Academic

Graph Attention Based Prioritization of Disease Responsible Genes from Multimodal Alzheimer's Network

arXiv:2603.02273v1 Announce Type: new Abstract: Prioritizing disease-associated genes is central to understanding the molecular mechanisms of complex disorders such as Alzheimer's disease (AD). Traditional network-based approaches rely on static centrality measures and often fail to capture cross-modal biological heterogeneity. We propose NETRA (Node Evaluation through Transformer-based Representation and Attention), a multimodal graph transformer framework that replaces heuristic centrality metrics with attention-driven relevance scoring. Using AD as a case study, gene regulatory networks are independently constructed from microarray, single-cell RNA-seq, and single-nucleus RNA-seq data. Random-walk sequences derived from these networks are used to train a BERT-based model for learning global gene embeddings, while modality-specific gene expression profiles are compressed using variational autoencoders. These representations are integrated with auxiliary biological networks, includin

B
Binon Teji, Subhajit Bandyopadhyay, Swarup Roy · · 1 min read · 9 views
#cs.LG

arXiv:2603.02273v1 Announce Type: new Abstract: Prioritizing disease-associated genes is central to understanding the molecular mechanisms of complex disorders such as Alzheimer's disease (AD). Traditional network-based approaches rely on static centrality measures and often fail to capture cross-modal biological heterogeneity. We propose NETRA (Node Evaluation through Transformer-based Representation and Attention), a multimodal graph transformer framework that replaces heuristic centrality metrics with attention-driven relevance scoring. Using AD as a case study, gene regulatory networks are independently constructed from microarray, single-cell RNA-seq, and single-nucleus RNA-seq data. Random-walk sequences derived from these networks are used to train a BERT-based model for learning global gene embeddings, while modality-specific gene expression profiles are compressed using variational autoencoders. These representations are integrated with auxiliary biological networks, including protein-protein interactions, Gene Ontology semantic similarity, and diffusion-based gene similarity, into a unified multimodal graph. A graph transformer assigns NETRA scores that quantify gene relevance in a disease-specific and context-aware manner. Gene set enrichment analysis shows that NETRA achieves a normalized enrichment score of about 3.9 for the Alzheimer's disease pathway, substantially outperforming classical centrality measures and diffusion models. Top-ranked genes enrich multiple neurodegenerative pathways, recover a known late-onset AD susceptibility locus at chr12q13, and reveal conserved cross-disease gene modules. The framework preserves biologically realistic heavy-tailed network topology and is readily extensible to other complex disorders.

Executive Summary

This article proposes NETRA, a multimodal graph transformer framework for prioritizing disease-associated genes. NETRA integrates gene regulatory networks, protein-protein interactions, Gene Ontology semantic similarity, and diffusion-based gene similarity into a unified multimodal graph. The framework assigns disease-specific and context-aware relevance scores to genes using attention-driven relevance scoring. In an Alzheimer's disease case study, NETRA outperforms traditional centrality measures and diffusion models in gene set enrichment analysis. Top-ranked genes enrich multiple neurodegenerative pathways, recover a known late-onset AD susceptibility locus, and reveal cross-disease gene modules. NETRA preserves biologically realistic network topology and can be extended to other complex disorders.

Key Points

  • ▸ NETRA is a multimodal graph transformer framework for prioritizing disease-associated genes.
  • ▸ NETRA integrates multiple biological networks and assigns attention-driven relevance scores.
  • ▸ NETRA outperforms traditional centrality measures and diffusion models in gene set enrichment analysis.

Merits

Strength in addressing biological heterogeneity

NETRA effectively captures cross-modal biological heterogeneity by integrating multiple biological networks, which is a significant improvement over traditional network-based approaches.

Improved gene prioritization

NETRA's attention-driven relevance scoring and multimodal graph structure enable more accurate and context-aware gene prioritization, which is essential for understanding complex disorders.

Demerits

Scalability limitations

NETRA's performance and scalability may be affected by the size and complexity of the integrated biological networks, which could limit its application to extremely large datasets.

Limited domain generalizability

NETRA is specifically designed for Alzheimer's disease, and its performance and generalizability to other complex disorders remain to be evaluated.

Expert Commentary

NETRA presents a promising approach for prioritizing disease-associated genes by integrating multiple biological networks and applying attention-driven relevance scoring. The framework's performance in gene set enrichment analysis and its ability to recover known susceptibility loci are notable achievements. However, scalability limitations and limited domain generalizability are areas that require further investigation. As the field of network medicine continues to evolve, NETRA's contributions to the development of graph-based approaches and its application to Alzheimer's disease research highlight its potential for advancing our understanding of complex disorders.

Recommendations

  • ✓ Further evaluation of NETRA's performance and scalability on larger datasets and its application to other complex disorders is necessary to fully assess its potential.
  • ✓ The integration of additional biological networks and the development of more advanced relevance scoring methods may further enhance NETRA's performance and generalizability.

Sources

Related Articles