Why Diablo result is scaled to between -2 and 2?

Analysis
1

Hi mixomics team,

I noticed all my diablo results have legend ranged from -2 to 2 in heatmap created by cimdiablo() function. So I checked that I can get this matrix from my final.model object (final.model$mat).

Somehow I need to interpret this heatmap, but I just figured out that I can not explain why using [-2,2] here. We describe correlation which is between [-1,0].

Thanks

yz

2

Hey, I’m just another mixOmics but if I well understand, there is no correlation in cimDiablo. Indeed, it just takes the input data, center and scale them (mean = 0, standard deviation = 1), then print these values on a heatmap with variables (columns) and samples (rows) reordered by a hierarchical clustering.

However, even after the data were centered and scaled, we could have some high values which could drive all the color code of the heatmap (basically, just one or two values would be in deep blue/red, and all other values would be in yellow so we couldn’t see anything interesting).

To avoid this problem, cimDiablo use a ‘cutoff’ where all values lower than -2 are fixed to -2, and all values higher than 2 are fixed to 2, so the heatmap shows variability of the data.

Hope it helps, and that the mixOmics team will correct me if I misunderstood this function.

Best,
Emile

1 Like
3

Thanks Emile ,
You are right. I did not think from this point of vue

1 Like