SOBA (feature Selection by Orthogonal Burden Association) is a framework for selecting biologically plausible features for machine learning models in liquid biopsy applications. In this study, Hamilton, Almanza, Wang, et al. applied the framework to early lung cancer detection using cell-free DNA (cfDNA) methylation sequencing. Using SOBA features, the authors developed the lung methylation-based cancer likelihood in plasma (Lung-mCLiP) model to non-invasively predict lung cancer probability. Code available through this website extracts features used for model training from a bam file.