Predictive Modeling of Preterm Birth Risk utilizing Multi-Omic Vaginal Biomarkers and Machine Learning Algorithms

International Journal of Obstetrics and Gynecology | Vol. 14, No. 8, August 2026 | pp. 325–336

DOI: 10.46882/2026/IJOG/000504

Original Research Article

Title: Predictive Modeling of Preterm Birth Risk utilizing Multi-Omic Vaginal Biomarkers and Machine Learning Algorithms

Names of Authors: Oluwaseun A. Adebayo¹, Chinedu O. Okafor², Aminat Z. Yusuf¹

Authors’ Affiliations:
¹ Department of Obstetrics and Gynecology, College of Medicine, University of Lagos, Lagos, Nigeria
² Department of Maternal-Fetal Medicine, University of Nigeria Teaching Hospital, Enugu, Nigeria

Abstract:
Spontaneous Preterm Birth (sPTB) is driven by diverse pathogenic pathways that are often inadequately captured by single clinical screening variables. This prospective cohort study investigated the predictive value of combining vaginal microbiome profiling, cervicovaginal fluid cytokines, and machine learning models to identify women at risk for spontaneous preterm birth prior to 34 weeks of gestation. Vaginal swabs were obtained from 500 asymptomatic singleton pregnancies between 16 and 20 weeks. Microbial compositions were identified through 16S rRNA gene sequencing, while cytokine concentrations (IL-1beta, IL-6, and TNF-alpha) were quantified via enzyme-linked immunosorbent assay. Predictive risk classification was performed using a Random Forest algorithm compared against standard cervical length screening (≤ 25 mm). Forty-five women (9.0%) delivered spontaneously before 34 weeks. The multi-omic machine learning classifier achieved a high predictive performance, yielding a sensitivity of 88.9%, a specificity of 92.3%, and an Area Under the Curve (AUC) of 0.94. Key predictive features included high relative abundances of Gardnerella vaginalis and Prevotella, coupled with elevated cervicovaginal fluid IL-6 concentrations (> 250 pg/mL) and a shortened cervical length. In contrast, utilizing cervical length measurements alone produced a sensitivity of 53.3% and a specificity of 86.8% (p < 0.001). Integrating cervicovaginal multi-omic data with machine learning algorithms provides a highly accurate, individualized screening framework for spontaneous preterm birth, facilitating targeted prophylactic strategies during early gestation.

Keywords: Spontaneous Preterm Birth; Vaginal Microbiome; Cytokines; Machine Learning; Random Forest; Risk Stratification.

Manuscript Timeline: Received: May 17, 2026; Revised: July 08, 2026; Accepted: July 30, 2026; Published: August 09, 2026.

Citation: Adebayo OA, Okafor CO, Yusuf AZ. Predictive Modeling of Preterm Birth Risk utilizing Multi-Omic Vaginal Biomarkers and Machine Learning Algorithms. International Journal of Obstetrics and Gynecology. 2026; 14(8): 325–336.

Subscribe to read the full article: https://mail.internationalscholarsjournals.org/subscribe-to-read