Machine Learning Application in cfDNA Analysis to Achieve Tumour Assessment

Abstract

Breast cancer (BC) is the leading cause of cancer in women and the second leading cause of cancer-related death. Early and accurate screening of BC is a promising way of reducing the proportion of patients with advanced stages of BC. In recent years, the non-invasive test of tumour diagnosis by assessing the level of Plasma cell-free DNA (cfDNA) has become a research hotspot. Here, we demonstrate the use of random forest models to predict BC by evaluating the levels of 26 known breast cancer-related cfDNA methylation molecular markers (model-tested accuracy of 67.88%). Then, we improved the accuracy of the model to 71.52% by parameter optimization. In addition, considering that the diagnosis of BC is closely related to the health of every female, we have extended the project from scientific research to social investigation by carrying out a sample survey of Chinese college students to understand various perspectives on the application of artificial intelligence in the diagnosis of diseases. We found that the response was rather optimistic, while some participants showed concerns about the maturity of the technology and the disclosure of privacy. Therefore, future research should focus on the optimisation of the machine learning model, so as to effectively improve the accuracy of diagnosis and provide better pre-service for the population at risk of cancer.