Polyconnectomic score: a novel quantification of brain signatures in neuropsychiatric disorders
Poster No:
1907
Submission Type:
Abstract Submission
Authors:
Ilan Libedinsky1, Koen Helwegen2, Laura Guerrero Simon1, Marius Gruber3, Jonathan Repple4, Tilo Kircher5, Udo Dannlowski6, Martijn van den Heuvel7
Institutions:
1Vrije Universiteit, Amsterdam, NH, 2Vrije Universiteit Amsterdam, Amsterdam, North Holland, 3University Hospital, Goethe University Frankfurt, Frankfurt am Main, Hesse, 4Department of Psychiatry, Psychosomatic Medicine and Psychotherapy, University Hospital Frankfurt, Frankfurt, Hesse, 5Department of Psychiatry and Psychotherapy, University of Marburg, Marburg, Hesse, 6Institute for Translational Psychiatry, Münster, North Rhine Westphalia, 7Vrije Universiteit, Amsterdam, N/A
First Author:
Co-Author(s):
Jonathan Repple
Department of Psychiatry, Psychosomatic Medicine and Psychotherapy, University Hospital Frankfurt
Frankfurt, Hesse
Department of Psychiatry, Psychosomatic Medicine and Psychotherapy, University Hospital Frankfurt
Frankfurt, Hesse
Introduction:
Emerging evidence has shown that neuropsychiatric disorders have distinct macroscale brain connectivity patterns [Fornito et al., 2015]. Detecting these altered brain patterns with MRI remains challenging, partly due to the numerous tests needed for whole-brain analysis, which increases the risk of missing crucial findings [Helwegen et al., 2023]. To overcome this constraint, we introduce the polyconnectomic score (PCS). Drawing inspiration from polygenic scores [Torkamani et al., 2018], PCS offers an interpretable way to quantify the extent of brain circuitry linked to a disorder that is present within a connectome. We demonstrate the utility of PCS in three applications: detecting individuals with brain disorders, stratifying patients by disease predisposition, and uncovering brain-behaviour associations.
Methods:
PCS is based on connectome summary statistics [Nichols et al., 2017], reflecting both the strength and direction of the association between brain connections and a phenotype (Fig. 1). These statistics are derived by estimating connectivity differences between patients and controls (using Cohen's d), either from a prior study or by aggregation from multiple studies via a meta-analytical approach. The PCS for an out-of-sample individual is computed as the weighted average of the summary statistics and the individual's brain connectivity map, resulting in a unique score per subject.
Resting-state functional MRI data from a total of 34,570 individuals were used to reconstruct the functional connectivity [De Lange et al., 2023] of controls (n = 5,551) and patients with autism spectrum disorder (n = 1,117), schizophrenia (n = 279), attention deficit hyperactivity disorder (n = 1,064), and Alzheimer's disease (n = 223). In each dataset, we regressed out the effects of covariates such as age, sex, site, and total in-scanner motion from the functional connectivity values. We computed the PCS for each disorder, comparing PCS levels between groups using Cohen's d and FDR-corrected p-values from Student's t-tests. We stratified individuals based on psychosis liability by computing the PCS for schizophrenia and compared PCS levels among patients with schizophrenia, schizoaffective disorder, and bipolar disorder (n = 126, 59, and 72, respectively), their first-degree relatives (n = 113, 71, and 75, respectively), and healthy controls (n = 88) [Ivleva et al., 2013], using Cohen's d and FDR-corrected p-values from Student's t-test statistic. We further explored brain-behaviour associations in the UK Biobank (n = 26,673) [Sudlow et al., 2015] by measuring the Pearson correlation coefficient between PCS for schizophrenia and behavioural measurements (FDR-corrected p-values).
Resting-state functional MRI data from a total of 34,570 individuals were used to reconstruct the functional connectivity [De Lange et al., 2023] of controls (n = 5,551) and patients with autism spectrum disorder (n = 1,117), schizophrenia (n = 279), attention deficit hyperactivity disorder (n = 1,064), and Alzheimer's disease (n = 223). In each dataset, we regressed out the effects of covariates such as age, sex, site, and total in-scanner motion from the functional connectivity values. We computed the PCS for each disorder, comparing PCS levels between groups using Cohen's d and FDR-corrected p-values from Student's t-tests. We stratified individuals based on psychosis liability by computing the PCS for schizophrenia and compared PCS levels among patients with schizophrenia, schizoaffective disorder, and bipolar disorder (n = 126, 59, and 72, respectively), their first-degree relatives (n = 113, 71, and 75, respectively), and healthy controls (n = 88) [Ivleva et al., 2013], using Cohen's d and FDR-corrected p-values from Student's t-test statistic. We further explored brain-behaviour associations in the UK Biobank (n = 26,673) [Sudlow et al., 2015] by measuring the Pearson correlation coefficient between PCS for schizophrenia and behavioural measurements (FDR-corrected p-values).
Results:
Patients with autism spectrum disorder, schizophrenia, and Alzheimer's disease showed significantly higher PCS compared to controls (Cohen's d range: [0.30, 0.87], p < 0.05; Fig. 2). Including the whole-brain connectome led to a better differentiation between groups than including the most significant connections from the summary statistics. PCS enabled stratification of individuals by psychosis predisposition, differentiating patients with schizophrenia from their first-degree relatives (d = 0.42, p = 4 x 10-3), and first-degree relatives from healthy controls (d = 0.34, p = 0.034). PCS also revealed that subjects with brain patterns similar to those with schizophrenia were more likely to show lower fluid intelligence (r = -0.037, p = 1.1 x 10-5), higher neuroticism scores (r = 0.031, p = 1.5 x 10-5), and decreased levels of happiness (r = -0.023, p = 6.4 x 10-4), among other factors.
Conclusions:
PCS is a valuable tool for identifying individuals with neuropsychiatric disorders, stratifying disease risk, and uncovering brain-behaviour associations. The potential of PCS to quantify connectivity patterns across the entire connectome could significantly enhance our understanding of both healthy and diseased brain functioning.
Disorders of the Nervous System:
Neurodevelopmental/ Early Life (eg. ADHD, autism)
Psychiatric (eg. Depression, Anxiety, Schizophrenia)
Modeling and Analysis Methods:
fMRI Connectivity and Network Modeling 2
Methods Development 1
Multivariate Approaches
Keywords:
Attention Deficit Disorder
Autism
Computational Neuroscience
DISORDERS
FUNCTIONAL MRI
MRI
Schizophrenia
Other - Connectomics
1|2Indicates the priority used for review
Provide references using author date format
Fornito A, Zalesky A, Breakspear M (2015): The connectomics of brain disorders. Nat Rev Neurosci 16:159–172.
Helwegen K, Libedinsky I, van den Heuvel MP (2023): Statistical power in network neuroscience. Trends Cogn Sci:S136466132200328X.
Ivleva EI, Bidesi AS, Keshavan MS, Pearlson GD, Meda SA, Dodig D, Moates AF, Lu H, Francis AN, Tandon N, Schretlen DJ, Sweeney JA, Clementz BA, Tamminga CA (2013): Gray Matter Volume as an Intermediate Phenotype for Psychosis: Bipolar-Schizophrenia Network on Intermediate Phenotypes (B-SNIP). Am J Psychiatry 170:1285–1296.
Nichols TE, Das S, Eickhoff SB, Evans AC, Glatard T, Hanke M, Kriegeskorte N, Milham MP, Poldrack RA, Poline J-B, Proal E, Thirion B (2017): Best practices in data analysis and sharing in neuroimaging using MRI. CO M M E N TA RY 20.
Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, Downey P, Elliott P, Green J, Landray M, Liu B, Matthews P, Ong G, Pell J, Silman A, Young A, Sprosen T, Peakman T, Collins R (2015): UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age. PLOS Med 12:e1001779.
Torkamani A, Wineinger NE, Topol EJ (2018): The personal and clinical utility of polygenic risk scores. Nat Rev Genet 19:581–590.