06 Apr Psychiatric Research Focuses On Major Hubs of Complex Brain Systems
MedicalResearch.com Interview with:
Glenn Saxe, MD
Arnold Simon Professor of Child and Adolescent Psychiatry and
Chair, Department of Child and Adolescent Psychiatry
NYU Langone’s Child Study Center
Dr. Saxe’s bio page
MedicalResearch.com: What is the background for this approach? What are the main advantages and drawbacks to the CS-CN method in psychiatry research?
Dr. Saxe: Psychiatric disorders are complex and, in all likelihood, emerge and are sustained over time because they form what is called a complex system, involving the interaction between a great many variables of different types (e.g. molecules, neurons, brain circuits, developmental, social variables). There is a strong literature on complex systems in other fields that show remarkably similar properties between vastly different types of systems. Unfortunately, data methods used in research in psychiatry are not designed to ‘see’ the possible complex systems nature of a psychiatric disorder. Our method is designed to identify networks of variables related to psychiatric disorders that, together, have properties of complex systems. If such a system is identified, it may reveal new ways to treat these disorders.
MedicalResearch.com: How can the identification of such a system lead to new treatments?
Dr. Saxe: The nature of complex systems is that they are highly robust. They exist and last in nature because they can maintain their functioning within a diversity of environments and withstand a great amount of stress. This robust nature is built into their properties. However, the properties of complex systems also reveal their vulnerabilities. We know this from a great variety of complex systems. Complex systems are most vulnerable at their most connected points: their hubs, so to speak. We conduct analyses on data sets with a great many variables related to psychiatric disorders. These analyses examine the possible connection between each pair of variables within the data set and draw a network of all connected variable pairs. What we have found is that—like known complex systems—not all variables have the same number of variables connected to them. Most have few connections, or ‘links’. A small number of variables within the network have an extraordinary number of links—a great deal more than would be expected by chance. Like other complex systems, it is through these hub variables that a system acquires its robust, adaptive properties and it is through these hub variables that a system may lose its adaptive properties. If you can identify the hub variables related to a psychiatric disorder then you may identify the targets for treatment for that psychiatric disorder.
MedicalResearch.com: How would that work?
Dr. Saxe: Think about what some consider a prototype of a complex adaptive system that many people are familiar with: the Internet. The Internet is highly adaptive. It has amazing functionality. It was largely unplanned so it is self-organizing. It also has the known properties of complex adaptive systems, including the fact that most Internet sites have few links but a small number of sites have an extraordinarily large numbers of links: Google, Facebook, etc. Here’s the main point. If you want to stop the functioning of the Internet, you can destroy an enormous number of websites and the Internet will keep going with no decrease in functioning. If, however, you take down Google and Facebook and a few more highly connected sites, you can dramatically decrease the functioning of the Internet. This is the main implication of our research. We are looking for the Googles and Facebooks of psychiatric disorders. If we know what they are then we can design interventions to target them.
MedicalResearch.com: And is that what your research showed?
Dr. Saxe: Yes. We focused on my specialty, Posttraumatic Stress Disorder. We used the CS-CN method with two independent data sets related to PTSD in children. Using this method, we were able to identify networks of variables related to PTSD, and then we tested the network for its adaptive properties. This was clearly demonstrated including the identification of a small set of variables with hugely disproportionate number of links. We then modeled what would happen to the network when these hub variables were removed. As hypothesized, this destroyed the functioning of the network. What were the hub variables? Here are a few that we found in a longitudinal data set of PTSD in injured children: CRHR1 gene, FKBP5 gene, anxiety symptoms right after injury, depressive symptoms after the injury. The two genes code for specific biological processes related to the functioning of the HPA axis. This could suggest either existing treatments or the development of new ones related to these biological processes. Anxiety and depressive symptoms right after the injury are important variables, and treatable. Our results indicate they should be vigilantly assessed and treated.
This was the first time this unique method was ever used and clearly it needs to be replicated, for PTSD and other psychiatric disorders. The most meaningful takeaway from this study is that we proved the concept. The CS-CN method now enables psychiatric researchers to have a powerful tool to identify complex systems related to psychiatric disorders and then to use knowledge of this system to identify promising treatment targets.
MedicalResearch.com: Is there anything else you would like our readers to know about the CS-CN method?
Dr. Saxe: There is one more thing, and I wasn’t able to describe it in detail but it is important: The definition of ‘link’ between two variables in the data set. If this definition is not set well then it is unlikely our method would be successful. We began this method simply using bivariate correlations as the definition of link. We discovered that this approach was not strong enough. We then integrated advanced causal discovery algorithms to define these connections. In the last 5 or 10 years there have been tremendous advances in the mathematics of causality such that experimental designs are no longer necessary. This is really important in psychiatry since most knowledge in our field is correlational. One of my colleagues and a co-author on the manuscript, Dr. Constantin Aliferis, is a world expert on causal discovery algorithms. We embedded this approach as the definition of ‘link’ to obtain the network we analyzed for complex system properties. Thus our method—for the first time—both causal and complex systems inference for psychiatric research.
MedicalResearch.com: Thank you for your contribution to the MedicalResearch.com community.
A Complex Systems Approach to Causal Discovery in Psychiatry
Note: Content is Not intended as medical advice. Please consult your health care provider regarding your specific medical condition and questions.
Glenn Saxe, MD (2016). Psychiatric Research Focuses On Major Hubs of Complex Brain Systems MedicalResearch.com