parklab/ShatterSeek

Error in chisq.test(obs2, p = rep(1/4, 4)) : at least one entry of 'x' must be positive

Closed this issue · 14 comments

alvinwt commented

Hi,

I think this error message appears when the obs2 variable for the chisq.test is c(0,0,0,0).

When I debugged line by line, I think there are a number of TRA events but no event from the DEL, DUP, h2hINV, t2tINV categories. Perhaps a test for non-zero before the chisq.test is needed.

isidroc commented

Hello Alvin,
Thanks for your comment. Please note that in function_criteria.R we only use that function when the number of SVs > 0. If you give me more information I might be able to help with your particular issue.
Best,
Isidro

alvinwt commented
isidroc commented

Thanks Alvin. I will have a look at this. Please note that TRA assumes that the SVs are interchr. That is, there are no intrachr SVs in the TRA category.

alvinwt commented

Thanks @isidroc, I checked and there are no intrachr SVs in the TRA category. The manta calls include INS, causing nrow > 1 but not falling into the main svclasses. How would you classify the SV type for INS for shatterseek? Would you filter it out for the analysis.

isidroc commented

Hi Alvin, apologies for my tardiness (I reply by email more quickly). INS will generate two breakpoints, which you can classify into the 4 categories we consider. Happy to help you further if needed.

sainadfensi commented

Hi Isidro,
I am having the same issue and the program corrupted.

Error in chisq.test(obs2, p = rep(1/4, 4)) : 
  at least one entry of 'x' must be positive
In addition: Warning messages:
1: In min(c(SVsnow$pos1, SVsnow$pos2)) :
  no non-missing arguments to min; returning Inf
2: In max(c(SVsnow$pos1, SVsnow$pos2)) :
  no non-missing arguments to max; returning -Inf
3: In min(c(SVsnow$pos1, SVsnow$pos2)) :
  no non-missing arguments to min; returning Inf
4: In max(c(SVsnow$pos1, SVsnow$pos2)) :
  no non-missing arguments to max; returning -Inf
5: In min(c(SVsnow$pos1, SVsnow$pos2)) :
  no non-missing arguments to min; returning Inf
6: In max(c(SVsnow$pos1, SVsnow$pos2)) :
  no non-missing arguments to max; returning -Inf
7: In chisq.test(obs2, p = rep(1/4, 4)) :
  Chi-squared approximation may be incorrect
Called from: chisq.test(obs2, p = rep(1/4, 4))

Looking at the SV data, there are TRA, DEL, DUP, and INS, but I still has the error. I appreciate any suggestions.

  • 😕2
alhafidzhamdan commented

Any futher progress regarding this error? I got the same error when i used a GRIDSS-called VCF (all svtype "BND").
Any suggestions as to how one can convert BND to your preferred svtypes?

Evaluating the statistical criteria
Error in chisq.test(obs, p = rep(1/4, 4)) : 
  at least one entry of 'x' must be positive
In addition: Warning messages:
1: In min(c(SVsnow_window$pos1[which(SVsnow_window$chrom1 == chr_inter)],  :
  no non-missing arguments to min; returning Inf
2: In max(c(SVsnow_window$pos1[which(SVsnow_window$chrom1 == chr_inter)],  :
  no non-missing arguments to max; returning -Inf
alhafidzhamdan commented

Any futher progress regarding this error? I got the same error when i used a GRIDSS-called VCF (all svtype "BND").
Any suggestions as to how one can convert BND to your preferred svtypes?

Evaluating the statistical criteria
Error in chisq.test(obs, p = rep(1/4, 4)) : 
  at least one entry of 'x' must be positive
In addition: Warning messages:
1: In min(c(SVsnow_window$pos1[which(SVsnow_window$chrom1 == chr_inter)],  :
  no non-missing arguments to min; returning Inf
2: In max(c(SVsnow_window$pos1[which(SVsnow_window$chrom1 == chr_inter)],  :
  no non-missing arguments to max; returning -Inf

I used StructuralVariantAnnotation and convert the none DEL/DUP/INS to TRA and it works a treat.

isidroc commented

Thanks for providing a solution @alhafidzhamdan . Alternatively, you can preprocess your calls using https://github.com/ljdursi/mergevcf
Please check the documentation re the input requirements, as e.g. BND are not accepted.
Best and thanks,
Isidro

yutito commented

Thank you for your great work.
Probably my problem is related to this issue#9.

When I ran the package, I have two different warning message;
In chisq.test(obs2, p = rep(1/4, 4))Chi-squared approximation may be incorrect
In max(c(SVsnow_window$pos1[which(SVsnow_window$chrom1 == chr_inter)],no non-missing arguments to max; returning -Inf

If I understand correctly, the number of SVs is too small in calculating "Fragment join test". So, I think it is an unavoidable message. But, if there is any solution to put these warning messages away, could you tell me how to manage.

Thank you in advance for your generous support.

isidroc commented

Can you please send me a minimal working example (i.e. the SV and CN files/data you are using) so I can see how to fix this? Thanks

isidroc commented

I have modified the code now in any case. @yutito @alvinwt Can you please check whether it works for you now? Thanks!

isidroc commented

I assume it works as I have not received any further comments. Thanks for reporting

Giuseppe1995 commented

Hi, sorry for this intromission, but I had the same problem and I resolved it just by converting "BND" with "TRA" (trivially, gsub("BND","TRA", mySVtab$mySVType_column ) ). I also had a Manta output, which I converted to bedpe format via svtools . Hope this is useful.

  • 👍2