Machine Learning Model for T-cell Identification from Flow Cytometry Panel of Melanoma Patients

Flow cytometry (FC) is an essential tool in clinical and research laboratories for quantifying and identifying cells in biological samples, particularly in the early diagnosis of hematological malignancies. However, traditional analysis methods are often time-consuming and subjective. Recent advancements have introduced the use of unsupervised machine learning (ML) models to classify immune cells, but research on supervised ML models for immune cell identification remains limited. This study investigates the use of ML models to identify T-cells in melanoma patients, employing key phenotypic biomarkers such as TCRγδ, CD25, CD127, CD95, CD45RA, CCR7, and CD56. The XGBoost and Neural Network models achieved accuracies of 80.9% and 84% during a train-test split, while 5-fold cross-validation improved these to 86.82% and 88.30%, respectively. Our findings suggest that supervised ML can effectively characterize T-cell subtypes in melanoma patients, reducing the time needed for FC data analysis and driving biomedical research in identifying connections between biomarkers. The dataset and code are available on GitHub: https://github.com/Muraam-Abdel-Ghani/FC-TCell-Classification/.