Integrative graph deep learning with directional decoding for robust gene regulatory network inference from single-cell transcriptomics
Journal of Zhejiang University-SCIENCE B · 2026-10 · 27 (10) · 1129-1139
DOI: 10.1631/jzus.b2500424 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Zhejiang University-SCIENCE B, online erschienen am 22.09.2026. Verfasst von Binhua Tang und 4 weiteren Autor:innen. Beteiligte Einrichtung: Key Laboratory of Maritime Intelligent Cyberspace Technology (Ministry of Education of China), Hohai University, Nanjing 213200, China. bh.tang@hhu.e…. Hauptthemen laut MEDLINE: Gene Regulatory Networks, Deep Learning, Single-Cell Analysis, Transcriptome. Die Arbeit nennt 20 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: the StarGuard Research Project, Guangzhou BonHeart Foundation, China.
Abstract (Originalsprache)
While gene regulatory networks (GRNs) are fundamental to understanding complex cellular mechanisms, the accurate inference from single-cell RNA-sequencing (scRNA-seq) data is hindered by the inherent noise, transcriptional dropout events, and dynamic nature of gene regulation. Current computational approaches often overlook long-range regulatory interactions, lack explicit modeling of regulatory directionality, and suffer from limited generalizability. To address these limitations, we introduce scGRAIL, a supervised deep learning framework that combines inductive local-aware graph aggregation (LGA) with an asymmetric relation decoder to enable robust GRN inference. Within this framework, LGA performs multi-hop neighborhood aggregation on a transcription factor (TF)-target graph, jointly leveraging topological structure and gene expression to capture long-range regulatory dependencies. A node reconstruction autoencoder (NAE) imposes gene expression reconstruction constraints on the latent embeddings, thereby enhancing robustness to noise and data heterogeneity and improving generalization. An asymmetric difference mechanism is introduced during decoding for distinct modeling of TFs and target genes; by computing an ordered difference vector between their embeddings, the model explicitly captures the directionality and target specificity of TF-target regulatory interactions, facilitating more accurate identification of potential causal regulatory relationships. Comprehensive evaluations demonstrate the model's superiority: it achieves optimal performance on 86.4% of benchmark datasets (38/44), surpassing state-of-the-art methods. Notably, scGRAIL achieves a 4.25% increase in average area under the receiver operating characteristic curve (AUROC) across four real-world networks compared to the graph contrastive link prediction (GCLink) model, yielding a 17.64% AUROC improvement for the TF500+ Non-Specific network in the mouse hematopoietic stem cells with lymphoid-lineage (mHSC-L) dataset, demonstrating its resilience to noise and scalability for large-scale GRNs. These results establish scGRAIL as a transformative tool for uncovering gene regulatory logic in single-cell transcriptomics. Overall, scGRAIL enhances structural modeling and directional discrimination in single-cell GRN inference by using multi-hop graph aggregation to capture long-range regulatory dependencies, expression reconstruction constraints to improve cross-dataset robustness, and an asymmetric decoding mechanism to explicitly model the regulatory directionality between TFs and target genes. 基因调控网络(GRNs)是揭示细胞内复杂调控机制的重要工具,但基于单细胞RNA测序(scRNA-seq)数据的GRN推断常受噪声、转录本“dropout”以及调控关系动态性等因素影响而难以准确建模。现有方法往往忽略长程调控依赖,缺乏对调控方向性的显式刻画,并存在泛化能力不足等问题。为此,本文提出一种有监督的深度学习框架scGRAIL,将归纳式局部感知图聚合(LGA)与非对称关系解码器相结合,实现鲁棒的GRN推断。在该框架中,LGA在转录因子–靶标(TF–target)图上进行多跳邻域聚合,联合利用网络拓扑结构与基因表达信息以捕获长程调控依赖;同时引入节点重构自编码器(NAE)对潜在表示施加表达重构约束,从而增强对噪声和数据异质性的鲁棒性,并提升跨数据集泛化能力。解码阶段设计非对称差分机制,通过计算TF与Target嵌入的有序差分向量,对TF–target调控关系的方向性与靶基因特异性进行显式建模,从而更准确识别潜在的因果调控关系。综合实验结果表明,scGRAIL在44个基准数据集中的38个(86.4%)取得最优性能,整体优于现有先进方法;在四个真实网络上,相较GCLink的平均受试者工作特征曲线下面积(AUROC)提升4.25%,其中在Non-Specific网络,mHSC-L中TF500+数据集上AUROC提升17.64%,体现出对噪声的强鲁棒性及大规模网络的可扩展性。综上,scGRAIL通过多跳图聚合捕获长程依赖,以表达重构增强跨数据集鲁棒性,并利用非对称方向性解码提升调控判别能力,为单细胞转录组中的基因调控逻辑解析提供了有效工具。.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Substanzen
Autor:innen und Institutionen
- Key Laboratory of Maritime Intelligent Cyberspace Technology (Ministry of Education of China), Hohai University, Nanjing 213200, China. bh.tang@hhu.edu.cn.
- Shanghai Key Laboratory of Data Science, Fudan University, Shanghai 200438, China. bh.tang@hhu.edu.cn.
- Key Laboratory of Maritime Intelligent Cyberspace Technology (Ministry of Education of China), Hohai University, Nanjing 213200, China.
Förderung
- the StarGuard Research Project, Guangzhou BonHeart Foundation, China
Zitieren
APA
Tang, B., Feng, Y., Gao, X., Mao, M., & Zhang, Y. (2026). Integrative graph deep learning with directional decoding for robust gene regulatory network inference from single-cell transcriptomics. Journal of Zhejiang University-SCIENCE B, 27(10), 1129-1139. https://doi.org/10.1631/jzus.b2500424
Vancouver
Tang B, Feng Y, Gao X, Mao M, Zhang Y. Integrative graph deep learning with directional decoding for robust gene regulatory network inference from single-cell transcriptomics. Journal of Zhejiang University-SCIENCE B. 2026;27(10):1129-1139. doi:10.1631/jzus.b2500424
BibTeX
@article{Tang2026,
author = {Tang, Binhua and Feng, Yingying and Gao, Xinyu and Mao, Mengyao and Zhang, Yujia},
title = {Integrative graph deep learning with directional decoding for robust gene regulatory network inference from single-cell transcriptomics},
journal = {Journal of Zhejiang University-SCIENCE B},
year = {2026},
volume = {27},
number = {10},
pages = {1129--1139},
doi = {10.1631/jzus.b2500424}
}
EUVIMED