Hi everyone,
I am a little confused about the error complained by GISTIC:
GISTIC 2.0 input error detected:
169 segment overlaps detected in file '/path/***.seg.txt'.
First overlap detected between segments at lines 26227 and 26377.
I checked the first overlap lines in R, they are below:
line 26227:
Sample Chromosome Start End Num_Probes Segment_Mean
<chr> <int> <int> <int> <int> <dbl>
1 TCGA-G4-6317 1 3218610 14832648 6157 -0.160
line 26377:
Sample Chromosome Start End Num_Probes Segment_Mean
<chr> <int> <int> <int> <int> <dbl>
1 TCGA-G4-6317 1 3218610 49468025 24550 -0.099
I got the TCGA segmentation file from firehose, and only eliminated NAs and changed the names of Sample (save only first 12 digits). Does anyone know how to deal with this?
Thanks very much!
Thanks for the jet lag between me and most biostars, it gives me time to solve this problem by my own.
The reason now seems trivial, however, I will note it for completeness. I substringed the first 12 chars of Tumor-Sample-Barcode to get patient only barcode for further analysis. The problem is that, TCGA has harvested not exactly only one tumor sample from one patient, like here, in my dataset, I found there are '01' - Primary Solid Tumor, '02' - Recurrent Solid Tumor and '06' - Metastatic for a single patient (the numbers are the 14th-15th chars of Tumor-Sample-Barcode, details here: TCGA barcode and TCGA code tables).
Then, substring will cause different sample results from a single patient use the same patient barcode as ID, that would cause GISTIC complain, and would also import bias into the following analysis.
OK, that's my experience, put it here as a reminder for myself!
Thanks, but I filtered Primary Solid Tumor samples and it still shows error after GISTIC. I find out 22-25 chars of Tumor-Sample-Barcode can also differ in the same patient. It will result in error. Can you help me?
zhaoliang0302, you should consider contacting the GISTIC team at Broad Institute directly:
Log in to answer this question.
I think I may find a possibl cause, which is collapsing barcode to patients (first 12 chars) would cause different sample types of the same patient to have only one name. I will try this idea and let you know the results.
I think the reason is the cnv segment in line 26227 is included in segment in line 26377, they came from same sample , same chromosome, but the chromosome start and end is overlaped. you should consider to resolve this overlap. and it seems like you have 169 overlaps detected in your segmentation file. good luck!