mvpaShiny - step by step

# mvpaShiny

This slide deck contains looped
GIFs that show how to work with mvpaShiny, using the
HOMA-IR study dataset. However, not all plots shown in the manuscript are recreated.
For a detailed description on how the plots and tables
for the manuscript were created, please consult the <a href="https://liningtonlab.github.io/mvpaShiny_documentation/publication/R_script/" target="_blank" rel="noopener noreferrer">mvpa R workflow</a>.

## How to navigate

Press the left and right arrow keys to navigate through the slide deck.

Press **P** to open a presenter mode that reveils additional notes to each slide.

## Resources

Either download the mvpaShiny package and work locally <a href="https://github.com/liningtonlab/mvpaShiny" target="_blank" rel="noopener noreferrer">(link)</a>

or use the online version on <a href="http://www.mvpashiny.org" target="_blank" rel="noopener noreferrer">www.mvpashiny.org</a>!

---
# 1. Load, standardize and save the HOMA-IR dataset

<img src="img/1_LoadAndStandardizeData.gif"
	 style="object-fit:fill;
 width:800px;
 height:auto;
 border: solid 1px #CCC"/>
---
# 2. Inspect variable statistics

<img src="img/2_InspectFullDataset.gif"
	 style="object-fit:fill;
 width:800px;
 height:auto;
 border: solid 1px #CCC"/>
---
# 3. Create the HOMA-IR - lipoproteins subset

<img src="img/3_CreateHOMOIRLipoproteinSubset.gif"
	 style="object-fit:fill;
 width:800px;
 height:auto;
 border: solid 1px #CCC"/>
???
Press the control (Windows) or command (MacOS) key to select variables or objects one after another.

Press the shift key to select several items as shown in the clip.
			
---
# 4. Perform PLS-R to obtain target scores **tTP**

???
Setting a seed is not mandatory but 
enables a reproducible Monte-Carlo resampling procedure.
---
# 5. Inspect PLS-R model

---
# 6. Transfer **tTP** to dataset

<img src="img/6_TransferTTPToCurrentData.gif"
	 style="object-fit:fill;
 width:800px;
 height:auto;
 border: solid 1px #CCC"/>
???
The target score **tTP** is obtained from the Target Projection method on
the PLS-R model, using the determined optimal number of components
and using the full dataset.

Short reminder: Here, we use the target score as a proxy for the explained variance in HOMA-IR of all the lipoproteins. It is effectively a dimension reduction step.

---
# 7. Inspect customized correlation matrix

---
# 8. Perform covariate projection

???
Keep in mind that the order in which the covariates are selected matters and affects the resulting tables.
---
# 9. Use adjusted dataset as current dataset

???
When saving a dataset, mvpaShiny performs a check to see if the dataset
is fit for the downstream analyses, such as PLS-R. That is why the invariant (variance = 0)
variables have to be removed, before the dataset can be saved.
			
---
# 10. Create subset without lipoproteins

<img src="img/10_Figure4CCreation.gif"
	 style="object-fit:fill;
 width:800px;
 height:auto;
 border: solid 1px #CCC"/>
			
---
# 11. Perform PLS-R

<img src="img/11_Figure4CPLSR.gif"
	 style="object-fit:fill;
 width:800px;
 height:auto;
 border: solid 1px #CCC"/>
			
---
# 12. Inspect selectivity ratios and fractions