Feature attention graph neural network for estimating brain age and identifying important neural connections in mouse models of genetic risk for Alzheimer's disease

Abstract

Alzheimer's disease (AD) remains one of the most extensively researched neurodegenerative disorders due to its widespread prevalence and complex risk factors. Age is a crucial risk factor for AD, which can be estimated by the disparity between physiological age and estimated brain age. To model AD risk more effectively, integrating biological, genetic, and cognitive markers is essential. Here, we utilized mouse models expressing the major APOE human alleles and human nitric oxide synthase 2 to replicate genetic risk for AD and a humanized innate immune response. We estimated brain age employing a multivariate dataset that includes brain connectomes, APOE genotype, subject traits such as age and sex, and behavioral data. Our methodology used Feature Attention Graph Neural Networks (FAGNN) for integrating different data types. Behavioral data were processed with a 2D Convolutional Neural Network (CNN), subject traits with a 1D CNN, brain connectomes through a Graph Neural Network using quadrant attention module. The model yielded a mean absolute error for age prediction of 31.85 days, with a root mean squared error of 41.84 days, outperforming other, reduced models. In addition, FAGNN identified key brain connections involved in the aging process. The highest weights were assigned to the connections between cingulum and corpus callosum, striatum, hippocampus, thalamus, hypothalamus, cerebellum, and piriform cortex. Our study demonstrates the feasibility of predicting brain age in models of aging and genetic risk for AD. To verify the validity of our findings, we compared Fractional Anisotropy (FA) along the tracts of regions with the highest connectivity, the Return-to-Origin Probability (RTOP), Return-to-Plane Probability (RTPP), and Return-to-Axis Probability (RTAP), which showed significant differences between young, middle-aged, and old age groups. Younger mice exhibited higher FA, RTOP, RTAP, and RTPP compared to older groups in the selected connections, suggesting that degradation of white matter tracts plays a critical role in aging and for FAGNN's selections. Our analysis suggests a potential neuroprotective role of APOE2, relative to APOE3 and APOE4, where APOE2 appears to mitigate age-related changes. Our findings highlighted a complex interplay of genetics and brain aging in the context of AD risk modeling.

Keywords: APOE; Alzheimer’s disease; aging; brain connectomics; diffusion MRI; graph neural network.

Figure 1.

Overview of FAGNN method integrating AD risk factor traits, behavioral metrics and brain connectome. Diffusion MRI-derived connectomes undergo quadrant attention module for edge scoring prior to GNN analysis. AD risk traits and behavioral metrics are processed by 1D and 2D CNNs, respectively. The network outputs predicted brain age and we determine the difference between brain age and chronological age as a risk factor.

Figure 2.

Mean estimation difference (MED) of brain age for the risk factors used in the model. The risk factors include APOE genotype, sex, presence of HN gene and diet. (*** p<0.001, **p<0.01, *p<0.05).

Figure 3.

The top 30 connections with highest edge scores from continuous age prediction given by FAGNN. These connections were associated the most with age variation. The regions are connected to either the left or right cingulum forming two separate but mostly symmetric sub-networks.

Figure 4.

Visualization of the top 12 connections that contributed most to age prediction based on the quadrant attention module. (A) FA values along the corresponding tracts. (B) RTOP values along the corresponding tracts. (C) Illustration of the tractography with distinct colors for each connection.

Figure 5.

FA along tract profiles for the top 6 edges identified by FAGNN. The first row shows striatum-cingulum tract, second row shows corpus callosum-cingulum tract and last row shows hippocampus-cingulum tract. First column shows FA profile of left-left connection, second column shows the corresponding tractography with FA values. Third column shows FA profile of right-right connection, and the last column shows the corresponding tractography with FA values. The FA values along each tract for age groups were significantly different with p<0.001.

Figure 6.

RTOP along tract profiles for the top 6 edges identified by FAGNN. The first row shows striatum-cingulum tract, second row shows corpus callosum-cingulum tract and last row shows hippocampus-cingulum tract. First column shows RTOP profile of left-left connection, second column shows the corresponding tractography with RTOP values. Third column shows RTOP profile of right-right connection, and the last column shows the corresponding tractography with RTOP values. The RTOP values along each tract for age groups were all significantly different with p<0.001.

Figure 7.

RTAP along tract profiles for the top 6 edges identified by FAGNN. The first row shows striatum-cingulum tract, second row shows corpus callosum-cingulum tract and last row shows hippocampus-cingulum tract. First column shows RTAP profile of left-left connection, second column shows the corresponding tractography with RTAP values. Third column shows RTAP profile of right-right connection, and the last column shows the corresponding tractography with RTAP values. The RTAP values along each tract for age groups were all significantly different with p<0.001.

Figure 8.

RTPP along tract profiles for the top 6 edges identified by FAGNN. The first row shows striatum-cingulum tract, second row shows corpus callosum-cingulum tract and last row shows hippocampus-cingulum tract. First column shows RTPP profile of left-left connection, second column shows the corresponding tractography with RTPP values. Third column shows RTPP profile of right-right connection, and the last column shows the corresponding tractography with RTPP values. The RTPP values along each tract for age groups were all significantly different with p<0.001.