EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Effects of latent feature construction and fusion strategies in deep multi-omics classification

Md Rayhan Khan iD, Masrafe Bin Hannan Siam iD, Md Fazla Elahe iD, Md Shohel Arman, S M Hasan Mahmud

PLOS One · 17.11.2026 · 21 (11) · e0362839

DOI: 10.1371/journal.pone.0362839 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PLOS One, online erschienen am 17.11.2026. Verfasst von Md Rayhan Khan und 4 weiteren Autor:innen. Beteiligte Einrichtung: Data Science Lab, Department of Software Engineering, Daffodil International University, Daffodil Smart City, Dhaka, Bangladesh. Hauptthemen laut MEDLINE: Deep Learning. Die Arbeit nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Identifying the specific subtypes of a disease is a critical step in personalizing medical treatment, yet combining the massive amounts of data from different biological layers remains a major challenge. While deep learning is great at distilling biological data into latent features, it remains to be seen if these simplified features are actually enough to represent the full complexity of a disease. In this study, we compared four distinct strategies using the BRCA and ROSMAP datasets: a View-Specific Stacked Generalization Ensemble, an End-to-End Multi-Input Deep Neural Network (Late Fusion), a Two-Stage Early Fusion Network, and our proposed Hybrid Latent-Original Feature Fusion Network. We evaluated these models based on their classification accuracy, stability, and ability to handle imbalanced data. We identified the Hybrid Latent-Original Feature Fusion Network as the most effective approach. While other methods often struggled with smaller patient groups, our hybrid model achieved the highest test accuracy 87.07% for BRCA and 82.08% for ROSMAP and weighted F1-scores (0.88 and 0.82, respectively). The strength of this architecture is dual-pathway design, which balances the stability of simplified features with the original data. This ensures that vital biological clues are not lost during the compression process. Rather than replacing one with the other, our findings confirm that we achieve the best results by combining simplified features and original biological data. This establishes a new approach for building effective deep learning models in biomedical research.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Deep LearningHumansMultiomicsNeural Networks, ComputerClassification Algorithms

Autor:innen und Institutionen

  • Data Science Lab, Department of Software Engineering, Daffodil International University, Daffodil Smart City, Dhaka, Bangladesh.

ORCID-Profile: Md Rayhan Khan, Masrafe Bin Hannan Siam, Md Fazla Elahe

Zitieren

APA

Khan, M. R., Siam, M. B. H., Elahe, M. F., Arman, M. S., & Mahmud, S. M. H. (2026). Effects of latent feature construction and fusion strategies in deep multi-omics classification. PLOS One, 21(11), e0362839. https://doi.org/10.1371/journal.pone.0362839

Vancouver

Khan MR, Siam MBH, Elahe MF, Arman MS, Mahmud SMH. Effects of latent feature construction and fusion strategies in deep multi-omics classification. PLOS One. 2026;21(11):e0362839. doi:10.1371/journal.pone.0362839

BibTeX
@article{Khan2026,
  author = {Khan, Md Rayhan and Siam, Masrafe Bin Hannan and Elahe, Md Fazla and Arman, Md Shohel and Mahmud, S M Hasan},
  title = {Effects of latent feature construction and fusion strategies in deep multi-omics classification},
  journal = {PLOS One},
  year = {2026},
  volume = {21},
  number = {11},
  pages = {e0362839},
  doi = {10.1371/journal.pone.0362839}
}
Kostenlos starten →