Variables representation
We can also use variables plot to better highlight strong correlations between the variables (here the genes and the lipids). See also this page for more details about correlation circles (2D and 3D plots).
plotVar(result, comp = 1:2, cutoff = 0.5, X.label = TRUE, Y.label = TRUE, cex = c(0.8, 0.8))
You should obtain the following graph:
![](http://www.math.univ-toulouse.fr/~biostat/mixOmics/images/cs_nutri_plotVar.png)
The argument cutoff = 0.5
removed the variable points for which the correlation coordinate was less than 0.5. We can identify which genes (in red) are correlated with which lipids (in blue) by looking at formed clusters. The strength of the correlation is represented by the distance from the origin (the further from the origin the better).
A 3D plot is also available. Here the relevant genes are represented by spheres colored as in González et al. (2009), the others genes are colored in gray. The fatty acids are represented by tetrahedrons in black.
pch <- c("t", "s") col.lip <- rep("black", 21) col.gen <- rep("darkgray", 120) names(col.gen) <- result$names$Y genes1 <- c("CAR1") genes2 <- c("GSTpi2", "CYP3A11", "CYP2c29") genes3 <- c("S14", "ACC2", "cHMGCoAS", "HMGCoAred") genes4 <- c("ACBP", "AOX", "BIEN", "CPT2", "CYP4A10", "HPNCL", "L.FABP", "PECI", "PMDCI", "THIOL", "mHMGCoAS") genes5 <- c("FAS") genes6 <- c("PLTP") col.gen[genes1] <- "darkviolet" col.gen[genes2] <- "darkgreen" col.gen[genes3] <- "red" col.gen[genes4] <- "orange" col.gen[genes5] <- "cyan" col.gen[genes6] <- "magenta" col = list(col.lip, col.gen) plot3dVar(result, cutoff = 0.6, col = col, pch = pch, cex = c(1.5, 1.5), label.axes.box = "axes")
![](http://www.math.univ-toulouse.fr/~biostat/mixOmics/images/cs_nutri_plot3dVar.png)
Here the white circle has a radius of 0.5.
We can also use relevance networks to focus on the relationship between the two types of variables. More details about the methodology can be found in the reference below. Basically, given a rCCA result, we are constructing the network based on the resulting correlations between the variables.
## Interactive = TRUE allows the user to set a correlation threshold to only ## represent variables above that threshold network(result, comp = 1:3, interactive = TRUE, threshold = 0.55)
![]() |
![]() |
This network can be saved as a .glm for an input into Cytoscape, see here.
The Clustered Image Map (see reference below) also allows to visualize the correlation between the variables:
cim(result, comp = 1:3, xlab = "genes", ylab = "lipids", margins = c(5,6))
![](http://www.math.univ-toulouse.fr/~biostat/mixOmics/images/cim_rcc.png)
References
- Martin P.G.P., Guillou H., Lasserre F., Déjean S., Lan A., Pascussi J.-M., San Cristobal M., Legrand P., Besse P. and Pineau T. (2007) Novel aspects of PPARα-mediated regulation of lipid and xenobiotic metabolism revealed through a multrigenomic study. Hepatology 54, pp 767-777.
More on relevance networks and CIM:
- González I., Lê Cao K.-A., Davis, M.D. and Déjean S. (2011) Insightful graphical outputs to explore relationships between two ‘omics’ data sets. Submitted.