This case study shows how to apply multilevel sPLS to explore relationships between gene expression and clinical variables in liver toxicity data, using a simulated repeated measures design. By accounting for subject-specific effects, the multilevel framework enables the identification of gene–clinical associations while reducing within-subject variability. You’ll learn how to construct a multilevel sPLS model using pre-tuned parameters, and interpret sample separation, variable relationships, and feature clustering across components.
🔍 More on sPLS
🔍 More on multilevel methods
📄 Download R script
Data used on this page:vac18
Key functions used on this page:spls()plotIndiv()plotVar()cim()
Note:
This case study uses pre-tuned sPLS-DA parameters (optimal.ncomp and optimal.keepX), which were calculated externally and are available for download here.
References:
1. Bushel, P., Heinloth, A., Li, J., Huang, L., Chou, J., & Boorman, G. et al. (2007). Blood gene expression signatures predict exposure levels. Proceedings Of The National Academy Of Sciences, 104(46), 18211-18216. https://doi.org/10.1073/pnas.0706987104