Fig. 1
From: Shambhala: a platform-agnostic data harmonizer for gene expression data
Schematic representation of Shambhala pipeline for harmonization of gene expression data. Various profiles from samples (1… N) obtained at different platforms are taken one-by-one, merged with an auxiliary calibration dataset P0 and then quantile-normalized with it. This produces the dataset P, which is then transformed into the shape of the definitive dataset Q; during transformation, only the dataset P changes, while Q remains constant. The result of such a conversion, dataset P1, contains the transformed profile for sample i, which is considered harmonized. Profiles from all other samples (1,…,N) are harmonized one-by-one using the same algorithm