Forecasting the convergence of artificial intelligence and precision medicine

Image of artificial DNA, which in combination with other artificial intelligence could contribute to an artificial model of the immune system
Will an artificial model of the immune system be the key to discovering new, precision vaccines?

This is part I of a two-part blog series recapping the 2018 BIO International Convention.

At the 2018 BIO International Convention last week, it was clear what’s provoking scientific minds in industry and academia — or at least those of the Guinness-world-record-making 16,000 people in attendance. Artificial intelligence, machine learning and their implications for tailor-made medicine bubbled up across all BIO’s educational tracks and a majority of discussions about the future state of biotechnology. Panelists from Boston Children’s Hospital also contributed their insights to what’s brewing at the intersection of these burgeoning fields.

Isaac Kohane, MD, PhD, former chair of Boston Children’s Computational Health and Informatics Program, spoke on a panel about how large-scale patient data — if properly harnessed and analyzed for health and disease trends — is a virtual goldmine for precision medicine insights. Patterns gleaned from population health data or electronic health records, for example, could help identify which subgroups of patients who might respond better to specific therapies.

According to Kohane, who is currently the Marion J. Nelson Professor of Biomedical Informatics and Pediatrics at Harvard Medical School (HMS), we will soon be leveraging artificial intelligence to go through patient records and determine exactly what doctors were thinking when they saw patients.

“We’ve seen again and again that data abstraction by artificial intelligence is better than abstraction by human analysts when performed at the scale of millions of clinical notes across thousands of patients,” said Kohane.

And based on what we heard at BIO, artificial intelligence will revolutionize more than patient data mining. It will also transform the way we design precision therapeutics — and even vaccines — from the ground up. 

Toward artificial models of the human immune system

Ofer Levy, MD, PhD, director of Boston Children’s Precision Vaccines Program, says that the traditional, one-size-fits-all concept that a vaccine should work for everyone — males, females, the young and the old — is outdated.

A precision vaccine approach is needed to immunize individuals at specific stages of life,” said Levy, also a professor of pediatrics at HMS.

For example, Levy explained during a panel about vaccine R&D that most HIV vaccine discovery efforts have historically been driven toward providing immunity to adults. But another arguably more effective approach could call for protecting individuals from early life onward, long before they become sexually active, which would move HIV vaccines from an adult to a pediatric arena.

The key to discovering new precision vaccines for HIV and other infectious diseases? Wayne Koff, PhD, of the Human Vaccines Project says discovery will be unlocked through artificial intelligence and the mapping of the human immunome, made up of all the genes that comprise the adaptive receptors on human B and T immune cells.

“It took about ten years and billion dollars to sequence the human genome and it’s transformed everything. The human immunome, however, is billons of times larger than the genome,” Koff said. He believes advances in sequencing technologies and informatics will make mapping the human immunome finally possible.

“At some point, we’ll have an artificial model of the human immune system, which will enable artificial intelligence and in silico modeling to discover all new vaccines,” Koff said.

Machine learning and discovery-driven science

Meanwhile, during a panel about next generation treatments for pain and addiction, Clifford Woolf, MB, BCh, PhD, director of Boston Children’s F.M. Kirby Neurobiology Center, discussed how he predicts machine learning will advance our understanding of pain and enable us to take a precision medicine approach to pain management. So far, the underlying challenge has been the extremely subjective nature of pain perception, varying from person to person.

“We can never know what pain feels like for another person; instead, we must identify, separate and understand the elements that contribute to pain, such as the central nervous system, anxiety, fear, etc.,” said Woolf. “Now that we are living in the big data era, we must collect and analyze as much patient information as possible — such as sleeping patterns, waking activities and more — to identify the complex signatures of pain that can serve as objective biomarkers.”

Woolf, who is also a professor of neurology and neurobiology at HMS, also spoke about the impact that he sees the newest technologies having in his laboratory.

“My graduate and postdoctoral researchers are conducting science in a totally new way, using genome-wide screen assays to identify and test new targets,” Woolf said. “We are shifting from hypothesis-driven science to discovery-driven science.”

Check back tomorrow for part II of our BIO 2018 recap.

Our latest coverage of advances in precision medicine.