Error of netROC function

[TarsLi]

Hi,
I tried to use the NetMoss2 to identify some biomarkers in multiple diseases, however, an error occurred when I tried the netROC function using the testdata. The error information is "Error in y - ymean: non-numeric argument to binary operator", then I change the column type column of the metadata file to binary (healthy:0; disease:1) and the error is gone. After that, another error appeared: Error in order (importance_rf$MeanDecreaseAccuracy, decreasing = T): argument 1 is not a vector“. So could you please offer me some suggestions? Thanks!

[xiaolw95]

Which data have you used, single or multiple? How about trying to set the "type" column of the metadata as factors?

[TarsLi]

I used the example for multiple files. Setting the column "type" of the metadata as factors worked.
Another question is about the abundance files, is it feasible to use only some of the high prevalence genera to identify the key taxon? In other words, the total abundance of these genera is less than 100% and there is variation between samples. Thanks!

[xiaolw95]

Technically, it's ok to use the highest prevalence genera to do your job.
The point is how you construct the ecological network. It's a complicated problem actually. In short, if too many nodes are missing, the co-occurrence network you build may not reflect the authentic ecological interaction. NetMoss indeed has the ability to identify the key taxon as long as there are differences between case and control networks; however, in this case, the results need further confirmation.

[Damianyangyang]

@xiaolw95 @TarsLi
Hi~,I got this error when I ran the netROC, do you know why?
netROC(case_dir = case_dir, control_dir = control_dir, marker = marker, metadata = metadata, plot.roc = TRUE, train.num = 1)

[xiaolw95]

@xiaolw95 @TarsLi Hi~,I got this error when I ran the netROC, do you know why? netROC(case_dir = case_dir, control_dir = control_dir, marker = marker, metadata = metadata, plot.roc = TRUE, train.num = 1)

Sorry for the delayed reply. There was something wrong with the read command of metadata and now has been fixed. You can try it with the NEW VERSION.