Abstract:Inferring cell-type-specific gene regulatory networks (GRNs) from single-cell RNA sequencing (scRNA-seq) data is a complex task, primarily due to data sparsity, noise, and the dynamic, context-dependent nature of gene regulation across cell types and states. Recent advancements in the collection of experimentally validated data on transcription factor binding have facilitated GRN inference via supervised machine learning methods, where models learn from known TF-gene pairs to guide predictions. However, these methods still face challenges in 1) effectively representing and integrating prior knowledge, and 2) capturing regulatory mechanisms across diverse cellular contexts. To tackle the above challenges, we introduce a novel GRN inference method, scRegNet, that learns a joint representation from graph neural networks (GNNs) and pre-trained single-cell foundation models (scFMs). scRegNet combines rich contextual representations learned by large-scale, single-cell foundation models trained on extensive unlabeled scRNA-seq datasets with the structured knowledge embedded in experimentally validated networks through GNNs. This integration enables robust inference, the prediction of unknown gene regulatory interactions by simultaneously accounting for gene expression patterns and established gene regulatory networks. We evaluated our approach on seven single-cell scRNA-seq benchmark datasets from the BEELINE study, outperforming current state-of-the-art methods in cell-type-specific GRN inference. scRegNet demonstrates a superior ability to capture intricate regulatory interactions between genes across various cell types, providing a more in-depth understanding of cellular processes and regulatory dynamics. By harnessing the capabilities of large-scale pre-trained single-cell foundation models and GNNs, scRegNet offers a scalable and adaptable tool for advancing research in cell type-specific gene interactions and biological functions.

GenoHoption: Bridging Gene Network Graphs and Single-Cell Foundation Models

GenePT: A Simple But Effective Foundation Model for Genes and Cells Built From ChatGPT

scGPT: Towards Building a Foundation Model for Single-Cell Multi-omics Using Generative AI

DeepGene: An Efficient Foundation Model for Genomics based on Pan-genome Graph Transformer

scGPT: toward building a foundation model for single-cell multi-omics using generative AI

Scgen Predicts Single-Cell Perturbation Responses

Generative modeling and latent space arithmetics predict single-cell perturbation response across cell types, studies and species

GENet: A Graph-Based Model Leveraging Histone Marks and Transcription Factors for Enhanced Gene Expression Prediction

GeneCompass: Deciphering Universal Gene Regulatory Mechanisms with Knowledge-Informed Cross-Species Foundation Model

scLong: A Billion-Parameter Foundation Model for Capturing Long-Range Gene Context in Single-Cell Transcriptomics

Generating Multi-Modal and Multi-Attribute Single-Cell Counts with CFGen

ClOneHORT: Approaches for Improved Fidelity in Generative Models of Synthetic Genomes

GeneCompass: deciphering universal gene regulatory mechanisms with a knowledge-informed cross-species foundation model

scGNN+: Adapting ChatGPT for Seamless Tutorial and Code Optimization

Gene Regulatory Network Inference with Joint Representation from Graph Neural Network and Single-Cell Foundation Model

scGREAT: Transformer-Based Deep-Language Model for Gene Regulatory Network Inference from Single-Cell Transcriptomics

Learning Asynchronous Boolean Networks from Single-Cell Data Using Multiobjective Cooperative Genetic Programming

scHybridBERT: integrating gene regulation and cell graph for spatiotemporal dynamics in single-cell clustering

GEN Model: An Alternative Approach to Deep Neural Network Models

Cell-Graph Compass: Modeling Single Cells with Graph Structure Foundation Model

EpiAgent: Foundation model for single-cell epigenomic data