Biomarkers to improve schizophrenia diagnosis and treatment are getting closer, with researchers at Neuroscience Research Australia (NeuRA) harnessing the potential of artificial intelligence (AI) for molecular testing.
Schizophrenia affects approximately one in 100 Australians, with symptoms often overlapping between psychiatric conditions, hindering early intervention, and creating an urgent need for more objective, accessible biomarkers that distinguish mental health conditions.
Researchers Dr Timothy Couttas and Associate Professor Claire Shepherd, with support from Professor Markus Leweke, Dr Michael Allwright and Anna Hoffman from the University of Sydney, used AI to explore how naturally occurring lipids, or fats, can be screened as key features in schizophrenia.
“We urgently need objective accessible biomarkers to assist with timely and accurate schizophrenia diagnosis,” Dr Couttas said.
“This study used an established artificial intelligence tool called IDEARS, to move us toward identifying potential biomarkers in the blood and cerebrospinal fluid of patients with schizophrenia. We analysed paired blood and CSF samples from more than 150 people, including individuals with untreated schizophrenia, people receiving antipsychotic medication, and healthy controls.
“This helped us identify distinct patterns of sphingolipids and endocannabinoids that differentiated people with schizophrenia from healthy individuals, with high accuracy.”
What they found
The results identified several patterns that differed between the people receiving or untreated for schizophrenia, when compared to control donors. These changes were seen in both blood and cerebrospinal fluid, which raises the possibility that blood measurements could eventually help detect changes related to brain processes.
Anandamide, a naturally occurring endocannabinoid involved in regulating brain signalling, was particularly important in distinguishing untreated schizophrenia, alongside a group of sphingolipids that also showed consistent changes in both blood and cerebrospinal fluid.
“We also observed treatment-related changes in distinct ceramide patterns, a type of fat involved in metabolism,” Dr Couttas said.
“These changes may be linked to the metabolic effects associated with these treatments, including weight gain and altered blood lipid levels.
“Some findings appeared to differ between people who were not taking antipsychotic medication and those who were being treated. If these patterns are confirmed in future research, they might help researchers develop tools to distinguish biological subgroups or monitor treatments efficacy and/or side effects.”
What’s next
Dr Couttas said using AI allowed the team to gather more insight on complex information with greater speed and efficiency.
“The machine learning approach also helped identify combinations of measurements that would have been overlooked when assessing individual results that are not statistically clear on their own,” Dr Couttas said.
“Although further research is needed, these findings suggest that combining AI with molecular testing is a novel step to help improve the biological understanding and future diagnosis of schizophrenia beyond its current symptom-based classifications.
“It gives us useful candidates to investigate further and with AI growing, I think we are going to see even more opportunities for its use in the field.”
The report was published in Molecular Psychiatry as ‘Identification of fluid endocannabinoid and sphingolipid biomarkers in schizophrenia: a machine learning approach toward clinical value’ and is available here.
