EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Integrating clinlabomics‐based machine learning with in silico gene knockout to identify benign and malignant jaw tumors

Bo Wang, Beini Xu, Ling Zhang, Lin Zou, Yue He

VIEW · 10.11.2026

DOI: 10.1002/viw2.70220 ↗

Kurz gefasst

Fachartikel in der Zeitschrift VIEW, online erschienen am 10.11.2026. Verfasst von Bo Wang und 4 weiteren Autor:innen. Die Arbeit nennt 37 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurately differentiating benign from malignant jaw tumors is critical but challenging due to imaging misdiagnoses and biopsy false‐negative rates. This study aims to develop a clinlabomics‐based machine learning model for tumor differentiation and to investigate the biological correlations of malignancy. A multicenter case‒control study was conducted, enrolling 927 retrospective (338 malignant and 591 benign) and 273 prospective (59 malignant and 214 benign) jaw tumor cases. Demographic data and 52 clinlabomics parameters were collected. Following least absolute shrinkage and selection operator (LASSO) feature selection, 12 machine learning algorithms were evaluated. The optimal model was validated across internal, external, and prospective cohorts. Mechanistically, in silico knockout of signal transducer and activator of transcription 3 (STAT3) was performed on osteosarcoma single‐cell RNA sequencing data to identify downstream effectors. A Naive Bayes model using 10 core clinlabomics features showed robust discrimination malignant from benign, after optimizing 12 machine learning algorithms, which achieved an area under the curve of 0.812 in the training cohort, 0.774 in internal cohort, and 0.785 in external. In the prospective cohort, this model was yielded 79.1% accuracy, 87.4% specificity, and 86.2% negative predictive value for tumor discrimination. Mechanistic analysis revealed that interleukin‐6 (IL‐6) and STAT3 upregulation in malignant jaw tumors. Virtual STAT3 knockout disrupted the cell‒substrate adhesion pathway, significantly affecting the expression of integrin‐binding sialoprotein (IBSP). High IBSP expression was correlated with poorer prognosis (hazard ratio = 1.73). A validated and accessible Naive Bayes model is presented for distinguishing benign from malignant jaw tumors. The IL‐6/STAT3/IBSP axis is provided for biological plausibility and clinical interpretability, which could help clinical decision making for jaw tumors.

Zitieren

APA

Wang, B., Xu, B., Zhang, L., Zou, L., & He, Y. (2026). Integrating clinlabomics‐based machine learning with in silico gene knockout to identify benign and malignant jaw tumors. VIEW. https://doi.org/10.1002/viw2.70220

Vancouver

Wang B, Xu B, Zhang L, Zou L, He Y. Integrating clinlabomics‐based machine learning with in silico gene knockout to identify benign and malignant jaw tumors. VIEW. 2026. doi:10.1002/viw2.70220

BibTeX
@article{Wang2026,
  author = {Wang, Bo and Xu, Beini and Zhang, Ling and Zou, Lin and He, Yue},
  title = {Integrating clinlabomics‐based machine learning with in silico gene knockout to identify benign and malignant jaw tumors},
  journal = {VIEW},
  year = {2026},
  doi = {10.1002/viw2.70220}
}
Kostenlos starten →