Multiblock variable influence on orthogonal projections (MB-VIOP) for enhanced interpretation of total, global, local and unique variations in OnPLS models
share
A method for variable selection in multiblock analysis, called multiblock variable influence on orthogonal projections (i.e., Multiblock-VIOP or MB-VIOP) is explained in this paper. Multiblock-VIOP is a model based variable selection method that uses the data matrices, the scores and the normalized loadings of an OnPLS model in order to sort the input variables of a large number of data matrices according to their importance for both simplification and interpretation of the total OnPLS model, and also of the unique, local and global model components separately. The previously published OnPLS algorithm finds the relationships among multiple data matrices (blocks) by calculating latent variables; however, a method for improving the interpretation of these latent variables by assessing the importance of the input variables was missing. In this paper, we provide evidence for the usefulness, the efficiency and the reliability of MB-VIOP for the abovementioned purposes by means of three examples, (i) a synthetic four-block dataset, (ii) a real three-block omics dataset related to plant sciences, and (iii) a real six-block dataset related to the food industry.
READ FULL TEXT
