sciClone input, output and parameters

Question

[genomicist]

Dear all,

I tried sciClone (pubmedID 25102416) and having encountered some issues I'd appreciate hearing some thoughts on these.

1. My copy number data used for input is log-ratio data. It seems though as sciClone interpreted it as absolute copy number data, since only cn=2 were used (and according to the paper, sciClone only looks at VAFs in regions of normal cn). Is it possible to make the program use the log-ratio data?
2. The program seems to have filtered away mutations with depth<100. Is it possible to decrease this limit?
3. Finally, the output contains one cluster. However, visually there are 3 subclusters in that cluster. I understand they are too close to each other to be considered separate clusters. Is it possible to make the program output them?

Thanks in advance.

Answer 1

[poisonAlien]

1. Why don't you convert your log-ratio into absolute cn values before inputting into sciclone. 2*(2^log-ration)
2. You can set depth limit manually with the argument minimumDepth
3. For this I guess you will have to play around with maximumClusters argument.

Answer 2

[Chris Miller]

If you look at the help (with ?sciClone), you'll find many of the answers you're looking for.

1. My copy number data used for input is log-ratio data. It seems though as sciClone interpreted it as absolute copy number data, since only cn=2 were used (and according to the paper, sciClone only looks at VAFs in regions of normal cn). Is it possible to make the program use the log-ratio data?*

When calling the sciClone() function, set cnCallsAreLog2=TRUE

2. The program seems to have filtered away mutations with depth<100. Is it possible to decrease this limit?

SciClone works best with deep readcounts, as they decrease the uncertainty associated with each VAF. That said, if you want to decrease it, when calling the sciClone() function, change the minimumDepth parameter.

3. Finally, the output contains one cluster. However, visually there are 3 subclusters in that cluster. I understand they are too close to each other to be considered separate clusters. Is it possible to make the program output them?

That would defeat the purpose of clustering automatically. See if adding more points (by decreasing your minimum depth) helps. If not, feel free to email me a screenshot of your plots at [e-mail] - I'll see if there are any suggestions I can offer.

[genomicist]

Thanks a lot for the tips. Now it works very nicely. And there no need anymore to try and tweak subclusters into clusters, since clusters are readily reported.

It might be useful to put the help pages somewhere online (e.g. like the readme installation guide at github), to make it easier for someone like me who have looked for this information all over (except for in R).

[genomicist]

If sciClone is run with two or three samples, am I right to think that LOH regions common for all samples (intersect) should be used?

[Chris Miller]

Yep - just merge them into one file and use that as your "regions to exclude" file.

[Lisa]

When using one sample, sciclone just work fine. But when tring two samples comparison, an error warning generated:

[1] "checking input data..."
[1] "Not all variants fall within a provided copy number region. The copy number of these variants is assumed to be 2."
[1] "Not all variants fall within a provided copy number region. The copy number of these variants is assumed to be 2."
[1] "ERROR: no sites are copy number neutral and have adequate depth in all samples"

The code is: sc = sciClone(vafs=list(v1,v2),copyNumberCalls=list(cn1,cn2), sampleNames=names[1:2],cnCallsAreLog2=TRUE,minimumDepth=50)

Is there anything wrong?

[Chris Miller]

Hi Lisa. I'd need more info to help you figure out why you're getting this erro. I'd suggest opening up a new Biostar question, and posting at least the header of your input files there so that I have some information to use for troubleshooting. Feel free to shoot me an email when that question is up ([e-mail]) and I'll take a look.

[Lisa]

Hi Chris, I post the question at sciClone error for two sample

Thanks.