kumquatum / gwena Goto Github PK

How to conduct a phenotypic association using eigengenes of detection$modules_premerge?

Thank you!

Hi @Kumquatum @nturaga I am getting the following error for build_net while using rnaseq counts ( > 20 samples) with default option as per the tutorial. Everything went well until this step.

the standard deviation is zeroError in if (min(cor_mat < -1) | max(cor_mat) > 1) stop("Provided correlation function returned values outside [-1,1].") : 
  missing value where TRUE/FALSE needed

My command is:
net = build_net(counts_fil, cor_func = "spearman", n_threads = 2)

Any pointers please?
Thanks.

https://github.com/Kumquatum/GWENA/blob/cff602da716700b94548c964a752368d8a4e60c3/R/biological_integration.R#LL344C1-L347C23

In the R/biological_integration.R line 344-347 (as referred to above), the returned error should not be eigengene content but the phenotype content. This gives me quite some confusion when the error message appears. Hope it can be corrected in the update! Thank you!

I have annotated my transcripts (from non model species) with GO and KEGG numbers. How to import this information into R and analysis using GWENA?

Hello,
The comparison function to compare two networks is quite memory-heavy.
Is there a way to use multiple cores for this operation?
Mayan

Hi,
the value "e" of the use argument you pass to WGCNA::bicor() in .cor_func_match() is not recognized and results in error:

net_bicor <- GWENA::build_net(
  vst_counts_filtered,
  cor_func = "bicor",
  n_threads = 4
)

> Error in WGCNA::bicor(x, use = "e") : 
  Unrecognized parameter 'use'. Recognized values are 
 'all.obs', 'pairwise.complete.obs'

If I'm not mistaken, inside WGCNA::bicor() the use parameter is used when calculating pearson correlation, so in theory, it should accept "e", but authors decided to disallow it.

> packageVersion("WGCNA")
[1] ‘1.70.3’
> packageVersion("GWENA")
[1] ‘1.0.1’

Hi Gwenaelle, I wanted to export the network data in tabular format for further analysis and wasn't sure which bit in the summarizedexperiment data object needs exporting.

I am trying to use bicor as option to build_net,

network = build_net(
        data,
        fit_cut_off=0.8,
        cor_func="bicor",
        network_type="signed",
        n_threads=32,
    )

but doing so returns the following error:

R[write to console]: Error in WGCNA::bicor(x, use = "e") : 
  Unrecognized parameter 'use'. Recognized values are 'all.obs', 'pairwise.complete.obs'

Please find attached my session info: sessionInfo.txt

@Kumquatum Any idea what is causing the error?

When I try to use the function it gives me this error:

Error in associate_phenotype(modules$modules_eigengenes, SummarizedExperiment::colData(se_kuehne) %>% :
Number of row should be the same between eigengene and phenotypes (samples)

I follow the tutorial so the module should be created automatically... so it is strange

I am trying to run the build_net command on my own expression data, however I run into this error:

Warning message:
In get_fit.cor(cor_mat = cor_mat, fit_cut_off = fit_cut_off, network_type = network_type,  :
  No fitting power could be found for provided fit_cut_off. Taking power for maximum fit. See FAQ for known causes.

For data_expr, I have created a SummarizedExperiment with expression data (counts) and colData for 3 biological replicates of the same sample type. I am able to get all the steps to work for kuehne_expr, but not for my own data, and I am at a loss for what might be causing this error.

When I try to run the Networks Comparison approach described in 2.8 of the vignette, this is the error I get for the build_net step:

Error in if (min(cor_mat < -1) | max(cor_mat) > 1) stop("Provided correlation function returned values outside [-1,1].") : missing value where TRUE/FALSE needed In addition: Warning message: In stats::cor(x, y, use, method) : the standard deviation is zero

This happens with ether counts or log2(counts). Thanks!

Hello there,
I would like to ask whether it is right to use GWENA on proteomics normalized intensities. Any suggestions and recommendations are appreciated.

Thank you,
Abed.

Minimal example :

eigengene_mat <- data.frame(mod1 = rnorm(20, 0.1, 0.2),
                            mod2 = rnorm(20, 0.2, 0.2))
phenotype_mat <- data.frame(phenA = sample(c("X", "Y", "Z"), 20,
                                           replace = TRUE))
association <- associate_phenotype(eigengene_mat, phenotype_mat)

>> Error in cor(x, y, use = use, ...) : 'y' has a zero dimension.

Supposed origin to check : the data.frame with one column is somewhere transformed into vector

Hi @Kumquatum,

I trying the GWENA package which seems very intuitive and easy to use - thank you! My goal s detect gene models between two conditions, specially identify genes which reflect a mutant state (vs control).

The one aspect I am struggling with is the modules differential co-expression. In my experiment I have two conditions, mut and wt, for which 6 and 2 modules were identified with:

lapply(
   se_expr_by_cond, build_net,
   cor_func = "spearman", 
   n_threads = threads_to_use,
   keep_matrices = "both"
)

But when I ran the differential analysis using wt as a reference, the plot_comparison_stats shows the 2 wt modules. Is this expected? Since I am interested in those gene modules which are "specific" to the mutant, I naively thought that I would get those 6(+2) which reflect this.

Dear all,

Thank you very much for the integrative package with a lot of helpful functions!

I would like to know, after bio_enrich() of the module genes, where to find the genes which are enriched in each significant pathways annotated?

Thanks for your help again!

Best,
Yuling