Figure 1.

A schematic flow of object-oriented regression, a) A omics data as a covariate matrix, b) organizing omics data via an unsupervised learning method, c) Clustered omics, resulted from two-way clustering analysis, leading to identification of exemplars, d) Computing similarity measurements with every exemplars and treating them as covariates, e) Dense covariate matrix of similarity measurements, useful for building a predictive model, f) Under the model, one can use penalized likelihood to select informative exemplars to build predictive models, g) One can perform validation analysis by assessing associations of predicted risk scores with clinical outcome, and h) Under a proposal hazard model for time-to-death, one can perform systemic association with individual exemplars