Network localization of genetic risk for schizophrenia and bipolar disorder

Shanwen Yao, Fan Mo, Zhonghao Rao, Yu Shi, Jiajia Zhu, Yongqiang YuPsychological 2025 · DOI

TL;DR

This study uses a novel network mapping approach to pinpoint how genetic predispositions for schizophrenia (SZ) and bipolar disorder (BD) manifest in brain networks. It reveals distinct functional and structural network alterations for each disorder, but a significant overlap when both modalities are combined, suggesting shared genetic underpinnings for their overlapping symptoms.

What Problem Does This Paper Address?

Schizophrenia (SZ) and bipolar disorder (BD) are distinct psychiatric diagnoses, yet they share overlapping clinical features and genetic risk factors. This overlap has led to ongoing debate about whether they represent fundamentally different disorders or variations on a common theme. Previous neuroimaging studies have identified brain abnormalities in relatives of patients (RELs) with SZ and BD, suggesting these abnormalities might reflect genetic vulnerability. However, these studies often focused on individual brain regions and yielded inconsistent findings. The challenge is to understand how genetic risk translates into brain alterations at a network level, bridging the gap between genes and clinical presentation. This paper aims to address this by localizing genetic risk for SZ and BD to specific brain networks using a novel functional connectivity network mapping (FCNM) approach.

Methods

The core of this study is the Functional Connectivity Network Mapping (FCNM) approach. This method integrates findings from published neuroimaging studies with large-scale resting-state functional MRI (rs-fMRI) datasets. Here's a breakdown:

  1. Literature Synthesis: The researchers systematically reviewed 103 published studies to identify reported brain locations (coordinates) of functional and structural damage in relatives of SZ and BD patients.
  2. Network Construction: These identified brain locations were used as "seeds" to map functional connectivity patterns within large rs-fMRI datasets. Essentially, they looked at which brain regions were consistently active or deactivated together with these "seed" locations.
  3. Disorder-Susceptibility Networks: This process generated four distinct networks:
  4. SZ-susceptibility functional damage network
    • BD-susceptibility functional damage network
    • SZ-susceptibility structural damage network
    • BD-susceptibility structural damage network
  5. Combined Networks: To capture a more holistic picture, they also created combined SZ- and BD-susceptibility brain damage networks by merging the functional and structural findings.
  6. Network Similarity Analysis: They quantified the spatial overlap between the SZ and BD networks using the Dice coefficient to assess shared and differential genetic influences.
  7. Relation to Canonical Networks: The identified networks were compared to 14 well-established canonical brain networks (e.g., executive control, default mode) to understand their functional relevance.
  8. Validation: The robustness of the findings was tested using an independent dataset (SALD), varying seed sizes, and focusing on younger adult subgroups.

This FCNM approach is novel in its application to localizing genetic risk for these disorders, moving beyond regional analyses to a network perspective.

Datasets

The study utilized two primary rs-fMRI datasets:

  • Anhui Medical University Dataset (AMUD): This served as the discovery dataset, comprising 656 healthy adults.
  • Southwest University Adult Lifespan Dataset (SALD): This served as the validation dataset, comprising 329 healthy adults.

These datasets are publicly available resources, allowing for reproducibility and further investigation.

Key Insights and Ideas

  • Distinct Network Signatures: SZ and BD, despite their overlap, show distinct patterns of brain network vulnerability. This supports the idea that they are neurobiologically distinct disorders.
  • Functional vs. Structural Dissociation: The functional and structural damage networks for each disorder were quite different, suggesting that genetic risk might impact these two aspects of brain function through separate pathways.
  • The Power of Combined Modalities: While functional and structural networks showed low similarity between SZ and BD individually, combining both functional and structural data revealed a substantial overlap. This suggests that the shared genetic mechanisms underlying both disorders manifest in overlapping brain regions when considering both types of brain alterations.
  • Network Localization as a Bridge: FCNM proves to be a powerful tool for bridging the gap between genetic predispositions and observable brain alterations, offering a more nuanced understanding than traditional regional analyses.

Key Results

  • SZ-susceptibility functional damage network: Primarily involved the executive control and salience networks.
  • BD-susceptibility functional damage network: Principally implicated the default mode and basal ganglia networks.
  • SZ-susceptibility structural damage network: Predominantly involved the auditory and default mode networks.
  • BD-susceptibility structural damage network: Mainly implicated the language and executive control networks.
  • Network Similarity:
    • Functional damage networks: Low spatial similarity between SZ and BD (Dice coefficient = 0.04).
  • Structural damage networks: Low spatial similarity between SZ and BD (Dice coefficient = 0.23).
  • Combined SZ- and BD-susceptibility brain damage networks: Substantially increased spatial similarity (Dice coefficient = 0.45), with overlapping regions including the medial prefrontal cortex, anterior cingulate cortex, temporal pole, hippocampus, amygdala, thalamus, caudate, and nucleus accumbens.

Limitations and Future Work

The authors acknowledge several limitations:

  • Sample Characteristics: The study used healthy relatives, and while this minimizes disease-specific confounds, future studies could benefit from samples more closely matching the demographic and clinical profiles of patients.
  • Retrospective Design: The study design was retrospective, meaning some included relatives might eventually develop the disorders, potentially influencing interpretations. Prospective studies are recommended.
  • Overlap Threshold: A less conservative overlap threshold (60%) was used to identify networks, acknowledging that factors like statistical power and scanner variations can influence findings.
  • Focus on Task-Based fMRI: The analysis was primarily based on resting-state fMRI. Future research should incorporate dynamic functional connectivity and task-based fMRI for a more comprehensive understanding.
  • Neuroimaging Biomarkers Only: The study focused solely on neuroimaging. Integrating genetic data, behavioral assessments, and cognitive data could provide a more complete picture of disease mechanisms.

Future work could explore whether these shared neuroimaging features can serve as biomarkers for early intervention or prediction of disease progression. The authors also highlight the ongoing challenge of addressing reproducibility issues in neuroimaging research due to small sample sizes, clinical heterogeneity, and methodological variability.

Summary updated

Abstract

These findings may support the concept that SZ and BD represent distinct diagnostic categories from a neurobiological perspective, helping to clarify the common network substrates via which the shared genetic mechanisms underlying both disorders give rise to their overlapping clinical phenotypes.
BibTeX
@inproceedings{yao2025network,
  author = {Shanwen Yao and Fan Mo and Zhonghao Rao and Yu Shi and Jiajia Zhu and Yongqiang Yu},
  title = {{Network localization of genetic risk for schizophrenia and bipolar disorder}},
  booktitle = {Psychological},
  year = {2025},
  doi = {10.1017/s0033291725101992},
}
Figure