you could play music for 6,000 years.
The reason Ruddle needs so much computing power sits directly across the street from the data center—Yale Center for Genome Analysis (YCGA). There, scientists and technicians use million-dollar gene-sequencing machines and other tools to help faculty members and students perform their research. The gene sequencing takes place at YCGA, while the analysis of the results is done with Ruddle. In some cases, diseases are caused by variants in multiple genes, so analysis of the interplay among the genes requires a large amount of computing power.
The scale of this type of computation is tremendous. Each individual human genome contains about 3 billion DNA base pairs. It took more than a decade and about $2.7 billion to sequence the first human genome—a feat completed in 2003. Now, because of advances in gene-sequencing techniques and high-performance computing, technicians can sequence and analyze an entire genome in a couple of days for about $1,000. Sequencing of the whole exome, which is the protein coding subset of the genome in which most disease-causing variants occur, costs about $200 and can be completed in a few hours.
The PGDP research team uses whole-exome sequencing to make discoveries in such cases as Elsa Violante’s. Using techniques initially developed by Lifton, a family’s blood samples are submitted to the YCGA or another facility like it for sequencing. Next, the computers are used to identify potential variants of interest. The research team looks at these results, as well as the specific problems a child has, to find the very best candidates for testing. In the case of Elsa and another child with similar seizures, variants in the NEUROD2 gene became the focus.
The PGDP scientists used CRISPR-Cas9 genome editing in frog eggs to explore further. They introduced CRISPR-Cas9 molecules into the eggs through a tiny needle. The molecules worked together to target and eliminate the NEUROD2 sequence. These eggs were then incubated and tadpoles emerged over the next few days. Researchers observed the tadpoles’ behavior as they developed, and they noticed convulsions. Bolstered with other evidence, they concluded that NEUROD2 variants were the cause of the seizures in both children.
“We hope this discovery helps other children with infant-onset epilepsy caused by NEUROD2 variants to be identified sooner, to hopefully have appropriate seizure treatments sooner,” says Lauren Jeffries, DO, the clinical genetics coordinator for PGDP. “Having a specific genetic diagnosis can help doctors and families have better understanding of treatments, outcomes, and predictions.”
While giving a tour of the YCGA lab, its director, Shrikant Mane, PhD, professor of genetics, proudly shows off the latest gene sequencing machines, which are the size of small refrigerators. Then, with a twinkle in his eye, he removes a device the size of a smartphone from a box and waves it in the air like a magic wand. “The latest technology that’s on the market—this is a sequencing machine,” he says.
The device he’s holding is a portable DNA sequencer developed by Oxford Nanopore Technologies, which is designed to enable physicians and researchers to sequence small fragments of DNA or RNA. In one scenario, health workers in Africa could use the device to test patients in remote villages for Ebola fever. The sequencer costs just $1,000. “You can carry this in your backpack. All you need is the human samples and a laptop. This is the future,” Mane says.
While miniaturization in electronics will make such devices ever smaller and cheaper, the data sets that scientists and clinicians work with are getting ever larger. But that’s a good thing. By combining detailed genetic information from hundreds of thousands (and soon millions) of people with data from electronic medical records and information from research databases, researchers and clinicians can take into account a wealth of information when they make decisions that guide research or patient care.
A new initiative at Yale is aimed at rapidly collecting genetic data that can be used for research and care. The Generations Project, launched earlier this year by Yale Medicine, Yale New Haven Health, and other partners, aims to create a DNA biobank with genetic data from more than 100,000 people over the next three to five years. Their genetic data will be linked to their electronic health records. Participants will have their genes sequenced and screened for early detection of diseases and for variants that pose risks. In addition, the data will be available to researchers performing broad-based studies.
Electronic health records are essential. Without the ability to examine an individual patient’s biology closely and compare what’s happening with that person to what has happened to many others, it’s difficult to truly understand diseases, to treat patients optimally, and to measure treatment outcomes.
For Harlan Krumholz, MD, the Harold H. Hines, Jr. Professor of Medicine (Cardiology) and director of Yale’s Center for Outcomes Research and Evaluation, all of this new data and computing power is a bonanza. “In the past, we looked at large averages and said the results applied to everybody,” he says. “Now, with the combination of all that data and remarkable computing power at our fingertips, we can make inferences that are much more focused on individuals.”
Yale New Haven Health is in the process of building a computer system designed to fulfill the promise of precision medicine. It’s bringing data from many of its existing computing systems into a centralized “data lake” where information can be readily accessed and analyzed. This huge collection of data includes not just basic electronic health records but also genetic data; test results from imaging systems and the pathology department; and real-time data collected from monitoring devices in the intensive care unit.
Krumholz predicts that medicine will be transformed by data within a decade. “We will move from a place where research and clinical care are separate domains to one where we learn from every interaction with a patient—it’s built into the system,” he says. “With every click, the system gets smarter and more customized. We leverage everything that everyone has done and is doing.”
It’s not just the amount of data that matters; it’s also how quickly you can analyze them. High-performance computers are essential here. Krumholz can imagine a time in the not-too-distant future when a clinician who is meeting with a patient with complex health issues will be able to query a computer and get advice from an artificial intelligence program about a diagnosis or a personalized treatment plan on the spot—before the patient leaves the office.
In research settings, this kind of real-time response is already possible. Nicholas Turk-Browne, PhD ’09, a professor of psychology and neuroscience, uses high-performance computing along with functional magnetic resonance imaging (fMRI) to better understand how the brain works. Measuring brain activity by detecting changes in blood flow, it correlates activities in various segments of a person’s brain with different kinds of thinking and emotions. Today, with a collaborative research team, Turk-Browne is using the system to train people diagnosed with major depressive disorder to avoid negative thoughts and feelings. “If we can get people to change what they’re focusing on, we can avoid some of the negative symptoms,” he says.
With all of this mingling of medicine and computers, medical researchers and even clinicians are being asked to develop computer science expertise. Yale’s Center for Research Computing provides tutorials and training sessions in software coding for faculty researchers, and graduate programs have been created to teach data science and bioinformatics.
In the Yale Computational Biology and Bioinformatics program, doctoral candidates use high-performance computing to investigate such biomedical questions as how genes work together at the molecular level to mediate complex biological processes. William Meyerson, an MD/PhD candidate in the program, is studying non-disease-causing mutations—which will contribute to the understanding of disease-causing mutations. In his planned career as a physician-bioinformatician, Meyerson wants to facilitate the adoption of new computer-aided decision-making tools. “There are language and culture differences between doctors and computing experts,” he says. “My role will be to bridge the gap between my computational colleagues and my clinical colleagues.”
While many of the innovations that drive progress in medicine take place within the walls of medical schools, much of the discovery in the realm of medical technology happens in the wilds of the business world. That’s why we increasingly see faculty members who have one foot planted in academia and the other in commerce.
Jonathan Rothberg, the gene sequencing innovator, shows how this two-step is done. While he remains on the adjunct faculty at Yale, his day job is running a startup incubator based in nearby Guilford, where seven companies so far are targeting health care. In all cases, they’re developing medical devices that use deep learning, a form of artificial intelligence, to continually improve their performance. One of his companies, Butterfly Network, developed a handheld ultrasound device that costs less than $2,000.
Rothberg predicts that the combination of technology and medicine will have profound effects on the future of the human race—for instance, he believes that people will someday live to be 200 years old. “We have sequencing, we have medical records, we have high-performance computing, and we have AI,” he says. “We’ll understand the complex diseases, and we’ll be able to intervene and extend life.”