Preeclampsia is a life-threating pregnancy disorder, especially preterm preeclampsia and early-onset preeclampsia, the heterogeneity and complexity of preeclampsia make it difficult to predict risk and develop treatments. Plasma cell-free RNA carries unique information from human tissue and may be useful for noninvasive monitoring of maternal, placental, and fetal dynamics during pregnancy. This study aimed to investigate various RNA biotypes associated with preeclampsia in plasma and develop classifiers to predict preterm preeclampsia and early-onset preeclampsia before diagnosis. We performed a novel cell-free RNA sequencing method named PALM-Seq to investigate the cell-free RNA characteristics of 715 healthy and 202 preeclampsia pregnancies before symptom onset. We explored changes in the abundance of different RNA biotypes in plasma between healthy and preeclampsia samples and built preterm preeclampsia and early-onset preeclampsia prediction classifiers using machine learning methods. Furthermore, we validated the performance of the classifiers using the external and internal validation cohorts and assessed the area under the curve and positive predictive value. We detected 77 differentially expressed genes including mRNA (44%) and miRNA (26%) between healthy and preterm preeclampsia mothers before symptom onset, which could separate preterm preeclampsia from healthy samples and play critical functional roles in preeclampsia physiology. We developed two classifiers for predicting preterm preeclampsia and early-onset preeclampsia before diagnosis based on 13 cell-free RNA signatures and two clinical features (IVF and MAP), respectively. Notably, both classifiers showed enhanced performance compared to the existing methods: the preterm preeclampsia prediction model achieved 81% area under the curve and 68% positive predictive value in an independent validation cohort (preterm PE, n=46; control, n=151); the early-onset preeclampsia prediction model had an area under the curve of 88% and a positive predictive value of 73% in an external validation cohort (early-onset PE, n=28; control, n=234). Furthermore, we demonstrated downregulation of miRNAs may play vital roles in preeclampsia through the upregulation of preeclampsia-relevant target genes. Our cohort study presented the first comprehensive transcriptomic landscape of different RNA biotypes in preeclampsia and developed two advanced classifiers with significant clinical importance for preterm preeclampsia and early-onset preeclampsia prediction before symptom onset. We demonstrated for the first time that messenger RNA, microRNA, and long noncoding RNA can simultaneously serve as potential biomarkers of preeclampsia, holding the promise of prevention of preeclampsia in advance. Abnormal cell-free messenger RNA, microRNA and long noncoding RNA molecular changes help to elucidate the pathogenic determinants of preeclampsia, and open new therapeutic windows to effectively reduce pregnancy complications and fetal morbidity.

Gestational diabetes mellitus (GDM) is a common complication of pregnancy, which has significant adverse effects on both the mother and fetus. The incidence of GDM is increasing globally, and early diagnosis is critical for timely treatment and reducing the risk of poor pregnancy outcomes. GDM is usually diagnosed and detected after 24 weeks of gestation, while complications due to GDM can occur much earlier. Copy number variations (CNVs) can be a possible biomarker for GDM diagnosis and screening in the early gestation stage. In this study, we proposed a machine-learning method to screen GDM in the early stage of gestation using cell-free DNA (cfDNA) sequencing data from maternal plasma. Five thousand and eighty-five patients from north regions of Mainland China, including 1942 GDM, were recruited. A non-overlapping sliding window method was applied for CNV coverage screening on low-coverage (~0.2×) sequencing data. The CNV coverage was fed to a convolutional neural network with attention architecture for the binary classification. The model achieved a classification accuracy of 88.14%, precision of 84.07%, recall of 93.04%, F1-score of 88.33% and AUC of 96.49%. The model identified 2190 genes associated with GDM, including DEFA1, DEFA3 and DEFB1. The enriched gene ontology (GO) terms and KEGG pathways showed that many identified genes are associated with diabetes-related pathways. Our study demonstrates the feasibility of using cfDNA sequencing data and machine-learning methods for early diagnosis of GDM, which may aid in early intervention and prevention of adverse pregnancy outcomes.

Objectives: Genetic screening can benefit early detection and intervention for hearing loss. The frequency of common deafness-associated variants in general populations is highly important for genetic screening and genetic counseling tailored to different ethnic backgrounds. We aimed to analyze the frequency of common deafness-associated variants in a large population-based Chinese newborn cohort and to explore the population-specific features in diverse populations worldwide. Design: This population-based cohort study analyzed the frequency of common deafness-associated variants in 3,555,336 newborns in the Chinese Newborn Concurrent Hearing and Genetic Screening cohort. Participants were newborn infants born between January 2007 and September 2020. Limited genetic screening for 20 variants in 4 common deafness-associated genes and newborn hearing screening were offered concurrently to all newborns in the Chinese Newborn Concurrent Hearing and Genetic Screening cohort. Sequence information of 141,456 individuals was also analyzed from seven ethnic populations from the Genome Aggregation Database for 20 common deafness-related variants. Statistical analysis was performed using R. Results: A total of 3,555,326 Chinese neonates completed the Newborn Concurrent Hearing and Genetic Screening were included for analysis. We reported the distinct landscape of common deafness-associated variants in this large population-based cohort. We found that the carrier frequencies of GJB2, SLC26A4, GJB3, and MT-RNR were 2.53%, 2.05%, 0.37%, and 0.25%, respectively. Furthermore, GJB2 c.235delC was the most common variant with an allele frequency of 0.99% in the Chinese newborn population. We also demonstrated nine East-Asia-enriched variants, one Ashkenazi Jewish-enriched variant, and one European/American-enriched variant for hearing loss. Conclusions: We showed the distinct landscape of common deafness-associated variants in the Chinese newborn population and provided insights into population-specific features in diverse populations. These data can serve as a powerful resource for otolaryngologists and clinical geneticists to inform population-adjusted genetic screening programs for hearing loss.