Machine learning driven dashboard for chronic myeloid leukemia prediction using protein sequences

Abstract

The prevalence of Leukaemia, a malignant blood cancer that originates from hematopoietic progenitor cells, is increasing in Southeast Asia, with a worrisome fatality rate of 54%. Predicting outcomes in the early stages is vital for improving the chances of patient recovery. The aim of this research is to enhance early-stage prediction systems in a substantial manner. Using Machine Learning and Data Science, we exploit protein sequential data from commonly altered genes including BCL2, HSP90, PARP, and RB to make predictions for Chronic Myeloid Leukaemia (CML). The methodology we implement is based on the utilisation of reliable methods for extracting features, namely Di-peptide Composition (DPC), Amino Acid Composition (AAC), and Pseudo amino acid composition (Pse-AAC). We also take into consideration the identification and handling of outliers, as well as the validation of feature selection using the Pearson Correlation Coefficient (PCA). Data augmentation guarantees a comprehensive dataset for analysis. By utilising several Machine Learning models such as Support Vector Machine (SVM), XGBoost, Random Forest (RF), K Nearest Neighbour (KNN), Decision Tree (DT), and Logistic Regression (LR), we have achieved accuracy rates ranging from 66% to 94%. These classifiers are thoroughly evaluated utilising performance criteria such as accuracy, sensitivity, specificity, F1-score, and the confusion matrix.The solution we suggest is a user-friendly online application dashboard that can be used for early detection of CML. This tool has significant implications for practitioners and may be used in healthcare institutions and hospitals.