Most diseases, such as breast cancer and lung cancer, are not just one disease. They are a kind of disease. For one type of disease, the pathogenesis, molecular expression profile, or histopathology of each subtype may have a high degree of heterogeneity. Accordingly, patients with the same disease may experience huge differences in disease progression, clinical efficacy, prognosis, etc. The classification of disease subtypes based on molecular characteristics is conducive to more accurate and effective treatment for patients. In confidence analysis, unsupervised machine learning is usually used to cluster samples according to the molecular expression of patients.
Annotation: As shown in the figure, the samples of tumor patients can be divided into three subtypes through unsupervised clustering, and the three subtypes are obviously different in the expression profile of protein molecules. Through survival analysis, it was found that the survival probability of the three subtypes of patients was also significantly different. Furthermore, we can find that the expression of some proteins in the three subtypes is significantly different, which is expected to become a potential biomarker for predicting the disease subtypes.