Genetically-Informed Brain Networks: A Deep Learning Approach to Mapping Neuroanatomical Variability
Poster No:
1545
Submission Type:
Abstract Submission
Authors:
Marilena De Pian1, Junhao Wen2, Christos Davatzikos3
Institutions:
1Department of Bioengineering, University of Pennsylvania, Philadelphia, PA, 2Columbia University, New York, NY, 3Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA
First Author:
Co-Author(s):
Introduction:
Advances in brain imaging genetics (Shen, 2020) have propelled the field of precision medicine by generating imaging-derived phenotypes linked to an array of genetic variants (Wen, 2023) (Yang, 2024). This is particularly relevant in polygenic neurodegenerative diseases like Alzheimer's Disease (AD), in which genetic variants exert pleiotropic effects along the whole genome (Lin, 2021). Here, we introduce a deep learning model that provides interpretable summarizations of brain structural networks jointly guided by variations from brain MRI and genetics.
Methods:
Our approach integrates two generative models: an orthonormal projective Nonnegative Matrix Factorization (Sotiras, 2015) autoencoder and a Variational Autoencoder (VAE). This combination identifies genetically influenced patterns of brain structure from MRI data, partitioning the brain into anatomical brain elements (ABEs) that correlate with specific genetic variations, including those associated with disease, in a data-driven manner. To enhance interpretability, the generative models are designed to disentangle imaging latent variables into components influenced by genetic and non-genetic factors, providing a nuanced understanding of brain structure. We trained and evaluated our model on simulated data (N=1500) to validate patterns in brain structure influenced by genetic variations. We further applied the model to clinical data from CN, AD, and MCI patients (N=1564) from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset for 54 AD-related genetic variants.
·Model overview
Results:
The model demonstrated robust capability in detecting specific imaging and genetic patterns and elucidating the associations between them when applied to both simulated and real voxel-wise brain volumetric measurements and SNP data. In simulated experiments, the model effectively clustered and disentangled voxels based on ground truth, demonstrating superior performance in both qualitative and quantitative contexts. Specifically, it outperformed imaging-only models in two evaluation experiments. First, the genetically-informed latent representation of the brain achieved better performance in classifying the ground truth atrophy types of simulated subjects compared to imaging-informed representations alone. Second, the genetically-informed representation exhibited a stronger correlation with the ground truth latent structure than the non-genetically influenced representations. In clinical settings, it successfully detected AD-informed ABEs and established a comprehensive brain anatomy dictionary that supports further analysis. In particular, when applied to the ADNI data, the model effectively grouped brain regions based on structural covariance and the covariance of their genetic associations. This led to a novel flexible brain parcellation distinct from those based solely on structural covariance. We extracted 5 genetically-informed ABEs. More specifically, the model uniquely clustered the subcortical regions and the orbitofrontal cortex together, distinguishing them from the rest of the frontal lobe. Additionally, the cerebellum was partitioned, with most regions grouped as non-genetically informed, except for the inferior posterior lobe of the cerebellar vermis and hemisphere. These differentiations were absent when genetic information was excluded. We also pinpointed specific SNPs associated with each ABE. For instance, rs429358 (APOE) and rs41289512 (NECTIN2) were associated with the genetically-informed ABE of the temporal and hippocampus regions, aligning with existing literature (Kim, 2022). Several ABCA7 polymorphisms (rs4147929, rs3752246, rs111278892) and CASS4 polymorphisms (rs7274581, rs6014724, rs6024870) were clustered together in association with specific ABEs, respectively.
·Application of AE – VAE on ADNI dataset for 54 AD – related SNPs
Conclusions:
These insights highlight the model's ability to design a brain atlas that captures the complexity of genetic and neuroanatomical heterogeneity, refining our understanding of how genetic factors influence brain anatomy.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s)
Genetics:
Genetic Modeling and Analysis Methods 2
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
Methods Development 1
Segmentation and Parcellation
Keywords:
Computational Neuroscience
Degenerative Disease
Machine Learning
Modeling
Phenotype-Genotype
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:
Provide references using APA citation style.
Lin, E., Lin, C. H., & Lane, H. Y. (2021). Deep learning with neuroimaging and genomics in Alzheimer’s disease. International Journal of Molecular Sciences, 22(15), 7911.
Shen, L., & Thompson, P. M. (2020). Brain imaging genomics: Integrated analysis and machine learning. Proceedings of the IEEE, 108(1), 125–162.
Sotiras, A., Resnick, S. M., & Davatzikos, C. (2015). Finding imaging patterns of structural covariance via non-negative matrix factorization. NeuroImage, 108, 1–16.
Wen, J., Nasrallah, I. M., Abdulkadir, A., Satterthwaite, T. D., Yang, Z., Erus, G., Robert-Fitzgerald, T., Singh, A., Sotiras, A., Boquet-Pujadas, A., Mamourian, E., Doshi, J., Cui, Y., Srinivasan, D., Skampardoni, I., Chen, J., Hwang, G., Bergman, M., Bao, J., Veturi, Y., … Davatzikos, C. (2023). Genomic loci influence patterns of structural covariance in the human brain. Proceedings of the National Academy of Sciences of the United States of America, 120(52), e2300842120.
Yang, Z., Wen, J., Erus, G., Govindarajan, S. T., Melhem, R., Mamourian, E., Cui, Y., Srinivasan, D., Abdulkadir, A., Parmpi, P., Wittfeld, K., Grabe, H. J., Bülow, R., Frenzel, S., Tosun, D., Bilgel, M., An, Y., Yi, D., Marcus, D. S., LaMontagne, P., … Davatzikos, C. (2024). Brain aging patterns in a large and diverse cohort of 49,482 individuals. Nature Medicine, 30(10), 3015–3026.