Multi-model biological and sequence information fusion for gene regulatory network inference from single-cell transcriptomics
Identification of transcription factor target gene interactions and construction of the gene regulatory networks (GRNs) are essential for understanding the molecular mechanisms underlying transcriptional gene regulation. Large scale single cell transcriptomics across different tissues offers unprecedented resolution of cellular diversity and regulatory dynamics by capturing gene expression…
The article discusses the development of scMGFGRN, a multi-model deep learning framework designed to fuse single-cell transcriptomic profiles with Gene Ontology hierarchical relationships, gene sequences, and DNA language models to infer gene regulatory networks (GRNs). This innovative approach addresses the limitations of existing methods by effectively integrating multimodal data and capturing hierarchical structures in biological knowledge.
By leveraging denoising auto encoders, graph attention feature extraction, and a pertained DNA language model, scMGFGRN identifies informative regulatory signatures and integrates complementary features from different sources to accurately predict GRNs. Benchmarked against seven datasets of human and mouse cells, scMGFGRN surpasses state-of-the-art methods in GRN identification and reconstruction, revealing novel transcription factor-gene interaction pairs and cell-type specific GRNs.
The interpretability analysis further demonstrates scMGFGRN's capability to integrate transcriptomic profiles with multi-model structure information, highlighting the contributions of heterogeneous biological sources towards improved GRN inference.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
