Poster preview
ePoster
DUAL-NET: A HYBRID DEEP LEARNING FRAMEWORK FOR SNP PRIORITIZATION IN ALZHEIMER'S DISEASE, VALIDATED IN 5,570 INDEPENDENT INDIVIDUALS
Taeho Jo
Indiana University School of Medicine
FENS Forum 2026 (2026)
Barcelona, Spain
Board PS05-09AM-086
Presentation
Date TBA
Board: PS05-09AM-086
Event Information
Poster Board
PS05-09AM-086
Poster
View posterAbstract
Genome-based prediction of Alzheimer's disease (AD) remains challenging because whole-genome sequencing (WGS) data is high-dimensional and variants exhibit local interactions and long-range functional relationships. Current approaches typically analyze local genomic patterns or functional annotations in isolation, missing complementary information. We developed DuAL-Net (Dual Approach Local-global Network), a hybrid deep learning framework capturing local SNP interactions through sliding genomic windows and global relationships through functional annotation grouping. The framework integrates Random Forest and TabNet classifiers via out-of-fold stacking, with SNP prioritization based on optimized weighting of local and global scores. DuAL-Net was trained on ADNI WGS data (n=1,050; 607 AD, 443 cognitively normal) using 14,094 SNPs within the APOE ±50kb region, achieving AUC=0.698 (95% CI: 0.659-0.737), outperforming logistic regression (p=0.019) and univariate feature selection (p=0.026) baselines. Critically, we validated generalizability by transferring ADNI-derived SNP rankings to the completely independent ADSP ADC cohort (n=5,570; 3,837 AD, 1,733 CN). Top-ranked SNPs consistently outperformed bottom-ranked SNPs: AUC=0.686 vs. 0.516 for 100 SNPs (Δ=+0.170), 0.671 vs. 0.569 for 500 SNPs (Δ=+0.102), and 0.691 vs. 0.610 for 1,000 SNPs (Δ=+0.081). The framework identified established risk variants rs429358 (APOE ε4) and rs7412 (APOE ε2) across all cross-validation folds. These results demonstrate that integrating local genomic context with global functional annotations improves variant prioritization, and that DuAL-Net rankings generalize to independent populations, supporting its utility for identifying predictive genetic variants in AD.
Recommended posters
FUNCTIONAL EVALUATION OF GENETIC VARIANTS RELATED TO POLYGENIC RISK FOR ALZHEIMER’S DISEASE USING A MASSIVELY PARALLEL REPORTER ASSAY
Danna Perlaza Meneses, Laia Lidón, Alba Cervantes-González, Xuanyang Bai, María Dolores Capilla-López, Esther Álvarez-Sánchez, Natalia Valle-Tamayo, Aleyda Benitez Amaro, Borja Moya-Llamas, Érika Sánchez-Aced, Joaquim Aumatell Escabias, Laia Muñoz, Laura Cervera-Carles, Sònia Sirisi-Dolcet, Oriol Dols-Icardo, Alberto Lleó, Olivia Belbin
A SCALABLE WORKFLOW FOR OMICS-BASED AD DRUG SCREENING USING A NOVEL 3D NEUROGLIAL MODEL
Jessica Valentina Montgomery, Julien Klimmt, Carolina Gonçalves, Jonas Schulte-Schrepping, Elena De Domenico, Marc D. Beyer, Joachim L. Schultze, Caterina Carraro, Dominik Paquet
INTEGRATIVE ANALYSIS OF MICROGLIAL SINGLE-CELL TRANSCRIPTOMES IN AD MOUSE MODELS AND HUMAN DISEASE
Marina Guillot Fernández, Aysha Bhojwani Cabrera, Alejandro Expósito Coca, Yasmina Manso Sanz, Eduardo Soriano García, José López Atalaya
LINKING AXONAL PATHOLOGY TO ELN DYSFUNCTION VIA POLYGENIC RISK IN HUMAN TISSUE AND HIPSC MODELS OF ALZHEIMER’S DISEASE
Emily Turschak, Shannon Rose, Kevin Takasaki, Wan-Qing Yu, Steven Cook, Kareena Villalobos, Elizabeth Guadarrama, Russel Torres, Connor Laughland, Eric Perlman, Ayana Hellevik, Souyma Chatterjee, C. Dirk Keene, Jessica Young, R. Clay Reid
MULTI-OMIC PROFILING OF RETROTRANSPOSON-LINKED MECHANISMS DRIVING ACCELERATED ALZHEIMER’S DISEASE IN DOWN SYNDROME
Paulina Carriba, Hatice Recaioglu, Thomas Cahill, Xiangpeng Guo, Guanshunchen Shang, Miguel A. Esteban, Mara Dierssen
UNRAVELING AGE-DEPENDENT BEHAVIORAL AND MOLECULAR FEATURES IN NOVEL MOUSE MODELS OF ALZHEIMER’S DISEASE IN DOWN SYNDROME: IMPLICATIONS FOR PATHOGENESIS AND TARGET IDENTIFICATION
Monika Rataj Baniowska, Paige Mumford, Francesca Prestia, Pauline Stephan, Millie Beament, Marie-Christine Birling, Chiara Lanzillotta, Letizia Ciafardini, Eugenio Barone, Gloria Lau, Claire Chevalier, Chadia Nahy, Nadia Messaddeq, Thais Lestra, Yixing Wu, Valerie Nalesso, Fabio Di Domenico, Frances Wiseman, Yann Herault