Fingerprinting Bipolar Disorder stages through a multivariate Machine Learning approach
Poster No:
1140
Submission Type:
Abstract Submission
Authors:
Andrea Caporali1,2, Francesca Martella3, Mauro Pettorruso4, Giovanni Martinotti4, Nicolaja Girone5, Bernardo Dell'Osso5,6,7, Salvatore Saluzzi8, Annabella Giorgio8, Claudio D'Addario3, Francesco de Pasquale1
Institutions:
1Faculty of Veterinary Medicine, University of Teramo, Teramo, TE, 2School of Advanced Studies, Center for Neuroscience, University of Camerino, Camerino, MC, 3Department of Bioscience and Technology for Food, Agriculture and Environment, University of Teramo, Teramo, TE, 4University G. D'Annunzio Chieti-Pescara, Chieti, CH, 5Department of Mental Health, University of Milan, Milano, MI, 6Aldo Ravelli” Center for Neurotechnology and Brain Therapeutic, University of Milan, Milan, MI, 7- Department of Psychiatry and Behavioral Sciences, Bipolar Disorders Clinic, Stanford University, Stanford, CA, 8Department of Mental Health and Addictions, ASST Papa Giovanni XXIII, Bergamo, Bergamo, BG
First Author:
Andrea Caporali
Faculty of Veterinary Medicine, University of Teramo|School of Advanced Studies, Center for Neuroscience, University of Camerino
Teramo, TE|Camerino, MC
Faculty of Veterinary Medicine, University of Teramo|School of Advanced Studies, Center for Neuroscience, University of Camerino
Teramo, TE|Camerino, MC
Co-Author(s):
Francesca Martella
Department of Bioscience and Technology for Food, Agriculture and Environment, University of Teramo
Teramo, TE
Department of Bioscience and Technology for Food, Agriculture and Environment, University of Teramo
Teramo, TE
Bernardo Dell'Osso
Department of Mental Health, University of Milan|Aldo Ravelli” Center for Neurotechnology and Brain Therapeutic, University of Milan|- Department of Psychiatry and Behavioral Sciences, Bipolar Disorders Clinic, Stanford University
Milano, MI|Milan, MI|Stanford, CA
Department of Mental Health, University of Milan|Aldo Ravelli” Center for Neurotechnology and Brain Therapeutic, University of Milan|- Department of Psychiatry and Behavioral Sciences, Bipolar Disorders Clinic, Stanford University
Milano, MI|Milan, MI|Stanford, CA
Salvatore Saluzzi
Department of Mental Health and Addictions, ASST Papa Giovanni XXIII, Bergamo
Bergamo, BG
Department of Mental Health and Addictions, ASST Papa Giovanni XXIII, Bergamo
Bergamo, BG
Annabella Giorgio
Department of Mental Health and Addictions, ASST Papa Giovanni XXIII, Bergamo
Bergamo, BG
Department of Mental Health and Addictions, ASST Papa Giovanni XXIII, Bergamo
Bergamo, BG
Claudio D'Addario
Department of Bioscience and Technology for Food, Agriculture and Environment, University of Teramo
Teramo, TE
Department of Bioscience and Technology for Food, Agriculture and Environment, University of Teramo
Teramo, TE
Introduction:
Staging models are gaining prominence in psychiatry as tools for delineating how mental illnesses, such as Bipolar Disorder (BD), progress through recognizable phases (Kapczinski et al., 2014). While staging in BD could improve diagnostic accuracy and tailor treatment to specific illness stages (Vieta et al., 2011), the lack of progression biomarkers and the reliance on current clinical staging models remain significant limitations. In this context, recent research has emphasized the role of resting state networks in neurobiology of BD and its progression (Martino & Magioncalda, 2022). Among them, a key role seems to be played by the "triple network" model (Menon, 2011), encompassing the frontoparietal (FPN), default-mode (DMN), and salience (SAL) networks. In terms of biological markers, aberrations in neurotrophic and inflammatory pathways have been implicated in illness onset and progression, with factors such as proinflammatory cytokines and neurotrophic factors emerging as promising candidates (Rosenblat & McIntyre, 2016; Baykara et al., 2021). Thus, in this study, we combined fMRI and genes expression data in a machine learning algorithm to develop a data-driven biological staging model for BD patients. Specifically, we aimed to: I) identify the optimal set of mRNA levels and neuroimaging features to explain BD stages; II) train a data-driven classification model; and III) provide individualized multivariate fingerprints for BD patients.
Methods:
A total of 93 BD patients were recruited and clinically assigned to one of the five stages based on the Kupka & Hillegers staging model (Kupka & Hillegers, 2012). Resting-state fMRI (3T) data was acquired over 13.5 minutes, and blood samples were collected and processed to isolate Peripheral Blood Mononuclear Cells (PBMCs). For fMRI data, dynamic functional connectivity was computed using a sliding-window method, yielding 12 features: the mean and standard deviation of connectivity within and between the DMN, FPN, and SAL networks. The mRNA levels of 10 different genes (BDNF, IL-1β, IL-2, IL-4, IL-6, IL-8, IL-19, TNFα, IFNγ, and MB-COMT) were evaluated in PBMCs samples using real-time Polymerase Chain Reaction. By conducting exhaustive feature selection, a data-driven binary classification model (Support Vector Machine) was trained to provide individual multivariate fingerprints. The optimal model corresponded to the one that achieved the highest F1-score. The analysis pipeline is summarized in Fig.1.
Results:
Our optimal model achieved an accuracy of 83% and an F1-score of 78%. Despite no penalization terms on the model size being included in the cost function, our algorithm converged to a parsimonious 8-variable model, namely 'Average DMN', 'Average DMN-SAL', 'Average FPN-SAL', 'Std FPN-SAL', 'TNFα', 'IL-1β', 'IFNγ, and 'IL-2'. Models with a higher number of variables exhibited lower performance. The optimal model comprised an equal number of genes expression and fMRI features, highlighting the importance of a multivariate approach. In Fig.2 we report the fingerprints averaged across the two classes. We also tested genes expression-only and fMRI-only models, but the optimal ones achieved respectively 70% and 68% of accuracy, demonstrating the outperformance of a multivariate approach.
Conclusions:
Our findings highlight the potential of a data-driven, multivariate biomarker approach for staging BD. By integrating brain dynamic functional connectivity and gene expression data, we developed a more accurate classification model than using either modality alone. This framework holds promise for stage-informed diagnosis, monitoring, and personalized treatment strategies in BD. Future effort will focus on expanding the sample size and incorporating additional features, including structural MRI measures, to further enhance the model's performance and robustness.
Funding: Funded by the European Union - Next Generation EU - NRRP M6C2 - Investment 2.1 Enhancement and strengthening of biomedical research in the NHS.
Funding: Funded by the European Union - Next Generation EU - NRRP M6C2 - Investment 2.1 Enhancement and strengthening of biomedical research in the NHS.
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 2
Genetics:
Genetics Other
Modeling and Analysis Methods:
Classification and Predictive Modeling 1
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling
Keywords:
Computational Neuroscience
FUNCTIONAL MRI
Machine Learning
Modeling
Multivariate
Psychiatric Disorders
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
2. Kapczinski, F., Magalhães, P. V. S., Balanzá‐Martinez, V., Dias, V. V., Frangou, S., Gama, C. S., ... & Berk, M. (2014). Staging systems in bipolar disorder: an International Society for Bipolar Disorders Task Force Report. Acta psychiatrica scandinavica, 130(5), 354-363.
3. Kupka, R. W., & Hillegers, M. H. (2012). Staging and profiling in bipolar disorders. Tijdschrift Voor Psychiatrie, 54(11), 949-956.
4. Martino, M., & Magioncalda, P. (2022). Tracing the psychopathology of bipolar disorder to the functional architecture of intrinsic brain activity and its neurotransmitter modulation: a three-dimensional model. Molecular Psychiatry, 27(2), 793-802.
5. Menon, V. (2011). Large-scale brain networks and psychopathology: a unifying triple network model. Trends in cognitive sciences, 15(10), 483-506.
6. Rosenblat, J. D., & McIntyre, R. S. (2016). Bipolar disorder and inflammation. Psychiatric Clinics, 39(1), 125-137.
7. Vieta, E., Reinares, M., & Rosa, A. R. (2011). Staging bipolar disorder. Neurotoxicity research, 19, 279-285.