Multilayer Similarity Network Fusion Successfully Predicts Alzheimer's Disease Diagnosis
Poster No:
79
Submission Type:
Abstract Submission
Authors:
Bárbara Avelar-Pereira1,2, Saman Sarraf2, S. M. Hadi Hosseini2
Institutions:
1Karolinska Institutet, Stockholm, Stockholm, 2Stanford University, Stanford, CA
First Author:
Co-Author(s):
Introduction:
Alzheimer's disease (AD) is a multifactorial and heterogeneous disorder, which makes early detection particularly challenging (Ewers et al., 2011). Its pathology is driven by intricate interactions across multiple levels including brain structure and function, biospecimen information, genetics, and cognition (Tan et al., 2014). This indicates that multimodal integration of different data types allows for a more comprehensive understanding of the mechanisms underlying AD. Multilayer network community detection methods capture such complex interactions across modalities and, therefore, play a crucial role in AD prediction. In this study, we tested the performance of a Multilayer Similarity Network Fusion (SNF) framework (Wang et al., 2014; Mucha et al., 2010) in identifying and predicting AD in a multimodal dataset of 205 individuals across the AD spectrum.
Methods:
We included 205 participants (age 71.75±7.08, 95 women) from the Alzheimer's disease Neuroimaging Initiative (ADNI) who had complete data across five modalities: genetics, cerebrospinal fluid (CSF), cognition, structural MRI, and amyloid PET at baseline. The sample included cognitively healthy (CN; N=68), mild cognitive impairment (MCI; N=79), and AD (N=54) individuals. Four individuals had MCI or dementia unrelated to AD and were excluded from later statistical analyses. Data were normalized and covariates were regressed out (i.e., age, sex, intracranial volume). Spearman correlations between measures across subjects were calculated per modality to extract similarity matrices. We then applied in-house developed software that included standardizing input data, executing multimodal SNF, and 10-fold cross-validation. Lastly, the most optimized modalities were identified using an enhanced Silhouette score-based algorithm. This allows us to stabilize the outcomes and increase the robustness of results. Findings were also validated in an independent sample of 143 ADNI individuals with baseline and a 4-year follow-up.
Results:
We identified two communities: one primarily comprising AD (81.5%) and the other CN (92.6%) individuals. Those in the AD community had higher genetic risk (χ(4)=44.203, p<5.82x10-9) and poorer cognition as measured by the MMSE (t(101.636)=6.571, p<0.001), MoCA (t(203)=6.483, p<0.001), ADAS13 (t(111.827)=-7.082, p<0.001), and CDR scales (t(92.399)=-6.721, p<0.001). These individuals also had more pronounced AD pathology measured by amyloid PET (t(204)=-18.882, p<0.001) and CSF tau (t(80.451)=-9.735, p<0.001), ptau (t(88.589)=-12.359, p<0.001), and hippocampal volume (t(187) =4.391, p<0.001). AD and CN individuals in the AD-dominant community displayed similar levels of brain pathology (e.g., amyloid PET: p=0.895). Notably, CN individuals in the AD-dominant community had greater brain pathology than AD individuals in the CN-dominant community (amyloid PET: t(13)=-2.686, p=0.009; CSF tau: t(13)=-3.138, p=0.004; ptau: t(13)=-4.309, p<0.001). This suggests our approach was able to group participants deemed to be CN but who displayed higher pathology levels than individuals formally diagnosed with AD (Fig. 1). Sensitivity and specificity were 84.38% (95% CI:73.14-92.24) and 92.65% (95% CI:83.67- 97.57). The validation sample showed similar results, with the AD-dominant community also displaying the highest amyloid 4 years later (p<0.01). Finally, the most optimized modalities were amyloid PET and CSF in both samples.
·Figure 1. Values for community 1 (i.e., CN-dominant) and 2 (i.e., AD-dominant) by diagnosis group for (a) amyloid PET, (b) CSF tau, and (c) CSF ptau.
Conclusions:
Our findings indicate that multilayer SNF is promising in capturing signs of AD pathology such as brain damage and cognitive outcomes. SNF was also able to identify participants deemed as CN, but who had comparable biomarker levels to those formally diagnosed with AD. By predicting AD with high sensitivity and specificity, we show that SNF is a valuable tool to capture complex relations between different markers of a disease.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 1
Modeling and Analysis Methods:
Multivariate Approaches 2
Other Methods
Keywords:
Aging
Cerebro Spinal Fluid (CSF)
Cognition
Degenerative Disease
Positron Emission Tomography (PET)
STRUCTURAL MRI
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?
Are you Internal Review Board (IRB) certified? Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
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.
- Mucha, P. J., et al. (2010). Community structure in time-dependent, multiscale, and multiplex networks. Science, 328(5980), 876-878.
- Tan, C. C., et al. (2014). Biomarkers for preclinical Alzheimer's disease. Journal of Alzheimer's Disease, 42(4), 1051-1069.
- Wang, B., et al. (2014). Similarity network fusion for aggregating data types on a genomic scale. Nature methods, 11(3), 333-337.