Computational Psychiatry: How will researchers adapt to the A.I. revolution?

Digital health and “computational psychiatry”…how will clinical and basic researchers adapt to the artificial intelligence revolution in research?

Health care, drug development and disease diagnostics are increasing moving towards artificial intelligence supported solutions and big data. The result is that large companies, like Apple, Google and Amazon, are stepping into the field. We have even begun to see more and more articles in the top journals Nature, Science and Cell coming from companies with nearly infinite funds, talent and resources. This creates an incredibly competitive new world for individual academics surviving on NIH R01 grants. Artificial intelligence (A.I.) technologies change and evolve quickly, are extremely complex and specialized and can easily yield artifacts. Clinicians and basic researchers focused on making new discoveries and understanding the mechanistic basis of disease and different biological processes cannot and should not drop their work to become experts in A.I.

In the growing A.I. and big data revolution, how are foundational R01 researchers going to compete, discover and lead in a climate where the biggest companies in the world are doing more than just generating products – they are doing research?!

In fields like mental health, A.I. can have an enormous impact. Traditional psychiatric diagnoses based on the DSM are debated at best and have been criticized as “scientifically meaningless” in the worst cases. In frustration, the former leader of the NIMH, Tom Insel PhD., moved to Silicon Valley to find digital solutions for mental health in a commercial setting (TEDx talk).

The newly released draft of the NIMH 2020 Strategic Plan For Research from the new leadership, Dr. Joshua Gordon MD., PhD, strongly emphasizes the shift to digital health and a new field of “computational psychiatry”.

NIMH-supported research is “interested in mobile and other emerging technologies to develop, test, and deliver targeted prevention and treatment interventions. [They] seek digital technology that can trigger interventions based on information about the person’s current state and needs, and applications or that facilitate monitoring and early detection of changes in patient status.

NIMH is also interested in digital technologies as biomarkers and clinical outcome assessments for inclusion in clinical trials for monitoring responses to interventions.” The NIMH has a new vision of “computational psychiatry” that is aimed at developing mathematical and modeling frameworks to improve the understanding and prevention and treatment of mental illness.

The NIMH Strategic Plan for Research rightly “expects computational models can put into explicit mathematical terms testable hypotheses regarding how alterations in genes and neural circuits might affect brain function”.

From the NIMH Strategic Plan

“We seek to understand how the interplay of molecular, cellular, circuit-level, genetic, and environmental factors influence the development of mental illnesses through animal and human studies. Multidisciplinary approaches integrating statistics, mathematics, physics, computer science, and engineering will help us begin to explain how our brain predicts, interprets, alters, and responds to a complex world. Through basic science, we will achieve a more refined understanding of the brain mechanisms underlying complex behaviors, which will drive progress toward the novel interventions of tomorrow.“

Computational models can link phenotypes to circuit dysfunction and genetic variation and reveal how that dysfunction manifests in behavior and leads to progressive, chronic disorders. By taking advantage of large data sets, one can categorize brain dysfunction in a way that has the potential to lead to better diagnoses, improved biomarkers, and tailored treatments.”

This is a wonderful vision. It is the right vision. However, how does a basic researcher with expertise genetics and neural circuits now integrate big data, digital technologies and A.I. into their program to stay competitive and ahead of their field? How do we use A.I. to most effectively understand complex cognitive and behavioral patterns, rather than just derive predictive models from crude movement and heart rate tracked on a smartphone or wearable device or in a smart home – these approaches will be good for some applications, but are not powerful enough for most researchers pushing the boundaries of their fields! How can we best enable improved and radically innovative behavioral health research in medical other, like neurodegeneration, cancer or allergy?

Researchers need a platform built for discovery, which is lacking.

These are some of the many important problems we created Storyline to solve for researchers. I spent 7 years thinking about it in my own research program focused on understanding the genetic and epigenetic basis of complex behavior patterns using machine learning and our many studies in this area are just starting to come out in journals . We have learned a ton and I will write a different post about that.

In the genomics revolution, R01 researchers succeeded big time because companies built platforms that specifically aid researchers. Illumina Inc. won the genome sequencing technology race and also created powerful genotyping arrays for researchers. These products enable researchers to make new discoveries and innovate novel approaches that rely on the complex DNA sequencing instruments made by Illumina, and the cost-effective genotyping arrays that supported the genetics revolution based around genome-wide association studies (GWAS), polygenic risk association and epigenome-wide association. Bioinformatics initiatives, like Bioconductor, emerged to support the data analysis.

The A.I. revolution is different so far. Apple, Google, Amazon and other Silicon Valley companies are not just manufacturing the wearable devices and smartphones platforms – they are doing research with them and using the world’s best software engineers, data scientists, artificial intelligence scientists and mathematicians to do it.

For most R01 researchers, the learning curve involved in stepping in to compete with these large companies and organizations is way too steep.

I was an early adopter of RNASeq technology during the genomics revolution and saw how the revolution succeeded… and the mistakes that were made. I learned a lot from this.

As a founding team, we clearly see how the A.I. revolution is leaving brilliant R01 researchers and trainees behind – the real creative people that make truly foundational discoveries and create new fields.

We asked what would a platform for behavior A.I. look like? How should it work for every researcher? How could we avoid some of the systemic changes that the genomics revolution continues to suffer from.

It would be a beautiful user-friendly platform that is extremely powerful and flexible and focused on enabling discovery and understanding, as well as prediction. This is how to empower new and important projects for the research community.

Co-founder, Chief Science Officer - Storyline Health Inc.
Associate Professor of Neurobiology and Human Genetics, University of Utah School of Medicine