Single_cell_GRN: gene regulatory network identification based on supervised learning method and Single-cell RNA-seq...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) offers high resolution for studying gene expression heterogeneity.
• Inferring gene regulatory networks (GRNs) is crucial for understanding cellular mechanisms.
• Existing computational methods for GRN inference from scRNA-seq data require rigorous evaluation.

Purpose of the Study:

• To evaluate the performance of seven common machine learning classifiers for GRN inference using scRNA-seq data.
• To compare supervised learning methods against unsupervised methods for this task.
• To identify the most effective classifiers and specific configurations for scRNA-seq based GRN inference.

Main Methods:

• Investigated seven classifiers: Support Vector Machine (SVM), Random Forest (RF), Naive Bayesian (NB), Gradient Boosting Decision Tree (GBDT), Logistic Regression (LR), Decision Tree (DT), and K-Nearest Neighbors (KNN).
• Utilized three kernel functions (linear, polynomial, radial basis function) for SVM.
• Applied three real-world scRNA-seq datasets from mouse and human tissues.

Main Results:

• Supervised learning methods generally outperformed the unsupervised method GENIE3 in terms of Area Under the Curve (AUC).
• SVM, RF, and KNN demonstrated superior performance compared to NB, GBDT, LR, and DT.
• For SVM, linear and polynomial kernels were found to be more suitable for modeling scRNA-seq data than the radial basis function kernel.

Conclusions:

• Supervised machine learning approaches are effective for GRN inference from scRNA-seq data.
• SVM, RF, and KNN are recommended as robust classifiers for this application.
• Specific SVM kernel choices significantly impact the accuracy of GRN inference from scRNA-seq data.