TensorPLS: an R package for exploring molecular and temporal patterns in longitudinal omics data using PLS-DA

TensorPLS, applied to longitudinal transcriptomics tensors, achieved Q2Y = 0.670 with AUROC 0.971 and balanced accuracy 0.931 in GSE68310 (109 subjects) using Tucker3 reconstruction.

• Why it matters: Longitudinal omics has repeated high-dimensional measurements with missingness, making supervised modelling and temporal interpretation difficult without leakage-aware validation.
• What they did: The R package represents data as subjects × features × time tensors, imputes missing values via Tucker3 reconstruction or feature-time mean imputation, then unfolds along subjects for matrix PLS-DA with nested cross-validation and VIP-based feature selection.
• The result: In GSE182616 (66 subjects), feature-time mean imputation produced Q2Y = 0.294, AUROC 0.827, and balanced accuracy 0.756, enabling predictive assessment plus temporal variable-importance mapping and functional enrichment.

AI summary of the abstract · may contain errors · not medical advice · read the paper.