Fig. 1
From: Using LDpred2 to adapt polygenic risk score techniques for methylation score creation
Left: Schematic of a probe-probe block correlation matrix. Grey areas represent blocks around the diagonal, within which correlations are calculated. Most fields are empty (white). Right: When including singletons in the CMR blocks, only a partial block structure is created, which gives the algorithm too little correlation information to work with, since the singleton clusters do not contribute anything to the model