run ichorCNA on hg38

[dnaase]

Hi, I met a problem when run ichorCNA on hg38 chromsomes, i set up everything is in hg38 coordinate now, but it seems that there are some problem internally when loading reads from tumor wig file:

here is the error msg: ... Parsing: fixedStep chrom=chrY start=1 step=1000000 span=1000000 Sorting by decreasing chromosome size Correcting Tumour Loading required package: GenomeInfoDb Error in [[<-(*tmp*, name, value = c(-1, 0.572853, -1, 0.564709, 0.478991, : 3044 elements in value to replace 3021 elements Calls: loadReadCountsFromWig -> $<- -> $<- -> [[<- -> [[<- Execution halted

[gavinha]

Hi Yaping,

This error is likely due to the tumor wig file having fewer (3021) rows than the gc wig file (3044). We will try to add a check for this and return only bins that overlap between the two wig files. In the meantime, can you please double check that the tumor, normal, and gc wig files have identical rows?

Thanks, Gavin

[dnaase]

Hi Gavin, Yes, I think that is the problem. they showed different number of rows... thanks!

Yaping

[dnaase]

Hi Gavin, I just correct the wig length problem, but still got some errors:

Parsing: fixedStep chrom=chrY start=1 step=1000000 span=1000000 Sorting by decreasing chromosome size Correcting Tumour Loading required package: GenomeInfoDb Removed 0 bins near centromeres. Error in correctReadCounts(counts, chrNormalize = chrNormalize) : Missing one of required columns: reads, gc, map Calls: loadReadCountsFromWig -> correctReadCounts Execution halted

Could you help me on it?

Thanks,

Yaping

[gavinha]

Hi Yaping,

It appears that using R version 3.3.X should work. I suspect that the GenomeInfoDB package for R version 3.4.X has changed and is causing issues with chromosome naming conventions.

Best, Gavin

[tdido]

Related to this: I'm using the Gencode FASTA for GRCh38, where the chromosomes are named "chr1" instead of just "1". I replaced the parameter in the Snakefile for the read depth calculations and everything seems to work, but just wanted to check whether there were other parameters that should be modified as well (in particular "ichorCNA_chrs").

Thanks!

[enmce]

Hello!

I have same issue, i checked every input to loadReadCountsFromWig function in RunIchorCNA script, which fails to run without explicit error, it contains right prefix 'chr' in all required fields, but still, i have following error:

Loading required package: GenomeInfoDb
Removed 0 bins near centromeres.
Error in correctReadCounts(counts, chrNormalize = chrNormalize) : 
  Missing one of required columns: reads, gc

I have suspicion that GenomeInfoDB fails to fix prefix issue, but how to fix this? Please help.

[gavinha]

Hi @enmce

You can try to use --chrs "paste0('chr', c(1:22, \"X\"))". The same goes for --chrTrain "paste0('chr', c(1:22))". Let me know if that works.

I am in the process of doing a rehaul of the snakemake workflow to play better with hg38. Sorry for the inconvenience.

Best, Gavin

[enmce]

Hello!

I`tried this fix, but got the same error:

Correcting Tumour
Loading required package: GenomeInfoDb
Removed 0 bins near centromeres.
Error in correctReadCounts(counts, chrNormalize = chrNormalize) : 
  Missing one of required columns: reads, gc
Calls: loadReadCountsFromWig -> correctReadCounts

Then i convert chromosome id`s into character vectors using --chrs "paste0(c(1:22, "X"))" --chrTrain "paste0(c(1:22))", script goes a bit further, and produces this error:

Correcting Tumour
Loading required package: GenomeInfoDb
Removed 9743 bins near centromeres.
Applying filter on data...
Correcting for GC bias...
Correcting for mappability bias...
Filtering low uniqueness regions with mappability score < 0.9
Error in genomeStyle %in% seqlevelsStyle(as.character(x)) : 
  object 'genomeStyle' not found
Calls: loadReadCountsFromWig ... getGender -> loadReadCountsFromWig -> setGenomeStyle -> %in%