Multi-Omics AI Model Improves Preterm Birth Prediction Accuracy

Preterm birth (PTB) remains one of the leading causes of maternal and neonatal morbidity and mortality worldwide, with around 15 million premature births each year, or roughly 11% of all births. The earlier a baby is born, the greater the associated health risks. Despite decades of research, PTB incidence remains high, as predicting risk is extremely difficult due to its complex, multi-factorial causes. Researchers have now developed a model that boosts PTB prediction accuracy to nearly 90%.

The researchers, led by BGI Genomics (Shenzhen, China), created GeneLLM, a large language model integrating genomics and transcriptomics. The approach analyzed cell-free DNA (cfDNA) and cell-free RNA (cfRNA) circulating in maternal blood to build predictive models. This study marks the first time multi-omics and LLMs were combined for PTB risk prediction.

Image: The AI model boosts preterm birth prediction accuracy to nearly 90% (Photo courtesy of Adobe Stock)

The nested case-control study included 682 pregnant women, with plasma samples collected for cfRNA and cfDNA sequencing. Researchers designed three predictive models: cfDNA-only, cfRNA-only, and an integrated cfDNA+cfRNA version. Published in npj Digital Medicine, results showed all models achieved high accuracy above 80%, with the integrated model reaching an AUC of 0.89, outperforming single-omics methods.

Importantly, RNA editing levels were significantly higher in preterm cases, and models based on these features achieved an AUC of 0.82, suggesting a mechanistic role of RNA editing in PTB. This provides new molecular insights while validating RNA editing as a promising biomarker. The findings highlight how cfDNA and cfRNA provide complementary information to strengthen prediction.

By integrating more clinical data, the model could become even more precise and transform prenatal screening practices. It demonstrates the potential of multi-omics with AI to offer earlier identification and intervention for at-risk pregnancies. Apart from prediction, the framework reveals new biological targets such as RNA editing, paving the way for novel preventive or therapeutic strategies.

“Our study shows that integrating cfDNA and cfRNA with LLM outperforms conventional methods in predicting PTB,” said Dr. Zhou Si, Chief Scientist at BGI Genomics’ IIMR and first author of the study. “Importantly, the model is efficient, resource-light, and ready for clinical translation. Beyond prediction, our findings also reveal RNA editing as a promising new target for understanding and regulating PTB.”

Related Links:
BGI Genomics