I would like to know how to interpret these matrices if my input file for the WGCNA package was a matrix of normalized expression values for multiple arrays. Is correlation among the genes used for calculation of adjacency matrix?
WGCNA constructs two matrices, first it defines a correlation matrix up to a power beta so the degree distribution will fit a small-word network.
This matrix given only information about the expression correlation between genes.
WGCNA thinks that co-expression is not enough and the similarity between genes should be reflected at the expression and the network topology level.
This is why it defines the TOM matrix which uses the co-expression Adjacency matrix and build another adjacency matrix that considers topological similarity.
Normally the TOM matrix is the final result of WGCNA.
Dear Mohamed Nadhir Djekidel,
I understand that the adjacency matrix is calculated using Pearson correlation as the metric and it gives us correlation between each pair of genes from the input normalized expression matrix. But, if you look at the TOM matrix, we do not see any gene names. Can you tell me exactly how TOM is calculated. e.g. the first matrix uses correlation as the metric and the values in the first adjacency matrix are actually Pearson correlation values, what is the metric used for TOM. What do the values in TOM matrix actually mean, and what do the first row and first column of TOM matrix represent ?
Thanks for all the help.
The column order of TOM and the adjacency are the same, so you can just copy the names in the "colnames" and "rownames"
Just in case you have not read these yet I am attaching two papers. They were some of the more useful literature that I have come across for explaining WGCNA concepts.
The "WeightedNetwork2005" paper is a great detailed overview while the "signedTOM-explanation" report really helped to clear up some TOM concepts.
Hi, steps in WGCNA are:
1. Similarity co-expression matrix (Pearson's correlation).
2. Adjacency matrix (signed, unsigned, signed-hybrid).
3. Topological Overlap Matrix (TOM) constructed from adjacency matrix.
(TOM formula is presented in below image)
4. Dissimilarity matrix (1-TOM) obtained from TOM.
*By using TOM matrix, the network in WGCNA assumes not only co-expression information, also it assumes the network topological information.
WGCNA was used to construct co-expression network and
significant modules in network. view the file
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis