Machine Learning based Pan-cancer Diagnosis Prediction and Treatment Recommendation using DNA Methylation Markers
Wei Chen, Jiajie Zhang, Jiahao Ye, Huijun Hao, Guodong He · 2023
Cancer is currently the leading cause of death around the world, with about 19 million new cases and nearly 10 million deaths every year. DNA methylation markers can accurately determine the tissue source of tumors, which is of great significance for selecting the correct tumor treatment and improving the survival expectation of patients. In this study, we constructed machine learning models to predict the potential of DNA methylation markers for Pan-cancer diagnosis, survival prediction and treatment recommendation. Our results show that the model which is based on multi-layer perceptron could classify 33 cancer types and normal tissue using 114 markers with an overall accuracy of 90.7%, while its precision is 88.7%. Based on DNA methylation data, we constructed an effective survival prediction model using survival information (P < 0.001). By combining DNA methylation data with clinical information in a machine learning parseable format, we were able to predict the treatment recommendation with 98.7% accuracy. Due to this result was 22.7% more accurate than the accuracy without DNA methylation data, it suggests that DNA methylation plays a significant role in the treatment commendation of cancers.