The difference between supervised and unsupervised analyses in omic studies lies in the incorporation of experimental design information. Unlike unsupervised methods, supervised analyses incorporate prior information about the experimental design, making them useful for testing the effects of experimental factors and associating omic data with phenotypic features. Supervised analyses can be divided into two types: regression and classification.
Regression problems involve predicting a numeric variable or matrix based on the omic data and experimental factors, such as treatment or subject characteristics.
Classification problems involve classifying observations into groups based on their features across different omic layers.