Hot answers tagged

4

You have several options to approach this with WGCNA (weighted correlation network analysis). You can run a WGCNA on the combined set, identify modules and select those lncRNAs for further follow-up that have relatively high intra-modular connectivity (i.e., are intramodular hub genes). I would suggest this approach first. Second option is to run WGCNA on ...


3

It's hard to say without knowing how different the subtypes are. If you do a common WGCNA, you may find modules related to the differences between A and B as well as to drivers of expression variation in A and B (common and separate). Since A has fewer samples, the variation will have to be stronger to be reflected by WGCNA modules. You can run the WGCNA, ...


3

You can use exportNetworkToCytoscape or exportNetworkToVisANT to generate and egde list file with the two nodes at the ends and weight for each edge. Hopefully there's a way to import that into networkX.


2

For outlier identification I suggest using the sample network approach developed by Oldham et al. It basically amounts to constructing a inter-sample connectivity (assuming normalizedData contains the log-normalized data with genes in colunms and samples in rows) k = colSums(cor(t(normalizedData))), scaling the connectivities Z.k = scale(k) and then ...


2

The problem is that blockwiseModules split your data into (probably 3) blocks, because the default maxBlockSize is 5000 and smaller than the number of genes in your data. If you have enough RAM (16GB is recommended for 12k genes, but 8GB just might be enough), rerun blockswiseModules again with maxBlockSize raised to above the number of your genes. Since you ...


2

Looks like a bug in the code which I will try to hunt down and fix. In the meantime, I would suggest playing with maxBlockSize argument to blockwiseModules. Try increasing it as much as your available RAM allows (see the paragraph "A second word of caution concerning block size" on page 6 of WGCNA tutorial I, section 2c, for some guidance on how to set ...


2

A cursory search to bedtools documentation will reveal the bedtools closest feature - which might be exactly what you are looking in your third question. You can download the list of protein-coding genes from Ensembl biomart or UCSC table browser in bed format, and also convert the list of your lncRNAs into a bed compatible file. Rest is easy: bedtools ...


2

Your question is not clear, you seem to mix the relevant terminology... First, you cannot run WGCNA for individual samples, only for individual conditions (with or without WT, depending on what question you're trying to answer and assuming you have enough samples in each condition). If that's what you have done, fine. The number of modules in each ...


2

It is likely that the cut heights set in the code (80 and 100*ratio of sample numbers) are too low. Look at the sample trees plotted into a pdf file (SampleClustering.pdf) and note the merging heights. You need to set the cut heights appropriately so that most samples go into a single cluster. Peter


2

The bad news is that, indeed, projectiveKMeans is not parallelized and I am not sure how much of it is (easily) parallelizable. The good news is that with 15k features (genes) and 96GB RAM you don't need to run preclustering at all. Just analyze the whole data set in one block. If you for some reason insist on splitting the data into blocks, make sure the ...


2

Don't worry about it too much. I would go with power 8 based on my general experience (also reflected in WGCNA FAQ) and on the mean connectivity around 50 and median around 20, which seems reasonable to me.


2

I'm not sure I understand "I would like to run GSEA or a similar analysis to find WGCNCA clusters that differ based on the interaction between the two main variables". I would run an association analysis (regression) for module eigengenes and the appropriate interaction term. The (most) significant modules are your candidates. This step is simply ...


2

You CAN calculate M^b for b being float number, complex number or even another matrix. As well as you can calculate exp(M), log(M), sin(M) whatever... That is very standard for math people. Let me give two ways how you can think of it. 1) M^b = exp(b log(M)), so you need to define what is exp and log for matrices. That can be done using power series: exp(...


1

The most likely culprit here is that your expression data (the datExpr input to chooseTopHubInEachModule) has no colnames. Check that your data do have appropriate column names. Also check that the module color vector you supply has at least actual module (type table(moduleColors) and check that you have at least one module apart from "grey" or 0).


1

The biggest difference a large number of samples makes is that you can usually decrease the soft thresholding power. If you use power 6 (or 12 for a signed network), try decreasing it to 3 or even 2 (6 or 4 for signed network). Apart from that, it could also be that the co-expression modules in your data are simply smaller than what people usually find, ...


1

"Mutually exclusive" is not a precise statistical term because it could be seen as independence. Anyway if truly mutually exclusive AND their behaviour follows periodicity, i.e. sine waves, which it might do in a cell cycle, i.e. the cells divide at periodic intervals, you can do this via asynchronicity. If the gene expression can be described under ...


1

Simply select the module(s) you are interested in and look for hub genes using the standard calculations. Having done a module preservation analysis does not change the procedure in the slightest.


1

Your data don't have obvious outliers so you don't have to worry about removing them. The problem is that both sets show prominent sample clusters. I would try to figure out what drives those clusters (is it technical or biological), and whether the variation should be removed. See the comments in WGCNA FAQ under point 5 (My data are heterogeneous. Can I ...


1

The networks you want to compare do not have to have the same number of modules. (The human and chimp networks just happen to have the same number of modules.) In fact, you only need modules in the reference network; modules from the test network are not used at all and need not be defined. You may want to read the original article describing module ...


1

You can use the function corAndPvalue. Simply replace the lines moduleTraitCor= cor(MEs, status, use="p") moduleTraitPvalue = corPvalueStudent(moduleTraitCor, nSamples) by something like cp = corAndPvalue(MEs, status) moduleTraitCor = cp$cor moduleTraitPvalue = cp$p corAndPvalue takes automatically into account the number of valid samples for each ...


1

Thanks for pointing this out. Looks like we forgot to include two important lines. Something along the lines of ConnectivityHuman = colSums(adjHuman)-1 ConnectivityChimp = colSums(adjChimp)-1 This should do it; I'll update the tutorial.


1

What you see is a warning, not an error. Your calculation will run fine, just slower. Unless you see other errors, you should be able to complete all steps of the analysis.


1

2) If not WGCNA, if I use cor function in R with method Pearson, on what cutoff I should select the target genes? Good call, this is a nice approach. It is a good question because the central issue is that a purists approach of 0.7 results in insufficient information to understand the behaviour of the data. The guys here were using 0.4, in my experience ...


Only top voted, non community-wiki answers of a minimum length are eligible