Development and Validation of Machine Learning Models Incorporating TGFBIp Expression for Prognosis Prediction in Epithelial Ovarian Cancer.

Authors

Yi Guo, Zhen Li, Wenwei Song, Lansheng Zhang

Published in

Cancer reports (Hoboken, N.J.). Volume 9. Issue 11.

Abstract

Epithelial ovarian cancer (EOC) exhibits substantial heterogeneity in clinical outcomes, highlighting the need for diverse prognostic tools. Transforming growth factor-β-induced protein (TGFBIp) has been implicated in ovarian cancer progression, but its incremental prognostic value within machine learning frameworks remains unexplored.
To develop and validate machine learning models incorporating TGFBIp expression to predict vital status and recurrence in EOC patients.
We analyzed 155 EOC patients from a tissue microarray. TGFBIp expression in tumor cells and stroma was assessed by immunohistochemistry. Twelve machine learning algorithms were benchmarked across four feature inclusion strategies, with TGFBIp expression sequentially added to a baseline clinicopathological model. Model performance was assessed using the area under the receiver operating characteristic curve (AUC) with bootstrap 95% confidence intervals, the area under the precision-recall curve, calibration intercept and slope, Brier score, and decision curve analysis (DCA). The incremental value of TGFBIp was formally tested by paired bootstrap comparison of test-set AUCs. SHAP analysis was performed for model interpretation. The support vector machine (SVM) with Strategy 4 (incorporating both tumor and stromal TGFBIp) was selected for vital status prediction, achieving a test-set AUC of 0.951 (95% CI 0.870-0.998), accuracy of 0.872, sensitivity of 0.920, specificity of 0.818, and negative predictive value (NPV) of 0.900. The random forest (RF) with Strategy 4 was selected for recurrence prediction, achieving a test-set AUC of 0.936 (95% CI 0.816-1.000), sensitivity of 0.974, and NPV of 0.800. Adding TGFBIp features significantly improved vital status prediction (ΔAUC +0.038, 95% CI +0.002 to +0.092 for Strategy 4 vs. baseline). Calibration was acceptable for the vital status model (test-set calibration intercept -0.24, slope 1.25, Brier score 0.071). SHAP analysis identified maximum tumor diameter, grade, stage, and TGFBIp expression as leading predictors; tumoral TGFBIp contributed to both outcomes, whereas stromal TGFBIp contributed mainly to vital status prediction. DCA suggested potential clinical utility across a broad range of threshold probabilities.
We established a machine learning framework incorporating compartment-specific TGFBIp expression for EOC prognosis. TGFBIp in tumor cells provided incremental prognostic information beyond conventional clinicopathological parameters.

PMID:
43006646
Bibliographic data and abstract were imported from PubMed on 10 Nov 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.