9 Mar 2017
by Lesley McCollum
Similar to fingerprints, the brain contains patterns of functional connections unique to each person—a connectome “fingerprint.” A new study published February 20 in Nature Neuroscience reports that not only does this connectome individualize during adolescence, but its maturation is delayed in people with psychiatric symptoms. Conducted by a team of researchers at the University of Oslo and Oslo University Hospital, Norway, the study found that weaker connectome distinctiveness reflects a higher burden of psychopathology.
“This paper provides a clear proof of principle for how non-invasive neuroimaging techniques can be used to relate brain features to psychopathology,” said Alan Anticevic of Yale University, who was not involved in the study.
The difference in connectome distinctiveness did not emerge until about 14 years of age, which aligns with the idea that adolescence is a particularly vulnerable time during which genetics and environmental insults can combine to derail neurodevelopment and lead to the emergence of psychiatric disorders.
In an email interview with SRF, first author Tobias Kaufmann and senior author Lars Westlye observed that the neurodevelopmental sources of psychopathology are complex, and understanding them will require research on all levels, from the genetic architecture and environmental interaction, through synaptic pruning, up to the level of brain network development.
“Our current results are therefore only a small piece in the puzzle that we hope will be complemented by joint transdisciplinary efforts,” said Kaufmann and Westlye. They are already following up on these results by combining measures of brain-based similarity and individuality with genetic data.
“This is a really important leap that leverages the well-powered, carefully characterized, large-scale datasets that are beginning to be available in our community,” said Anticevic, referring to the study’s use of publicly available data from the Philadelphia Neurodevelopmental Cohort (PNC).
The analysis included functional magnetic resonance imaging data of 797 people between eight and 22 years old. Kaufmann and colleagues created multiple connectivity profiles for each person using data collected while the participants performed either a working memory or emotional recognition task. The tasks created individualized profiles, such that a participant’s connectome from one task could be used to fish out the person from a pool of connectomes generated by the other task.
The researchers considered the “connectome distinctiveness” to be how well people's connectomes could discriminate them from the rest of the group. Across the group, the connectomes became more distinct with increasing age.
Based on screening for psychiatric symptoms, the cohort was divided into subgroups of 153 healthy participants and 137 participants with increased levels of general psychopathology. At about 14 years of age, the rate of maturation in the healthy group pulled away from the clinical group, and connectome distinctiveness decreased with increasing symptom scores.
“At this stage we can only speculate on the sources of delayed distinctiveness,” said Kaufmann and Westlye. “[W]e think of it as a delayed tuning of networks needed to arrive at an individually tailored and stable connectome,” they said, noting that more research is needed to confirm the view.
“This study, like probably many others, is still subject to some of the debates of the field of what is signal and what is noise,” said Anticevic, who was concerned as to what extent these patterns may be driven by features other than psychopathology (see SRF webinar). This can be almost impossible to disentangle and will require additional converging studies that apply a variety of de-noising techniques. “That said, it’s clearly a very compelling effect.”
Promising evidence for the effect was the specificity of the findings to subnetworks, only showing differences in higher-level cortical areas such as medial-frontal and frontoparietal networks but not in subcortical networks. The data do not yet reveal the source of this specificity, which the authors suggest could arise from methodology, such as differences in signal quality, or physiology, such as more individualized wiring patterns in higher cortical areas.
The participants were also assessed based on specific psychiatric disorders, which included 107 with attention deficit disorder, 103 with prodromal schizophrenia, and 85 depression patients. Only the depression group showed significant associations with connectome distinctiveness. Whether this points to unique differences among disorders remains unresolved.
“MRI-based features may be sensitive to a general psychopathology factor, but they may lack the specificity needed to disentangle one diagnosis from another,” said Kaufmann and Westlye.
The authors also note that symptom domains for many illnesses overlap and that mental disorders are highly heterogeneous. Further research is needed into the specificity of connectome distinctiveness, which will depend not just on increasing the specificity and sensitivity of brain imaging, but also transforming the current diagnostic boundaries based on biology.
New findings bring more questions
“At the current stage, we are not yet targeting predictions in a clinical setting, as we are still in the stage of identifying and validating potentially useful features,” said Kaufmann and Westlye.
It opens up as many questions as it answers, as any impactful study may do, said Anticevic. Specifically, it remains unknown what is driving the delay in maturation, and how the maturational trajectories actually impact clinical outcomes. “The paper scratches the surface, but to define underlying neurobiological mechanisms will require careful translational work. It’s a promising lead,” he said.
According to the authors, in theory, connectome distinctiveness could indeed yield features for clinical prediction in the future, likely integrated with many features from different domains. They note that the success of the field will depend on further large-scale collaborations and data sharing, which are needed to increase reproducibility and generalizability.
Anticevic echoed the importance of collaboration and data sharing, emphasizing that the public release of the PNC dataset has been a “remarkable service to the field.
“And that kind of philosophy and framework is really what will be needed in the upcoming years to try to solve these problems, because they’re more complicated than any one lab or team could do,” he said.