Genome MuSic Mutation Relation RUntimes

Question

[Ido Ben-Zvi]

Hi,

I am trying to run the MuSic analyses on a some cohort data that I have. I am using a 4cpu 2xl EC2 machine. I have been able to apply all analyses up to mutation-relation (MRT). Here the application has been running for several days (running several cohorts in parallel, with sizes ranging between 20 and 200). Non of the runs have completed. Previously, even the most extensive calculations on these cohorts did not take up more than 24h.

Looking at the code I see potentially endless loops such as:

doit=1;
while(doit==1)
{
    xi=matrix.sample.v1(x);
    if (max(xi)==1) doit=0;
}

But there could also be less obvious pitfalls, as R loops - looking for restricted permutations.

Can anyone share from their experience running this? Should I abandon ship?

Thanks,
Ido

Answer 1

[Cyriac Kandoth]

Unfortunately it scales exponentially with the number of genes you're testing. I wouldn't run it on more than 30 genes. There are much better alternatives to this pairwise gene test, assuming you're trying to find pathway-level relations. Try MultiDendrix, MeMO, MUTEX, CoMet, etc.

[Ido Ben-Zvi]

Thank you Cyriac