Question: Regarding Somatic Mutation Signature
0
3.5 years ago by
sj110
sj110 wrote:

Hi Sir,

For plotMutationSpectrum function,I want to know the exact process of calculating contribution.

Regarding second question I have run the same code 10 times,please have a look at the link for the code https://www.dropbox.com/s/sug1i8v1zoqm1ce/Mutation_signature_package_data.R?dl=0

Please have a look at this link https://www.dropbox.com/s/p30pdl0a7ifdv7a/ss1_ss2.png?dl=0 where I have attached an image of two plots which are slightly different to each other. I have used this function

plotSignatures(sigs_nmf) + ggtitle("Somatic Signatures: NMF - Barchart") and run this 10 times on the dataset supplied with the package,the plot on the right side occurred 8 times whereas the plot on the left side occurred 2 times.

We have also run this for our dataset but the dissimilarity is more.

Another question is that for your dataset supplied with the package,what is the index column meant for?

Also,what do to if I have to run this package for chr X,chr Y,chr MT.Is this package supports this?

Thanks

Swati

somaticsignatures • 638 views
modified 3.4 years ago by Julian Gehring1.3k • written 3.5 years ago by sj110

Dear Swati, please consider asking separate questions individually. This makes it easier and quicker to answer them, and helps others who may have similar questions to find the right answer. Thank you.

Regarding "Another question is that for your dataset supplied with the package,what is the index column meant for?", can you elaborate what you mean by the dataset and the "index column"?

2
3.4 years ago by
Julian Gehring1.3k
Julian Gehring1.3k wrote:

1. "Also,what do to if I have to run this package for chr X,chr Y,chr MT.Is this package supports this?"

All chromosomes are supported, as long as you have a reference genome that contains them.

2. "plotMutationSpectrum function,I want to know the exact process of calculating contribution."

As I outlined in Regarding somatic signature package, the function visualises the observed mutational spectrum 'M'. The details are described in detail in the vignette and in the referenced publications. With normalize = FALSE, the observed counts are plotted. With normalize set to TRUE, the counts for each motif are divided by the total number of counts in the samples.

3. NMF rerun multiple times

The NMF uses numerical methods to decompose the mutational spectrum, and some of these methods are based on a random initialisation of the matrices. The fact that several runs produce slightly different results is therefore not surprising. If you want to avoid this behaviour, you can

- Select a NMF algorithm that uses a deterministic seeding (see the vignette of the NMF package for some nice details and choices)

- Use a different decomposition method (such as PCA, k-means, ...)

- Set the random seed before running the computation. This way, rerun will give you consistent results.

4. index

The index column refers to the row of this entry in the original MAF file. It is generally not relevant for analyses and is kept to find entries in the original data set.

There are several Data sets present in the package SomaticCancerAlterations and one of them is gbm_tcga which is a GRanges object.You have run this package for all the datasets.While exporting the gbm_tcga dataset I have converted it into a dataframe to see the structure of the dataset,please have a look at the link to see the structure https://www.dropbox.com/s/kg5x4wxijc3cfzj/gbm_tcga_test.txt?dl=0 ,so there are several columns,what is the index column meant for?

I am getting an error -

> sigs_nmf = identifySignatures(sca_mm, n_sigs, nmfDecomposition)
Error in rowQ(imat, ncol(imat)) : cannot handle missing values.
Calls: identifySignatures ... standardGeneric -> eval -> eval -> eval -> rowMax -> rowQ -> rowQ
In .local(x, rank, method, ...) :
NMF residuals: final objective value is NA
Execution halted

Please have a look at the link https://www.dropbox.com/s/wc0a0sprzqcydnq/Data.maf?dl=0 , where a toy example of maf file is there and the following link for the code https://www.dropbox.com/s/m6pf05hnyfj7gos/code.R?dl=0

Swati

A gentle reminder sir,if you could solve the above problem then it would be very helpful for us.

Swati

Please post new questions separately; this also increases significantly the chance that someone will see them. Please make sure to include all relevant information, as described in the posting guide.

If we could answer your original questions, please mark them as accepted, so those that try to help you as well as other users with similar issues can understand what worked for you.