Reject and Cascade Classifier with Subgroup Discovery for Interpretable Metagenomic Signatures

Overcoming cohort heterogeneity for the prediction of subclinical cardiovascular disease risk

“…For instance, clustering analysis has been used to determine phenotypic heterogeneity in patients with unique characteristics to better inform patient care for a broad range of diseases, 13 , 14 , 15 and data mining techniques were used to discover homogeneous subgroups within a disease population to make more precise predictions. 16 , 17 , 18 Given the complexity of CVD, including interactions with many comorbidities and predispositions, it is not surprising that the pool of individuals that develop CVD are diverse with differing risk factors and characteristics. This diversity, which can be framed as cohort heterogeneity, is important to elucidate as it has large implications on downstream metabolomic analyses.…”

“… 26 , 27 Data-driven approaches for uncovering cohort heterogeneity can be divided into unsupervised techniques which ignore the outcome being studied, and supervised techniques, which include information about the outcome. Unsupervised techniques include clustering 10 , 28 , 29 and latent class analysis 30 , 31 of clinical phenotypes; supervised techniques include mixture of experts (MoE) 32 , 33 , 34 , 35 and subgroup discovery algorithms 17 , 36 to subdivide large cohorts. Following stratification of individuals and identification of heterogeneity present, techniques including hierarchical modeling of subcohorts and ensemble learning can then be employed to improve the prediction of CAD in the whole cohort.…”