A Logistic Regression Model for Predicting Treatment Response to Elexacaftor/tezacaftor/ivacaftor in Cystic Fibrosis Patients with the F508Del Mutation

Abstract

Cystic Fibrosis (CF) is a genetic disease that affects the lungs and other organs, causing mucus and other fluids to become excessively thick. The F508Del mutation is the most common variant of CF, and the Elexacaftor/tezacaftor/ivacaftor (ETI) therapy is frequently used to treat this specific mutation. Despite this recent advancement, patient variability leads to differences in individual response to ETI, regardless of sharing the F508Del mutation. This study addresses the gap in CF machine learning models to predict treatment response by developing a logistic regression model to detect the responsiveness of a CF patient following the ETI Treatment. Specifically, the research will answer the question: to what extent can a Logistic Regression model accurately classify responsive and unresponsive cases of CF patients with the F508Del Mutation following ETI treatment? The study implements a two-part quantitative method, including a Differential Gene Expression (DGE) Analysis experimental approach and a Logistic Regression Model evaluation approach. The DGE Analysis found that LDLR, TNF, and PSMD5 were the differentially expressed genes (DEGs) across the CF genetic data. The model evaluation leveraged a confusion matrix, McNemar’s Test, and an ROC Curve. The model achieved an 85.71% accuracy, a 66.67% sensitivity, and a 100% specificity. The Area Under the ROC Curve (AUC) was 91.67%. The study concluded these evaluation factors as statistically significant (p-value = 0.00548). These findings suggest that machine learning can assist in personalized treatment prediction, and further validation with larger and more diverse cohorts is warranted to enhance generalizability.