A test that can be done using a blood plasma sample containing cell-free (cf)DNA can predict whether a pregnant woman will have a preterm birth or not to a good degree of accuracy.
“Preterm birth occurs in approximately 11% of all births worldwide, resulting in significant morbidity and mortality for both mothers and their offspring,” wrote Jia Tang, a researcher at the Guangdong Provincial Reproductive Science Institute in China, and colleagues in the journal PLoS Medicine.
“Identifying pregnancies at risk of preterm birth during early pregnancy may help improve interventions and reduce its incidence.”
Blood tests that sample cfDNA are already widely carried out to check on the genetic health of babies during pregnancy. This new research combines genomic sequencing and artificial intelligence (AI) led analysis on the same samples to test for preterm birth risk, which could keep costs down if rolled out on a larger scale.
Various biomarkers have been investigated for predicting spontaneous preterm birth, but it is still a hard outcome to correctly predict in early pregnancy.
In this study, Tang and colleagues carried out whole genome sequencing on cfDNA sampled from 2,590 pregnant women. Of these women, 2,072 reached full term and 518 had preterm births.
The research team carried out whole genome sequencing on the cfDNA samples, with a focus on gene promoter profiling. The samples were taken in week 15 of pregnancy in the initial discovery cohort. “In this study, we hypothesize that the nucleosome profiles of plasma cfDNA carry information about its originating tissues, which could be used to develop predictive methods for preterm birth,” they explained.
The researchers then used four machine learning models and two algorithms to analyze the data and develop a predictive genetic signature for preterm birth based on the results.
A model named PTerm (Promoter profiling classifier for preterm prediction) had the best statistical score using the area under the curve (AUC) analysis of 0.88, which is a method of differentiating people who will have a medical outcome (such as preterm birth) from those who will not. PTerm is a type of Support Vector Machine, a supervised machine learning algorithm used for classification, regression, and outlier detection tasks.
PTerm was tested in three independent groups of women to validate its accuracy and maintained a good AUC score of 0.85.
“Our data suggest that the PTerm could provide valuable preterm birth predictions in early pregnancy,” concluded the authors.
“We believe that our method serves as a critical stepping stone toward developing a noninvasive diagnostic for the early prediction of pregnancy complications. Currently, PTerm can distinguish preterm birth pregnancies from full-term pregnancies with high accuracy. Moving forward, leveraging additional data on promoter profiles across different gestational ages could facilitate developing a model for accurately predicting delivery time.”
