cobalt error on long read data

[jiadong324]

Dear developer,

I am running cobalt (v1.14) on tumor-normal paired long read data, and the errors are shown below. Is that due to the low coverage of normal sample (~20X)?

Looking forward to your reply, thanks!

Read Count Complete
13:31:42.574 INFO  [main] - calculating sample ratios for A-tumor
13:31:42.574 INFO  [main] - Applying ratio gc normalization
13:31:46.727 INFO  [main] - median read count: 3.77260592509782, sparse consolidation count: 100
13:31:46.848 INFO  [main] - using 25629 sparse consolidated buckets, from 2560357 raw ratios
13:31:47.084 INFO  [main] - Persisting tumor gc read count to ./cobalt
13:31:47.095 INFO  [main] - calculating sample ratios for A-normal
13:31:47.095 INFO  [main] - Applying ratio gc normalization
13:31:51.049 INFO  [main] - using 25629 sparse consolidated buckets, from 2560359 raw ratios
13:31:51.097 ERROR [main] - [Thread[main,5,main]]: uncaught exception: java.util.NoSuchElementException
java.util.NoSuchElementException
        at java.base/java.util.ArrayList$Itr.next(ArrayList.java:1000)
        at com.google.common.collect.AbstractMapBasedMultimap$WrappedCollection$WrappedIterator.next(AbstractMapBasedMultimap.java:473)
        at com.hartwig.hmftools.cobalt.lowcov.LowCoverageRatioBuilder.populateLowCoverageRatio(LowCoverageRatioBuilder.java:73)
        at com.hartwig.hmftools.cobalt.lowcov.LowCoverageRatioBuilder.<init>(LowCoverageRatioBuilder.java:34)
        at com.hartwig.hmftools.cobalt.ratio.RatioSupplier$SampleRatios.<init>(RatioSupplier.java:109)
        at com.hartwig.hmftools.cobalt.ratio.RatioSupplier$GermlineRatios.<init>(RatioSupplier.java:132)
        at com.hartwig.hmftools.cobalt.ratio.RatioSupplier.tumorNormalPair(RatioSupplier.java:222)
        at com.hartwig.hmftools.cobalt.CobaltApplication.run(CobaltApplication.java:138)
        at com.hartwig.hmftools.cobalt.CobaltApplication.main(CobaltApplication.java:87)

[p-priestley]

How long are your reads? We have not tested COBALT at all with long reads.

[jiadong324]

I found this study https://www.nature.com/articles/s41587-022-01468-y uses hmftools for long read data, so that I want to have a try. For my data, the average length for both tumor and normal sample is 15kb. Since it successfully calculate A-tumor, I think read length might not be the issue.

[p-priestley]

I was not aware of that study. 15kb is very long since we count the read starts in each 1kb bins so I think it could cause problems, but you are right that it may not be the cause here.

Can you provide your exact command that you ran on the bam please?

[jiadong324]

Here is the command:

java -jar ~/Biotools/hmftools/cobalt_v1.14.jar -threads 6 -reference A-normal -reference_bam ./A-normal.bam -tumor A-tumor -tumor_bam ./A-tumor.bam -output_dir ./A/cobalt -gc_profile ./GC_profile.1000bp.38.cnp

BTW, I can successfully run Amber with the same data.

[hongwingl]

Hi, this is a bug in cobalt, I will put out a fix

[jiadong324]

@hongwingl Thanks! Looking forward to your update.

[hongwingl]

I made a release that fixes this bug: https://github.com/hartwigmedical/hmftools/releases/tag/cobalt-v1.14.1

[jiadong324]

@hongwingl Great! Thanks for your help!

[jiadong324]

@hongwingl It works right now. But I have another error running purple:

failed to load ref genome: org.apache.commons.cli.ParseException: Supplied ref genome must have associated sequence dictionary

What is the sequence dictionary?

[hongwingl]

It is looking for a .dict file that gives the collection of sequence information of the fasta file. There are many tools you can use to create it. Such as samtools-dict

[jiadong324]

Thanks! But I also have the same error as mentioned in https://github.com/hartwigmedical/hmftools/issues/405. The GC profile is GC_profile.1000bp.38.cnp, and the reference file is the same one as I used for the read alignment.

[hongwingl]

Can I close this? Looks like has been fixed in #405?