Advanced Multi-omics AI Model Significantly Enhances Preterm Birth Risk Prediction to Nearly 90%

A groundbreaking study has introduced a highly accurate predictive framework for preterm birth (PTB), utilizing cutting-edge artificial intelligence (AI) and multi-omics data integration. This innovative approach has the potential to revolutionize precision obstetrics by enabling early and reliable identification of women at high risk for PTB.

The research, conducted by a collaborative team including BGI Genomics, Professor Huang Hefeng’s team, Shenzhen Longgang Maternal and Child Health Hospital, Fujian Maternity and Child Health Hospital, and OxTium Technology, was published in the journal npj Digital Medicine.

Preterm birth remains a major global health challenge, with approximately 15 million infants born prematurely each year, accounting for about 11% of all births worldwide. Premature infants face increased health risks, underscoring the importance of accurate early prediction. Despite ongoing efforts, the complexity of PTB’s causes has hindered the development of effective predictive tools.

This study introduces GeneLLM, a gene-focused large language model designed to interpret complex biological data effectively. By analyzing circulating genetic material—cell-free DNA (cfDNA) and cell-free RNA (cfRNA)—obtained from maternal blood, the researchers constructed predictive models capable of estimating PTB risk with high precision.

The study involved 682 pregnant women from whom plasma samples were collected for cfRNA and cfDNA sequencing. Three models were developed: one based solely on cfDNA, another on cfRNA, and a third combining both data types. All models employed Transformer-based architectures and achieved high performance, with the combined cfDNA + cfRNA model reaching an Area Under the Curve (AUC) of 0.89, indicating nearly 90% prediction accuracy.

Furthermore, the analysis revealed that RNA editing levels were significantly higher in preterm cases. Models based on RNA editing features alone achieved an AUC of 0.82, offering new molecular insights into PTB’s mechanisms. These findings suggest that RNA editing may play a crucial role in the onset of preterm birth.

Dr. Zhou Si, the study’s lead author and Chief Scientist at BGI Genomics' IIMR, stated that integrating cfDNA and cfRNA data with large language models surpasses traditional methods in predicting PTB. The model is efficient and resource-friendly, making it suitable for clinical application. Beyond prediction, the study highlights RNA editing as a promising molecular target for future research.

This pioneering work underscores how AI and multi-omics data integration are transforming prenatal care, providing a powerful tool for early detection and intervention of at-risk pregnancies. Such advancements pave the way for improved maternal and neonatal health outcomes worldwide.

For more detailed information, see the original publication: npj Digital Medicine.