MAF to MAE

[emmaroth]

Hi, I am working with some MAF data that I would ideally like to fit into a MAE. I am aware of bioconductor's MAFtools, but the output is not matrix based and thus not able to fit in to something like a SummarizedExperiment. Do you know of any ways to get MAF data into the ExperimentList?

[lwaldron]

RaggedExperiment (https://bioconductor.org/packages/RaggedExperiment/) is designed for ragged data like MAF and for compatibility with MultiAssayExperiment, with convenient features for subsetting by range, gene, etc and coercion to matrix format when needed.

The MAF class exported by MAFtools is just a list so doesn't provide any of the methods that would are required for MultiAssayExperiment, and it wouldn't be so easy to make it so.

[lwaldron]

I see that the MAF class has two separate data slots that are each equivalent to a RaggedExperiment, data and maf.silent, several slots that would all belong in colData, clinical.data +variant.classification.summary corresponding to colData, a gene.summary that is a matrix, and metadata. It kind of looks like it would convert to a MultiAssayExperiment containing two RaggedExperiments, a matrix, colData, and metadata.

[lwaldron]

(or to a single RaggedExperiment, involving combining the silent & non-silent mutations and just putting the gene.summary in metadata)

[PoisonAlien]

I haven't used RaggedExperiment, or MultiAssayExperiment to be honest, so I can't really comment on how easily it can be done. But I am more than happy to contribute :)
By the looks of it, it seems MAF can be coerced into RaggedExperiment since variant data can be transformed into GRanges. I will spend some time on this soon.

[PoisonAlien]

I see that the MAF class has two separate data slots that are each equivalent to a RaggedExperiment, data and maf.silent, several slots that would all belong in colData, clinical.data +variant.classification.summary corresponding to colData, a gene.summary that is a matrix, and metadata. It kind of looks like it would convert to a MultiAssayExperiment containing two RaggedExperiments, a matrix, colData, and metadata.

Here is a crude implementation based on your suggestion and my limited knowledge of MAE. Seems to work on my test MAF files but I am sure there will be a ton of bugs. It converts data and maf.silent slots of MAF object into RaggedExperiment and then to MultiAssay_MAF - which in-turn inherits from MultiAssayExperiment class.
There is a custom show method to display some intuitive info - but it can be taken out.

Example dataset from maftools vignette:

> library(maftools)
> library(MultiAssayExperiment)
# source from Gist
> devtools::source_gist(id = "https://gist.github.com/PoisonAlien/2bd04dd7ac86504a746e37dbe7dddfc4")

# Path to TCGA LAML MAF file
> laml.maf = system.file('extdata', 'tcga_laml.maf.gz', package = 'maftools') 
#clinical information containing survival information and histology. 
> laml.clin = system.file('extdata', 'tcga_laml_annot.tsv', package = 'maftools') 

> laml = maftools::read.maf(maf = laml.maf, clinicalData = laml.clin)
An object of class  MAF 
                   ID          summary  Mean Median
 1:        NCBI_Build               37    NA     NA
 2:            Center genome.wustl.edu    NA     NA
 3:           Samples              193    NA     NA
 4:            nGenes             1241    NA     NA
 5:   Frame_Shift_Del               52 0.269      0
 6:   Frame_Shift_Ins               91 0.472      0
 7:      In_Frame_Del               10 0.052      0
 8:      In_Frame_Ins               42 0.218      0
 9: Missense_Mutation             1342 6.953      7
10: Nonsense_Mutation              103 0.534      0
11:       Splice_Site               92 0.477      0
12:             total             1732 8.974      9

# convert MAF to MultiAssayExperiment
> laml_mae = maf2mae(m = laml)
> laml_mae
An object of class MultiAssay_MAF inherited from MultiAssayExperiment
ExperimentList class object of length 2:
 [1] maf_nonSyn: RaggedExperiment with 1732 rows and 193 columns
 [2] maf_syn: RaggedExperiment with 475 rows and 193 columns
193 patients
5 metadata slots:
  summary, variant.classification.summary, variant.type.summary, variants.per.sample, gene.summary
clinical data: 4 features

> experiments(laml_mae)
ExperimentList class object of length 2:
 [1] maf_nonSyn: RaggedExperiment with 1732 rows and 193 columns
 [2] maf_syn: RaggedExperiment with 475 rows and 193 columns

> methods(class = class(laml_mae))
 [1] [                [[               [[<-             [<-             
 [5] $                $<-              anyReplicated    assay           
 [9] assays           c                coerce           colData         
[13] colData<-        complete.cases   dimnames         experiments     
[17] experiments<-    exportClass      getClinicalData  getGeneSummary  
[21] getSampleSummary hasRowRanges     isEmpty          length          
[25] mergeReplicates  metadata         metadata<-       names           
[29] names<-          replicated       sampleMap        sampleMap<-     
[33] show             subsetByAssay    subsetByColData  subsetByColumn  
[37] subsetByRow      updateObject    
see '?methods' for accessing help and source code

# convert  MultiAssayExperiment (MultiAssay_MAF) to MAF
> mae2maf(x = laml_mae)
-Summarizing
-Processing clinical data
An object of class  MAF 
                   ID          summary  Mean Median
 1:        NCBI_Build               37    NA     NA
 2:            Center genome.wustl.edu    NA     NA
 3:           Samples              193    NA     NA
 4:            nGenes             1241    NA     NA
 5:   Frame_Shift_Del               52 0.269      0
 6:   Frame_Shift_Ins               91 0.472      0
 7:      In_Frame_Del               10 0.052      0
 8:      In_Frame_Ins               42 0.218      0
 9: Missense_Mutation             1342 6.953      7
10: Nonsense_Mutation              103 0.534      0
11:       Splice_Site               92 0.477      0
12:             total             1732 8.974      9

[lwaldron]

@PoisonAlien, amazing! Is the reconstructed MAF object actually identical to the original? It doesn't have to be, but if not, are you still able to complete the rest of the MAFtools vignette using the reconstructed object?

The nice thing about the MultiAssay_MAF class is that you could construct it by first importing the MAF file and converting to MultiAssay_MAF, then concatenating other assays and doing any data management activities. With some validity checks to make sure the required components are still present, you could always convert back to MAF to do the MAFtools plotting, even defining methods for the generic functions exported by MAFtools so that they "just work" on MultiAssay_MAF objects.

[PoisonAlien]

I have not tested it thoroughly but my guess would be it should be the same. I will do bit more digging over the weekend, test it with vignette and CN data.

[PoisonAlien]

Hi Marcel,

I think it's quite done now. Could you please review the updated PR and let me know?
I will then proceed to push the changes to BioC devel.

# Installation
# devtools::install_github("PoisonAlien/maftools")
# devtools::install_github("PoisonAlien/MultiAssayExperiment@maf_fix")

laml.maf = system.file("extdata", "tcga_laml.maf.gz", package = "maftools") #MAF file
laml.clin = system.file('extdata', 'tcga_laml_annot.tsv', package = 'maftools') #clinical data

#MAF
laml_maf = maftools::read.maf(maf = laml.maf, clinicalData = laml.clin)

#Convert to MAE
laml_maf2mae = maftools::maf2mae(m = laml_maf)

#Convert MAE to MAF
laml_mae2maf = MultiAssayExperiment:::MultiAssayExperimentToMAF(mae = laml_mae)

sessionInfo(package = c("maftools", "MultiAssayExperiment"))

other attached packages:
[1] maftools_2.7.10              MultiAssayExperiment_1.17.10

Thanks.

[LiNk-NY]

Hi Anand, @PoisonAlien

I've reviewed the changes and left a comment.
Please sync the PR with the changes in the maf branch.
I've moved the function to its own file.

Best,
Marcel

[PoisonAlien]

Hi Marcel,
I have pushed the updated PR to maf branch.

[LiNk-NY]

Great @PoisonAlien ! After you update maftools on Bioconductor, I would need a maftools
package version number to add to the DESCRIPTION file to ensure compatibility.

[PoisonAlien]

Hi Marcel,
Maftools updates have been pushed to the Bioconductor master branch (version: 2.7.10)

[LiNk-NY]

These functions have been added to support conversion from one class to the other:

MultiAssayExperiment::MultiAssayExperimentToMAF (1.17.12)
maftools::maf2mae (2.7.10)

@PoisonAlien Thanks Anand for all your help!

[PoisonAlien]

Thanks to you as well for coordinating. Hope this will be helpful for all the users.

[lwaldron]

Thank you @PoisonAlien, this is a really nice contribution!