{"id":{"repo_id":"brock","oai_identifier":"oai:brocku.scholaris.ca:10464/20376"},"canonical_url":"https://search.dev.ndltd.org/etd/brock/oai:brocku.scholaris.ca:10464/20376","repository":{"repo_id":"brock","name":"Brock University","base_url":"https://brocku.scholaris.ca/server/oai/request"},"display":{"title":"Graph Attention Mechanisms for Modeling Pathway-Level Importance from Gene Expression","abstract":"High-throughput sequencing technologies have revolutionized the analysis and profiling of gene expression data, enabling a comprehensive understanding of the complex mechanisms underlying molecular biology. However, the high dimensionality and noise inherent in gene expression data pose significant challenges for biomarker discovery and molecular subtype classification in cancer. While deep learning models have shown promise in capturing complex transcriptomic patterns, their black-box nature limits their interpretability for downstream biological insight and analysis. Given that cancer phenotypes may arise from the dysregulation of gene networks, incorporating pathway-level biological knowledge directly into model design offers a principled approach to improve robustness, interpretability, and predictive performance. This work presents a pathway-aware variational autoencoder which integrates curated gene-pathway structure into latent representation learning. This includes the integration of gene interactions through graph neural networks, and attentionbased hierarchical pooling from genes to pathways to latent sample representations. Across experimental trials, this architecture demonstrated stable training which had matched, if not improved, reconstruction and classification performance relative to gene-only baselines. Moreover, the use of attention mechanisms enabled the extraction of task-dependent gene importance scores, providing intrinsic interpretability to the model’s predictions. Conclusively, this work demonstrates that pathway-aware representation learning unifies predictive performance with biologically grounded interpretability, enabling robust systems-level biomarker discovery and molecular subtype classification in cancer.","abstract_html":"High-throughput sequencing technologies have revolutionized the analysis and profiling of gene expression data, enabling a comprehensive understanding of the complex mechanisms underlying molecular biology. However, the high dimensionality and noise inherent in gene expression data pose significant challenges for biomarker discovery and molecular subtype classification in cancer. While deep learning models have shown promise in capturing complex transcriptomic patterns, their black-box nature limits their interpretability for downstream biological insight and analysis. Given that cancer phenotypes may arise from the dysregulation of gene networks, incorporating pathway-level biological knowledge directly into model design offers a principled approach to improve robustness, interpretability, and predictive performance. This work presents a pathway-aware variational autoencoder which integrates curated gene-pathway structure into latent representation learning. This includes the integration of gene interactions through graph neural networks, and attentionbased hierarchical pooling from genes to pathways to latent sample representations. Across experimental trials, this architecture demonstrated stable training which had matched, if not improved, reconstruction and classification performance relative to gene-only baselines. Moreover, the use of attention mechanisms enabled the extraction of task-dependent gene importance scores, providing intrinsic interpretability to the model’s predictions. Conclusively, this work demonstrates that pathway-aware representation learning unifies predictive performance with biologically grounded interpretability, enabling robust systems-level biomarker discovery and molecular subtype classification in cancer.","abstract_has_math":false,"creators":["Villena, Marcus"],"institution":"Brock University","degree_name":"M.Sc. Biological Sciences","degree_level":"Master","degree_discipline":"Faculty of Mathematics and Science","degree_department":"Department of Biological Sciences","school":null,"contributors":[],"advisors":["Yifeng, Li"],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026-06-11","date_published":"2026-06-11","updated_at":"2026-07-24T01:23:21Z","subjects":["NATURAL SCIENCES::Biology::Cell and molecular biology","TECHNOLOGY::Information technology::Computer science::Computer science"],"languages":["eng"],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/10464/20376","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Yifeng, Li"]},{"key":"dc:contributor.department","label":"Department","values":["Department of Biological Sciences"]},{"key":"dc:creator","label":"Author","values":["Villena, Marcus"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-06-15T17:11:00Z"]},{"key":"dc:date.issued","label":"Date","values":["2026-06-11"]},{"key":"dc:publisher","label":"Institution","values":["Brock University"]},{"key":"dc:type","label":"Dc Type","values":["Electronic Thesis or Dissertation"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Faculty of Mathematics and Science"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Master"]},{"key":"thesis:degree_name","label":"Degree Name","values":["M.Sc. Biological Sciences"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["NATURAL SCIENCES::Biology::Cell and molecular biology","TECHNOLOGY::Information technology::Computer science::Computer science"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["eng"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/10464/20376"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["High-throughput sequencing technologies have revolutionized the analysis and profiling of gene expression data, enabling a comprehensive understanding of the complex mechanisms underlying molecular biology. However, the high dimensionality and noise inherent in gene expression data pose significant challenges for biomarker discovery and molecular subtype classification in cancer. While deep learning models have shown promise in capturing complex transcriptomic patterns, their black-box nature limits their interpretability for downstream biological insight and analysis. Given that cancer phenotypes may arise from the dysregulation of gene networks, incorporating pathway-level biological knowledge directly into model design offers a principled approach to improve robustness, interpretability, and predictive performance. This work presents a pathway-aware variational autoencoder which integrates curated gene-pathway structure into latent representation learning. This includes the integration of gene interactions through graph neural networks, and attentionbased hierarchical pooling from genes to pathways to latent sample representations. Across experimental trials, this architecture demonstrated stable training which had matched, if not improved, reconstruction and classification performance relative to gene-only baselines. Moreover, the use of attention mechanisms enabled the extraction of task-dependent gene importance scores, providing intrinsic interpretability to the model’s predictions. Conclusively, this work demonstrates that pathway-aware representation learning unifies predictive performance with biologically grounded interpretability, enabling robust systems-level biomarker discovery and molecular subtype classification in cancer."]},{"key":"dc:title","label":"Title","values":["Graph Attention Mechanisms for Modeling Pathway-Level Importance from Gene Expression"]}]}],"canonical_facts":{"dc:contributor.advisor":["Yifeng, Li"],"dc:contributor.department":["Department of Biological Sciences"],"dc:creator":["Villena, Marcus"],"dc:date.accessioned":["2026-06-15T17:11:00Z"],"dc:date.issued":["2026-06-11"],"dc:description.abstract":["High-throughput sequencing technologies have revolutionized the analysis and profiling of gene expression data, enabling a comprehensive understanding of the complex mechanisms underlying molecular biology. However, the high dimensionality and noise inherent in gene expression data pose significant challenges for biomarker discovery and molecular subtype classification in cancer. While deep learning models have shown promise in capturing complex transcriptomic patterns, their black-box nature limits their interpretability for downstream biological insight and analysis. Given that cancer phenotypes may arise from the dysregulation of gene networks, incorporating pathway-level biological knowledge directly into model design offers a principled approach to improve robustness, interpretability, and predictive performance. This work presents a pathway-aware variational autoencoder which integrates curated gene-pathway structure into latent representation learning. This includes the integration of gene interactions through graph neural networks, and attentionbased hierarchical pooling from genes to pathways to latent sample representations. Across experimental trials, this architecture demonstrated stable training which had matched, if not improved, reconstruction and classification performance relative to gene-only baselines. Moreover, the use of attention mechanisms enabled the extraction of task-dependent gene importance scores, providing intrinsic interpretability to the model’s predictions. Conclusively, this work demonstrates that pathway-aware representation learning unifies predictive performance with biologically grounded interpretability, enabling robust systems-level biomarker discovery and molecular subtype classification in cancer."],"dc:identifier.uri":["https://hdl.handle.net/10464/20376"],"dc:language.iso":["eng"],"dc:publisher":["Brock University"],"dc:subject":["NATURAL SCIENCES::Biology::Cell and molecular biology","TECHNOLOGY::Information technology::Computer science::Computer science"],"dc:title":["Graph Attention Mechanisms for Modeling Pathway-Level Importance from Gene Expression"],"dc:type":["Electronic Thesis or Dissertation"],"thesis:degree_discipline":["Faculty of Mathematics and Science"],"thesis:degree_level":["Master"],"thesis:degree_name":["M.Sc. Biological Sciences"]},"updated_at":"2026-07-24T01:23:21Z"}