SCN Individual contributions

[Gershon Spitz]

Dear Chris,

Thank you for the speedy responses thus far with brainGraph.

I have been attempting to extract individuals contributions for SCN. The SCN analysis comprises two groups. I have used the Scheffer atlas and Freesurfer thickness modality. I am using brainGraph 2.7.3

I have two queries:

1) I have encountered an error when I run the aop and loo functions:

Error in { : task 1 failed - "must have >4 observations”

Running “as.numeric(myResids$resids.all$Group)” returns a numeric vector, without any NA or NULL values.

Any ideas what could be causing this error?

2) I would like to calculate individual contributions of certain network properties. I saw there was a previous thread on this. If, for example, I wanted to extract individual contribution to global efficiency or nodal efficiency would the edited function look like the following:

function (resids, corr.mat, level = c("global", "regional"), 

    control.value = 1) 

{

    Group <- Study.ID <- i <- NULL

    stopifnot(inherits(resids, "brainGraph_resids"))

    groups <- resids$groups

    kNumSubj <- resids$resids.all[, tabulate(Group)]

    if (is.numeric(control.value)) {

        control.int <- control.value

        control.str <- groups[control.int]

    }

    else {

        control.int <- which(groups %in% control.value)

        control.str <- control.value

    }

    patient.str <- groups[-control.int]

    patient.int <- seq_along(groups)[-control.int]

    control.inds <- resids$resids.all[, which(Group == control.str)]

    level <- match.arg(level)

    if (level == "global") {

        IC <- sapply(groups[-control.int], function(x) NULL)

        for (j in patient.int) {

            pat.inds <- resids$resids.all[, which(Group == patient.str)]

            IC[[groups[j]]] <- foreach(i = seq_len(kNumSubj[j]), 

                .combine = "c") %dopar% {

                resids.aop <- resids[c(control.inds, pat.inds[i])]

                resids.aop$resids.all[, `:=`(Group, control.str)]

                resids.aop$resids.all <- droplevels(resids.aop$resids.all)

                setkey(resids.aop$resids.all, Group)

                new.corr <- corr.matrix(resids.aop, densities = 0.1)[[1]]$R

             g <- graph_from_adjacency_matrix(corrs[[group.vec[i]]]$R, mode='undirected', diag=FALSE)

g.new <- graph_from_adjacency_matrix(new.corrs[[1]]$R, mode='undirected', diag=FALSE)

g.eff <- efficiency(g, 'global', use.parallel=FALSE)

g.eff.new <- efficiency(g.new, 'global', use.parallel=FALSE)

g.eff - g.eff.new

            }

            IC[[groups[j]]] <- cbind(resids$resids.all[groups[j], 

                c("Study.ID", "Group")], IC[[groups[j]]])

        }

        DT <- rbindlist(IC)

        setnames(DT, "V2", "IC")

    }

    else if (level == "regional") {

        RC <- sapply(groups[-control.int], function(x) NULL)

        for (j in patient.int) {

            pat.inds <- resids$resids.all[, which(Group == patient.str)]

            RC[[groups[j]]] <- foreach(i = seq_len(kNumSubj[j]), 

                .combine = "rbind") %dopar% {

                resids.aop <- resids[c(control.inds, pat.inds[i])]

                resids.aop$resids.all[, `:=`(Group, control.str)]

                resids.aop$resids.all <- droplevels(resids.aop$resids.all)

                setkey(resids.aop$resids.all, Group)

                new.corr <- corr.matrix(resids.aop, densities = 0.1)[[1]]$R

                g <- graph_from_adjacency_matrix(corrs[[group.vec[i]]]$R, mode='undirected', diag=FALSE)    # Add "weighted=TRUE" if you want a weighted metric

g.new <- graph_from_adjacency_matrix(new.corrs[[1]]$R, mode='undirected', diag=FALSE)

n.eff <- efficiency(g, 'nodal', use.parallel=FALSE)
n.eff.new <- efficiency(g.new, 'nodal', use.parallel=FALSE)

n.eff - n.eff.new

            }

            RC[[groups[j]]] <- cbind(resids$resids.all[groups[j], 

                c("Study.ID", "Group")], RC[[groups[j]]])

        }

        RC.dt <- rbindlist(RC)

        DT <- melt(RC.dt, id.vars = c("Study.ID", "Group"), variable.name = "region", 

            value.name = "RC")

    }

    out <- list(method = "Add one patient", level = level, DT = DT)

    class(out) <- c("IC", class(out))

    return(out)

}

Let me know if you need any more information and thanks for the continued support!

Gershon.

[Chris Watson]

1. That error comes from the "rcorr" function in "Hmisc", which is called in "corr.matrix". I am not sure what is causing the error without more information.

2. It looks correct at a glance but I am not 100% sure. Best thing would be to try it out and see what happens.

Chris

--
You received this message because you are subscribed to the Google Groups "brainGraph-help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to brainGraph-he...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/brainGraph-help/CA900870-F580-44C0-ABE1-AFA000FD7023%40monash.edu.

[Gershon Spitz]

Hi Chris,

After spending a long time trying to diagnose the problem with ‘app’ I think I have found the issue.

Indeed, it was in the correct.matrix function.

I found that replacing the line groups groups <- resids$groups with groups <- resids$groups[[1]] seemed to do the job.

So, it had to do with how groups were being assigned.

I thought this would also fix the ‘loo’ function, but this did not seem to do the trick. I wonder if there is any information I could supply to help diagnose the problem?

Best wishes,

Gershon.

To view this discussion on the web visit https://groups.google.com/d/msgid/brainGraph-help/CAHVz74j--XCCVunJw%3DEQ1_NWLOZSrj%2BHPZ2KEqbtdGni7Qymsg%40mail.gmail.com.

[Chris Watson]

I am not sure what error you are seeing, or the exact code you are running, etc. I would need a lot more information to diagnose whatever issues you are experiencing but I am currently more focused on finishing up v3.0 which might take care of the issue itself.

Chris

To view this discussion on the web visit https://groups.google.com/d/msgid/brainGraph-help/D48D99D4-A7BA-4DC2-BC5D-296566A6676A%40monash.edu.

[Gershon Spitz]

Thanks Chris, Ill look forward to v3. Hopefully this solves it.

Best

To view this discussion on the web visit https://groups.google.com/d/msgid/brainGraph-help/CAHVz74iTkZD0NP16oWuu9WPpH_33uoNXGQWMjzYg2j%3D2YJ9sjw%40mail.gmail.com.