An Enviro-Geno-Pheno State Analysis Framework For Biomarker Study

In this study, we introduce the basic perspectives of E-GPS framework for biomarker study and demonstrate its advantages by a series of experiments on real bio-data in cancer research. Owing to the development of high-throughput technology on gene detection, an increasing number of disease-related biomarkers for diagnosis, subtyping, and prognosis have been discovered. Conventional methods for biomarker discovery mainly focus on detecting the relationship between intrinsic genotypes and extrinsic phenotypes, while recent efforts emphasize the role of a biomarker under particular condition, which extends to the integrated analysis on Enviro-, Geno-, and Pheno-measures. Samples sharing a similar pattern of genotype under certain environment are in a 'state' that is distinctive from other states. Such different states generated by machine learning from samples can further indicate the different outcome of the phenotype of interest, which makes the enviro-geno-pheno state (E-GPS) analysis as an analytic tool for biomarker studies.