Related Experiment Video
Updated: May 9, 2026
Generating the Transcriptional Regulation View of Transcriptomic Features for Prediction Task and Dark Biomarker Detection on Small Datasets
Published on: March 1, 2024
Biomarker interactions are better predictors of spontaneous preterm birth
G Bhat1, S M Williams, G R Saade
11Department of Obstetrics and Gynecology, Division of Maternal-Fetal Medicine Perinatal Research, The University of Texas Medical Branch, Galveston, TX, USA.
Objective:
The objective of this study was to assess the role of biomarker interactions as predictors of spontaneous preterm birth (PTB) using multifactor dimensionality reduction (MDR) analysis. With MDR, a nonparametric, unsupervised, model-free approach, we tested for biomarker interactions within maternal-fetal compartments in 2 racial groups: African Americans (AA) and Caucasians (C).
Study Design:
A total of 36 biomarkers from maternal plasma (MP), cord plasma (CP), and amniotic fluid (AF) were analyzed from 191 patients. The MDR combined attribute selection, construction, and classification to detect biomarker interactions that were assessed for generality and significance using 10× cross-validation and permutation testing. Selected significant interactive models were replicated with additional samples.
Results:
The interactive model containing interleukin (IL)-2, angiopoietin 2 (ANGPT-2), and IL-6 receptor was significant in AA MP. In AA CP, the IL-8 and tumor necrosis factor (TNF) receptor 1 model was significant. In AA AF, the ANGPT-2 and macrophage inflammatory protein 1 alpha model was significant. Replication of the AA MP model using 54 additional AA MP samples confirmed predictability of these biomarkers. In C AF, interaction was observed between ANGPT-2, monocyte chemotactic protein 3, and TNF-α, but no other interactions were significant in C.
Conclusions:
Using MDR, we identified biomarker interactions that are predictors of PTB even in the absence of a main effect with a single biomarker.
