search

PoisonAlien / maftools

Summarize, Analyze and Visualize MAF files from TCGA or in-house studies.
http://bioconductor.org/packages/release/bioc/html/maftools.html
MIT License
452 stars 222 forks source link

Signature plot #1017

Closed Ind2022 closed 7 months ago

Ind2022 commented 7 months ago

Describe the issue

Command Dear Maftools support,

I am using maftools, NMF and sigminer R packages for construction of signature plots from a whole exome data from 2 sets of data, whole exome. I did use "nrun = 100" and pConstant = 0.01.

While I am able to create the signatures but the diagnostic plots doesn't look all right i.e there is drop at 3 then again it goes up in cophenetic plot from first plot . And it second plot there is no drop, its a straight line.

Please advice if these plots looks fine.

There are 2 sets of data run, so I am sharing 2 plot sets below.

Thank you.

factorizationrank_scid factorizationrank_nude

PoisonAlien commented 7 months ago

Hi, How many samples do you have? Also it appears your mutation load is quite low.

CC: @ShixiangWang

Ind2022 commented 7 months ago

Thank you for your reply. I have 4 samples in first set and on 2nd set there are 3 samples. Kindly let me know if it will be fine to use this results or derived signatures.

PoisonAlien commented 7 months ago

Hi,

I would not use maftools for signature analysis with this low no. of samples. You should consider other tools such as deconstructSigs that can work at the sample level.

Ind2022 commented 7 months ago

Thanks for the suggestion. Does this method will do a similarity analysis with Cosmic signatures as maftools does because that's what we are trying to accomplish.

PoisonAlien commented 7 months ago

It won't extract any new signatures and compare them to COSMIC. It's meant to get the composition of known COSMIC signatures within each sample.

Ind2022 commented 7 months ago

Also does deconstructSigs work with mouse genome rather than human. I am not finding any ref for mouse genome with deconstructSigs

ShixiangWang commented 7 months ago

@Ind2022 Signature fitting is suitable for your data as @PoisonAlien suggested. Use the SigProfiler Assignment https://cancer.sanger.ac.uk/signatures/assignment/ or my sigminer sig_fit (https://shixiangwang.github.io/sigminer-book/basic-workflow.html#reference-signature-fitting) function as alternative is recommended.

Ind2022 commented 7 months ago

Thank you for your response. Like I have mentioned above does these methods work with mouse genome and where can I find mouse genome reference file like maftools provide.

Thanks again.

PoisonAlien commented 7 months ago

Hi, maftools does not provide any reference genomes. It uses BSGenome object to get the sequences. Please have a look at Bioconductor BSgenome packages for your mouse builds.

I don't know how well these signatures translate to the mouse genome, you will have to research these independently.

ShixiangWang commented 7 months ago

They both support the mouse genome. You should check corresponding usage/tutorial. @PoisonAlien I think this could be close now.

@Ind2022 Signature fitting is suitable for your data as @PoisonAlien suggested. Use the SigProfiler Assignment https://cancer.sanger.ac.uk/signatures/assignment/ or my sigminer sig_fit (https://shixiangwang.github.io/sigminer-book/basic-workflow.html#reference-signature-fitting) function as alternative is recommended.

Ind2022 commented 7 months ago

Thank you for your replies.