Plot feature loadings function

[ElySeraidarian]

According to #92, we need a way to plot feature loadings after performing a reduction method by several ways. Here is a first version with PCA that plots feature loadings in 4 different ways : Tree, Barplot, Screeplot & Heatmap. This will be available for LDA & NMF as soon as it is possible to retrieve loadings matrix from reducedDim slot. There are parameters that can be added so that we can talk about it (for example the number of features plotted). The parameter method is probably doomed to disappear as we will be able to recognize which method has been used in the reducedDim slot. There are examples for NMF & LDA at the end for now without using the function. This will need examples in OMA and strong documentation in order to tell people how to use this function (agglomerate tree if plotting the tree, only plotting few features, ...). It is currently possible to give either loadings matrix or TSE object to plot.

I will add tests & generate documentation also later.

[ElySeraidarian]

Changes have been made, heatmap has been changed and parameter add.tree instead of tree layout

[TuomasBorman]

I simplified the code a lot. Check that it still works as expected. Also run BiocCheck::BiocCheck() and fix issues related to this added R file (especially, do not exceed 80 characters in line width). Also use inheritParam tag in the documentation.

[TuomasBorman]

Thing to discuss: should rank be row.var? --> variable defining a column from rowData. That is what it is doing currently

[ElySeraidarian]

Thing to discuss: should rank be row.var? --> variable defining a column from rowData. That is what it is doing currently

Yes I believe this would make more sense

[antagomir]

Seems good.

[ElySeraidarian]

But we could also still create a parameter rank that agglomerates tree as in the other functions

[TuomasBorman]

But we could also still create a parameter rank that agglomerates tree as in the other functions

If we agglomerate the data, we lose the match between loadings and rows. Loadings are for features of un-agglomerated data --> that is my first impression, check if that is correct

[ElySeraidarian]

But we could also still create a parameter rank that agglomerates tree as in the other functions

If we agglomerate the data, we lose the match between loadings and rows. Loadings are for features of un-agglomerated data --> that is my first impression, check if that is correct

Yes that's right, agglomeration has to be done before performing the reduction method

[TuomasBorman]

1. I removed authors. See the issue: Remove authors from functions mia#628
2. You should perform first agglomeration and then transform the data (fixed)
3. I added a check that the feature names match in tree heatmap

[TuomasBorman]

Looks very nice. Are you interested in updating OMA: https://microbiome.github.io/OMA/docs/devel/pages/beta_diversity.html#sec-dbrda-workflow

If I remember correctly, there were some error?

[ElySeraidarian]

Looks very nice. Are you interested in updating OMA: https://microbiome.github.io/OMA/docs/devel/pages/beta_diversity.html#sec-dbrda-workflow

If I remember correctly, there were some error?

I will update using this function, I found same results when plotting. Also this example shows how to retrieve the loadings (for now we can plot only by getting matrix and then giving the matrix to plot).