METHOD FOR MINIMALLY INVASIVE DIAGNOSIS OF LUNG CANCER BY FRAGMENTED CIRCULATING FREE DNA BASED ON MACHINE LEARNING METHODS Russian patent published in 2024 - IPC A61B5/00 G01N33/48 

FIELD: medicine.

SUBSTANCE: invention refers to oncology, biomedical statistics and diagnostics of malignant neoplasms (MN). Invention includes a method for minimally invasive diagnosis of lung cancer by the circulating free DNA (cfDNA) fragmentation profile and the cfDNA fragment end motif sequence profile. Method is carried out by isolating cfDNA from blood plasma, preparing DNA libraries and sequencing them, as well as by using a unique algorithm for bioinformatics and statistical processing of sequencing data to create a binary classification model. Model enables to distinguish between healthy subjects (without diagnosed lung cancer, LC) and cancer patients (subjects with established LC) with AUC value from 0.87 to 0.875 with values of sensitivity of 0.81 and specificity of 0.79 to 0.9. CfDNA fragmentation profile contains records on the length of cfDNA fragments and the ratio of their lengths. Profile of sequences of end motifs of cfDNA fragments contains records on linear combinations of values of position-weight matrices for a genome section, according to which DNA fragmentation occurs most often inside this section. Classification model was created based on machine learning methods using L1, L2-regularized logistic regression, probabilistic classifiers based on Gaussian processes, a principal component method and nuclear logistic regression. Present method can be used as a basis for developing a diagnostic test which can be used to assist in making medical decisions.

EFFECT: model has good diagnostic value and does not require clinical data and data on other tumour-associated biomarkers.

1 cl, 15 dwg, 6 tbl