Modeling omics integration with HIVE identifies response signatures to multifactorial stress in plants.

CALIA, G.; MARGUERIT, S.; MOTA, A. P. Z.; VIDAL, M.; SCHULER, H.; BRASILEIRO, A. C. M.; GUIMARAES, P. M.; BOTTINI, S.

Modeling omics integration with HIVE identifies response signatures to multifactorial stress in plants.

Autores

Descripción

To address these challenges, we developed HIVE (Horizontal Integration analysis using Variational AutoEncoders), a method to jointly analyze multiple transcriptomics data from different experiments (i.e. unpaired). The current implementation is based on the use of a VAE, a generative model that learns low-dimensional representations of the observed data, using a variational Bayes methodology in an unsupervised framework. By coupling a random forest regression model and the SHAP explainer, HIVE selects relevant genes for the studied phenotype. The application of a nested stratified cross-validation technique allowed us not only to treat datasets with unbalanced classes but also to overcome the small sample size challenges (Fig. 1, Supplementary Note S1, Supplementary Figs. S1 to S5, and Supplementary Tables S1 to S3).