Predicting Heart Cell Types by Using Transcriptome Profiles and a Machine Learning Method

Area of Science:

• Cardiovascular Biology
• Computational Biology
• Genomics

Background:

• The heart comprises diverse cell types, including cardiomyocytes, endothelial cells, and fibroblasts, whose interactions are vital for cardiac function.
• Understanding the unique gene expression patterns of these cardiac cell types is essential for comprehending heart physiology and disease.

Purpose of the Study:

• To apply machine learning techniques to single-cell RNA sequencing data for precise identification of 11 distinct heart cell types.
• To uncover key genes and long non-coding RNAs (lncRNAs) that serve as biomarkers for differentiating cardiac cell populations.

Main Methods:

• Utilized machine learning, including light gradient boosting machine and incremental feature selection, to analyze heart single-cell gene expression profiles.
• Developed and optimized classification models, specifically decision trees (DT) and random forests, to classify cardiac cell types based on gene expression data.

Main Results:

• Achieved high classification accuracy, with decision tree and random forest models yielding weighted F1 scores of 0.957 and 0.981, respectively.
• Identified critical genes (e.g., NPPA, LAMA2, DLC1) and lncRNAs (e.g., LINC02019, NEAT1) crucial for distinguishing between different cardiac cell types.
• Enrichment analysis confirmed the role of selected features in cardiac structure and function.

Conclusions:

• Machine learning effectively distinguishes cardiac cell types based on gene expression, providing a robust method for cell-type identification.
• The identified key genes and lncRNAs offer potential as molecular diagnostic markers for cardiac diseases.
• This study lays the groundwork for advancing molecular diagnostics and biomarker discovery in cardiology.