Skip to content
DraconiticResearch
Search Sources API
194,216 studies31 sourcesUpdated 33 min ago
Peer-reviewedCC BY 4.0PLOS Computational Biology

Early vs late fusion of qMIDS with machine learning based classifiers for oral cancer diagnosis from gene expression data

  • Ferid Ben Ali1,2
  • Iosif Mporas1
  • Vaishnavi Balaji1
  • Dickson Ibaranyanakaye2
  • Barbara Carey3
  • Muy-Teck Teh4
  • Sola Adeleke2
4 affiliations
  1. 1School of Physics, Engineerng and Computer Science, University of Hertfordshire, Hatfield, United Kingdom
  2. 2Curenetics Ltd, London, United Kingdom
  3. 3Guy’s and St Thomas’ NHS Foundation Trust, London, United Kingdom
  4. 4Centre for Oral Immunobiology and Regenerative Medicine, Institute of Dentistry, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom
Read full textPDFDOICite

Abstract

Different fusion architectures of the quantitative Malignancy Index Diagnostic System (qMIDS) method with machine learning classifiers for oral cancer diagnosis from gene expression data are investigated. The proposed fusion architectures are tested on geographically and ethnically diverse datasets and the evaluation results showed performance improvements with late fusion, which significantly outperformed early fusion in more matched model configurations than the reverse in every cohort, although at the best-performing configuration the two architectures were statistically indistinguishable in most cohorts. Inter-dataset analysis showed robustness of the proposed architectures when the training/test samples are from different populations. Author summary: The study tests two fusion designs of the qMIDS gene-expression method combined with machine-learning classifiers for oral cancer diagnosis. Evaluations across geographically and ethnically diverse cohorts show that late fusion performs best overall, although a direct statistical comparison of the two designs shows that its advantage over early fusion, while consistent in direction, is modest and does not reach statistical significance at the best-performing setting in most cohorts. Cross-dataset experiments (training and testing on different populations) indicate the architectures are robust to population shifts.

Related

  1. Cross-Disease Transcriptomic Signatures Shared by NAFLD and MI: Multicohort Machine-Learning Analysis and Exploratory IDI1 Expression Assessment
    Jilei Qiao, Shuran Chen, Shuo Li +4
    International Journal of Molecular SciencesPeer-reviewedCC BY 4.0Full textMolecular biology & gene regulationCrossrefEurope PMC
  2. Breath molecular feature profiling with machine learning for noninvasive diagnosis of gout: a single-center case-control study
    Zebin Cai, Xin Huang, Xuao Luo +3
    Frontiers in molecular biosciencesPeer-reviewedCC BYFull textMolecular biology & gene regulationEurope PMCPMC
  3. Machine learning-assisted chemometric analysis of salivary ATR-FTIR spectra for the diagnosis of oral submucous fibrosis
    Priya Shree, Yogender Aggarwal, Manish Kumar +6
    Frontiers in OncologyPeer-reviewedCC BY 4.0Full textBiophysicsCrossrefEurope PMCPMCEurope PMC
  4. Lactate promotes mitochondrial biogenesis and fusion via GPR81 to enhance synaptic plasticity-related protein expression in hippocampal cells
    Qinghui Shang, Xuepeng Bian, Shujie Lou
    Frontiers in cell and developmental biologyPeer-reviewedCC BYFull textGeneral biochemistryEurope PMC
  5. Deep learning-based identification of N6-methyladenine sites via hybrid feature fusion and SHAP-driven feature selection
    Mai Alzamel, Islam Uddin, Salman Khan +2
    Scientific ReportsPeer-reviewedCC BY-NC-NDFull text6 citedMolecular biology & gene regulationOpenAlex

Cite

BibTeXRIS
BibTeX
@article{ali2026early,
  title = {{Early vs late fusion of qMIDS with machine learning based classifiers for oral cancer diagnosis from gene expression data}},
  author = {Ferid Ben Ali and Iosif Mporas and Vaishnavi Balaji and Dickson Ibaranyanakaye and Barbara Carey and Muy-Teck Teh and Sola Adeleke},
  journal = {PLOS Computational Biology},
  year = {2026},
  month = oct,
  doi = {10.1371/journal.pcbi.1014888},
  url = {https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014888},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Early vs late fusion of qMIDS with machine learning based classifiers for oral cancer diagnosis from gene expression data
AU  - Ferid Ben Ali
AU  - Iosif Mporas
AU  - Vaishnavi Balaji
AU  - Dickson Ibaranyanakaye
AU  - Barbara Carey
AU  - Muy-Teck Teh
AU  - Sola Adeleke
PY  - 2026
DA  - 2026/10/28
JO  - PLOS Computational Biology
DO  - 10.1371/journal.pcbi.1014888
UR  - https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014888
AB  - Different fusion architectures of the quantitative Malignancy Index Diagnostic System (qMIDS) method with machine learning classifiers for oral cancer diagnosis from gene expression data are investigated. The proposed fusion architectures are tested on geographically and ethnically diverse datasets and the evaluation results showed performance improvements with late fusion, which significantly outperformed early fusion in more matched model configurations than the reverse in every cohort, although at the best-performing configuration the two architectures were statistically indistinguishable in most cohorts. Inter-dataset analysis showed robustness of the proposed architectures when the training/test samples are from different populations. Author summary: The study tests two fusion designs of the qMIDS gene-expression method combined with machine-learning classifiers for oral cancer diagnosis. Evaluations across geographically and ethnically diverse cohorts show that late fusion performs best overall, although a direct statistical comparison of the two designs shows that its advantage over early fusion, while consistent in direction, is modest and does not reach statistical significance at the best-performing setting in most cohorts. Cross-dataset experiments (training and testing on different populations) indicate the architectures are robust to population shifts.
SN  - 1553-7358
SN  - 1553-734X
ER  -