top of page

Machine learning methodology for cell type classification of single-cell RNA sequencing data

Donggeon Woo (BME)

It is known that mature chondrocytes undergo hypertrophy and ossification. Chondrocyte dedifferentiation induces fibrosis and loss of functions in cartilage tissues. This project aims to classify cell types using the human OA cartilage data (Fig. 7A). In particular, it attempts to classify three types of chondrocytes: general chondrocyte, hypertrophic chondrocyte and fibrocartilage chondrocyte (Fig. 7B). Multiple machine learning models are used, including LDA, linear SVM, RBF SVM, logistic regression and MLP. Hyperparameters are optimized using Optuna. RBF SVM achieves the most accurate classification result of 96.4%. The result demonstrates that ML models can capture key marker gene expression patterns.

Created by Taeyang Yang @ BCI LAB

Link to UNIST website
BME_logo.png
bottom of page