Combining least absolute shrinkage and selection operator (LASSO) and heat map visualization for biomarkers detection of LGL leukemia

Large granular lymphocyte (LGL) leukemia is type of chronic leukemia disease that exhibits an unexplained elevation in large granular lymphocytes in the peripheral blood. In order to better define the LGL leukemia pathogenesis, we propose an approach that combines the least absolute shrinkage and selection operator (LASSO) and heat map visualization to detect possible biomarkers from a statistical perspective. In the research, we firstly classify our sample data into two classes: gene information from blood samples of normal donors and that of infected patients. We then apply the LASSO on the sample data for feature selection and dimension reduction to identify potential biomarkers. Based on our result, heat map visualization has revealed little information about possible biomarkers from the gene data, while the LASSO method shows promise in identifying biomarkers from over 10,000 candidate genes. Our final result shows that the method has successfully reduced the dimension of attributes to as low as 0.1% and has selected 21 features as potential biomarkers. Further analysis in the field of biomedical sciences is required to validate the relations between the potential biomarkers found and the LGL Leukemia disease.