When
Monday, November 10, 2025, at 12:00 p.m.
Atiyeh Fotoohinasab
PhD Candidate, Nan-kuei Chen's Lab
Simona Merlini
PhD Candidate, Roberta Brinton's Lab
Keating 103 | Zoom link
Hosts: Swarna Ganesh and Kellen Chen
Atiyeh Fotoohinasab
"Harmony to Fragility: Graph-Theoretical Dissection of Multilevel Brain Network Collapse"
Abstract: The human brain is an intricate network where cognitive functions arise from the harmonious collaboration of specialized regions. Graph theory offers a powerful lens to examine this structure, allowing us to quantify how connectivity patterns support healthy function, adapt to challenges, or falter in disease states. This research presents a unified graph-theoretical view of neurodegeneration, driven by the idea that various brain regions and systems show unique vulnerabilities and disruption patterns across different disorders.
To explore this varied fragility, we introduce a spectrum of graph-based methods that reveal how network organization breaks down under pathological conditions. The talk begins by pinpointing these differential patterns in the full-spectrum functional connectome, featuring a novel graph-attack model that simulates targeted disruptions to nodes and edges in connectivity matrices. This approach tracks how local failures cascade through the network, uncovering disease-specific disconnection pathways.
Shifting to the graph-spectral realm, we decompose connectivity into frequency components to expose changes at varying scales of organization. The Spectral Wavelet–Enhanced Network-Based Inference (SWENI) framework merges graph-signal processing with statistical tools to identify distributed and modular impacts across spectral bands. Complementing this, the Multiscale Spectral Gaussian Filtering (MSGauF) uses Gaussian kernels for a coherent multiresolution breakdown, highlighting frequency-specific dysconnectivity markers that mirror the brain's physiological hierarchies.
Drawing from resting-state fMRI data in Alzheimer's and Parkinson's patients, these methods converge on a shared understanding of network vulnerability, emphasizing consistent weaknesses in default-mode, medial temporal, and frontostriatal systems. Ultimately, they depict neurodegeneration as the gradual unraveling of multilevel network balance, where diverse pathologies disrupt regional integrity, functional synergy, and spectral harmony in the brain.
Simona Merlini
"Risk and Progression Profiling in Late-Onset Alzheimer’s Disease through Temporal Pattern Mining"
Abstract: Alzheimer's disease (AD) is a highly heterogeneous neurodegenerative disorder characterized by distinct timing, sequence, and cumulative burden of risk factors that evolve over decades. Traditional analytical frameworks often treat these risk factors as static variables, overlooking the dynamic and interdependent processes that drive disease onset and progression. Understanding when and how modifiable risk factors—such as hypertension, diabetes, and depression—interact over time is critical for developing precision prevention strategies, particularly given the profound sex and APOE genotype differences in disease trajectories.
In this presentation, I will describe an integrated framework for temporal pattern mining that addresses these challenges through three complementary approaches: Cox regression with time-varying covariates to identify critical ages and risk factor transitions associated with AD diagnosis; Cumulative Event Modeling (CEM), a novel process mining approach to reconstruct cumulative trajectories of comorbidities; and Transitive Sequential Pattern Mining Plus (tSPM+), which extracts higher-order temporal dependencies among clinical events. Data were derived from two large-scale longitudinal cohorts—the UK Biobank (UKB) and the Framingham Heart Study (FHS)—including demographic, clinical, and medication information collected across decades of follow-up.
The integration of these approaches enabled the identification of population-specific temporal profiles, revealing distinct sex- and APOE-dependent pathways to AD. Time-varying Cox models uncovered critical transition ages in risk factor impact; CEM delineated cumulative comorbidity pathways converging toward AD; and tSPM+ identified transitive event sequences predictive of progression. Together, these findings advance our understanding of AD heterogeneity and establish a data-driven foundation for stratified risk assessment and early intervention strategies.
Accessibility: Persons with a disability may request a reasonable accommodation by contacting the Disability Resource Center at 621-3268 (V/TTY).