Placenta · 2026
Evaluation of a deep generative computational framework under constrained conditions for multi-regional placental single-cell transcriptomic inference.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Tang Zhenjie, Li Kun, Liu Yuming, Lu Shijing, Wei Xingyu, Yang Yuxuan, Yin Hongtao, Guo Zhiwei, Sheng Chao, Li Fenxia, Yang Xuexi
Аннотация
INTRODUCTION: The placenta is a highly heterogeneous organ, and its dysfunction is central to preeclampsia (PE). Although single-cell RNA sequencing (scRNA-seq) enables high-resolution characterization, its high cost limits large-scale clinical use. scSemiProfiler integrates deep generative modeling with bulk RNA-seq to infer single-cell transcriptomes, but the original framework relies on an active learning module to select informative samples. In this study, we evaluate a simplified workflow that omits active learning, assessing its performance under these constrained conditions in complex, multi-regional placental tissues.
METHODS: We generated 14 scRNA-seq libraries from three placental regions-basal plate (BP), placental villi (PV), and chorioamniotic membranes (CAM)-obtained from 5 pregnant women (1 normotensive controls and 4 PE cases), and integrated these with 118 bulk RNA-seq datasets (detailed in Table S1). scSemiProfiler was applied without active learning, using existing scRNA-seq data as references. Semi-profiled and real-profiled data were compared in cell type composition, marker genes, and Gene Ontology (GO) enrichment. The impact of gene-count filtering was also assessed.
RESULTS: Semi-profiled data recapitulated major tissue-specific cell composition patterns across regions, despite deviations in some populations (especially in some rare populations). Gene expression intensity and detection rates were reduced, but canonical markers remained identifiable. Filtering cells with fewer than 10 detected genes improved compositional concordance and marker visualization, with minimal impact on GO consistency. scSemiProfiler captured pathway-level functional variation and disease-associated GO changes, but was less reliable for rare cell types and subtle gene-level differences, particularly in case-control contrasts across regions.
CONCLUSIONS: The simplified scSemiProfiler workflow, applied without active learning, largely preserved broad tissue-level transcriptional patterns but showed limited performance for rare cell types and disease contrasts.
METHODS: We generated 14 scRNA-seq libraries from three placental regions-basal plate (BP), placental villi (PV), and chorioamniotic membranes (CAM)-obtained from 5 pregnant women (1 normotensive controls and 4 PE cases), and integrated these with 118 bulk RNA-seq datasets (detailed in Table S1). scSemiProfiler was applied without active learning, using existing scRNA-seq data as references. Semi-profiled and real-profiled data were compared in cell type composition, marker genes, and Gene Ontology (GO) enrichment. The impact of gene-count filtering was also assessed.
RESULTS: Semi-profiled data recapitulated major tissue-specific cell composition patterns across regions, despite deviations in some populations (especially in some rare populations). Gene expression intensity and detection rates were reduced, but canonical markers remained identifiable. Filtering cells with fewer than 10 detected genes improved compositional concordance and marker visualization, with minimal impact on GO consistency. scSemiProfiler captured pathway-level functional variation and disease-associated GO changes, but was less reliable for rare cell types and subtle gene-level differences, particularly in case-control contrasts across regions.
CONCLUSIONS: The simplified scSemiProfiler workflow, applied without active learning, largely preserved broad tissue-level transcriptional patterns but showed limited performance for rare cell types and disease contrasts.