Predictive Modeling of Lung Cancer Detection Using Logistic Regression on Liquid Biopsy

Lung cancer remains one of the most important issues related to cancer mortality worldwide due to the late diagnosis and the need for invasive biopsy procedures. There has emerged a new and less cumbersome, painless method of testing for lung cancer called liquid biopsy, which relies on the assessment of the tumor demographic areas-circulating tumor Deoxyribonucleic Acid (ctDNA) of peripheral blood tissues and has shown a great potential for the early detection and follow up of lung cancer. This current work examines the extent to which logistic regression can be used as a predictive analytical tool aimed at the evaluation of liquid biopsy studies so that lung cancer detection can be more accurate and efficient. Using logistic regression a much celebrated machine learning approach owing to its ease of understanding and ability to withstand vagaries of practice, seek to analyze molecular data from liquid biopsy samples looking to segregate malignant and benign samples. Model training will involve incorporating datasets containing ctDNA mutations and other cancer related biomarkers within lung cancer's scope. Important features like mutation and variant allele frequencies and methylation profiles will be used as input parameters in constructing the predictive model. This work intends to focus on the logistic regression model refinement due to its sensitivity and specificity which concepts are very crucial in the early stages of lung cancer detection. In this way, a simple, affordable and efficient tool that is based on the readily understandable logistic regression to enable the diagnosis of patients in real time in clinical settings.