Bacterial signatures for diagnosis of colorectal cancer by machine learning

Abstract

Invasive methods such as colonoscopy are more commonly used in colorectal cancer (CRC) screening and diagnosis, but these methods are not easily accepted and have limitations. In this paper, we aim to exploit the close relationship between intestinal flora and the development of CRC. A T-test was used to screen and compare the intestinal flora of healthy individuals and patients, and strains with significant differences were selected as characteristic ones. In addition, three AI learning models, Random forest (RF), K-Nearest Neighbor (KNN), and Back propagation neural network (BPNN), were used to build a colorectal cancer diagnosis model based on intestinal flora. Overall, the investigation carried out by us has revealed six highly divergent species between healthy individuals and patients from t-tests and key species associated with CRC. The results were validated against each other, confirming the reliability of the obtained key strains, and providing a new idea for the clinical diagnosis of CRC.