mixOmics workflow

Once you’ve selected a mixOmics method, follow this general workflow:

1. Pre-process your data – Handle missing values or repeated measurements (see below).
2. Build an initial model – Apply the method to your data.
3. Tune model parameters – Optimise settings for better performance.
4. Train the final model – Use the best parameters for an optimised model.
5. Assess performance – Evaluate how well your model works.
6. Predict new samples – If applicable, use the model to predict outcomes.

Not all steps are necessary. For example, you can manually select parameters instead of tuning (Step 3), or if using an unsupervised method, prediction (Step 6) may not apply.

👉 For detailed examples of how to apply your specific method, refer to the case studies and the vignette.

Pre-processing

In general, data should be pre-processed before running mixOmics. This includes normalisation and prefiltering (see the vignette for more details). However, pre-processing data to handle missing values and repeated measures can be carried out using mixOmics functions. Pre-processing of microbiome data can also be done using mixOmics mixMC functions.