Shambhala‐2: A Protocol for Uniformly Shaped Harmonization of Gene Expression Profiles of Various Formats

Abstract

Uniformly shaped harmonization of gene expression profiles is central for the simultaneous comparison of multiple gene expression datasets. It is expected to operate with the gene expression data obtained using various experimental methods and equipment, and to return harmonized profiles in a uniform shape. Such uniformly shaped expression profiles from different initial datasets can be further compared directly. However, current harmonization techniques have strong limitations that prevent their broad use for bioinformatic applications. They can either operate with only up to two datasets/platforms or return data in a dynamic format that will be different for every comparison under analysis. This also does not allow for adding new data to the previously harmonized dataset(s), which complicates the analysis and increases calculation costs. We propose here a new method termed Shambhala‐2 that can transform multi‐platform expression data into a universal format that is identical for all harmonizations made using this technique. Shambhala‐2 is based on sample‐by‐sample cubic conversion of the initial expression dataset into a preselected shape of the reference definitive dataset. Using 8390 samples of 12 healthy human tissue types and 4086 samples of colorectal, kidney, and lung cancer tissues, we verified Shambhala‐2's capacity in restoring tissue‐specific expression patterns for seven microarray and three RNA sequencing platforms. Shambhala‐2 performed well for all tested combinations of RNAseq and microarray profiles, and retained gene‐expression ranks, as evidenced by high correlations between different single‐ or aggregated gene expression metrics in pre‐ and post‐Shambhalized samples, including preserving cancer‐specific gene expression and pathway activation features. © 2022 Wiley Periodicals LLC. Basic Protocol: Shambhala‐2 harmonizer Alternate Protocol 1: Linear Shambhala/Shambhala‐1 Alternate Protocol 2: Alternative (flexible‐format and uniformly shaped) normalization methods Support Protocol 1: Watermelon multisection (WM) Support Protocol 2: Calculation of cancer‐to‐normal log‐fold‐change (LFC) and pathway activation level (PAL)

Transcriptomic Harmonization as the Way for Suppressing Cross-Platform Bias and Batch Effect

“…The MAQC and SEQC projects investigated the correlations between the transcriptome profiles of the same biological type, yet obtained using the different experimental platforms. Although these correlations were high [ 17 , 18 , 19 , 27 ], without the some special cross-platform normalization methods (quantile normalization [ 28 ] was not enough), the overall collections of profiles were grouped according to the experimental platforms, rather than to the biological type of samples, in terms of both clustering dendrograms and of principal component analysis (PCA) [ 29 , 30 , 31 , 32 , 33 , 34 ].…”

“…More recently, we proposed a new family of uniformly shaped cross-platform harmonizers termed Shambhala [ 32 , 33 ]. Harmonization here is performed not simultaneously for all the profiles under harmonization, but for the gene expression profiles taken one by one, when each individual profile is merged and quantile-normalized [ 28 ] with an auxiliary calibration dataset that is pre-defined by the method developers.…”

“…Furthermore, such harmonization may use different mathematical transforms as the engine to reshape the transcriptional profiles. The first version of Shambhala used the piecewise linear method XPN [ 29 , 31 ] for profile reshaping [ 32 ], whereas the latest version [ 33 ] utilized the piecewise cubic transformation method CuBlock [ 34 ].…”

“…The following quantitative metrics and methods may be applied to estimate the effect of harmonization: First, different statistical criteria may be used to estimate the following endpoints: Correlation analysis for the gene expression profiles before and after harmonization [ 29 , 30 , 31 , 33 , 34 ]; Comparison of between- and within-class distances before and after harmonization [ 29 ]; Alternatively, one may classify the samples according to gene expression data after normalization, involving various machine learning (ML) methods: Logistic regression [ 92 ], used in [ 30 ]; SVM [ 93 ], used in [ 29 , 31 ]; Nearest shrunken centroids Prediction Analysis for Microarrays (PAM) [ 94 ], used in [ 29 ]. …”

“…First, different statistical criteria may be used to estimate the following endpoints: Correlation analysis for the gene expression profiles before and after harmonization [ 29 , 30 , 31 , 33 , 34 ]; Comparison of between- and within-class distances before and after harmonization [ 29 ]; …”