Only non-synonymous variants included?

[Joannavonberg]

Hi developers, first of all thank you for creating this package! It worked immediately out of the box for me, and I've been able to analyse all my samples so quickly :) When I accidentally tried to load an empty file, I got an error message from maftools saying there were no non-synonymous variants present in the file. I checked it for one of my other samples, and I get to the same number of variants that is reported in the katdetectr results if I filter on myloadedMAF$IMPACT %in% c("HIGH", "MODERATE") (i.e. filtering on non-synonymous variants).

So I have two questions:

1. Do you only include non-synonymous variants?
2. If yes, is it necessary / informative to do this? I would naively have thought that it makes sense to look for kataegis genome-wide, and not only in coding regions.

Looking forward to hearing from you!

[Joannavonberg]

I tried reading the same MAF in with the maftools::read.maf() function and I get the same number of variants as katdetectr (214 versus ~45000 that I get if I just read_table() the MAF). So it looks like read.maf() by default only keeps non-synonymous variants, something that I did not see mentioned in the documentation of their function but might be known to you?

[Joannavonberg]

Here is the code for the read.maf() function: https://github.com/PoisonAlien/maftools/blob/master/R/read_maf_dt.R They indeed explicitly keep only non-synonymous variants. They have a parameter vc_nonSyn to use different filters though, would you consider making this a parameter that is passed from detectKataegis, so users can change it themselves if they want (I would be an example of a user that would like to do that ;-)?

[Joannavonberg]

Thank you for the superquick edit! I'm going to try the new version :)

[daanhazelaar]

Dear @Joannavonberg,

Thank you for opening this issue and good to hear that katdetectr works well!

Also thank you for clearly explaining your relevant questions.

I have pushed an update of katdetectr that should make sure that now all variants (both synonymous and non-synonymous) are analysed when supplying a .maf file. Can you please try out this updated version and let me know if things are working to your satisfaction?

# download the latest version of katdetectr from github
devtools::install_git(url = "https://github.com/ErasmusMC-CCBC/katdetectr")

# load and attach katdetectr
library("katdetectr")

# detect kataegis
kd <- detectKataegis(genomicVariants = "path/to/file.maf")

# visualize in a rainfall plot
rainfallPlot(kd)

Hereby my extended answer to your questions:

I definitely recommend including both non-synonymous and synonymous variants when analysing a sample for kataegis. However, as you clearly noted; when using the previous version of katdetectr to analyse a .maf file only the non-synonymous were analysed.

This was indeed due to the maftools::read.maf() function that we use in katdetectr to load .maf files. This function reads a .maf file and constructs a MAF object. I have now realised that the non-synonymous variants are stored in the data slot and the synonymous variants are stored in the maf.silent slot of the MAF object (see code below). I was unaware of this so thanks again for pointing this out!

Previously, katdetectr only analysed the variants in the data slot and therefor only the non-synonymous variants. I now made sure both slots are included and therefor all variants present in the .maf file are analysed by katdetectr.

# Construct MAF object
maf <- maftools::read.maf(maf = system.file(package = "katdetectr", "extdata/APL_primary.maf"), verbose = FALSE)

# acces non-synonymous variants in MAF object
maf@data

# acces synonymous variants in MAF object
maf@maf.silent

# combine all variants in single tibble
maf@data |> 
    dplyr::as_tibble() |> 
    dplyr::bind_rows(maf@maf.silent)

Regarding the vc_nonSyn argument in the maftools::read.maf() function. This argument can be used to discern if a certain type of variant should be considered as non-synonymous or synonymous in a MAF object (see code below).

# Construct MAF object
maf <- maftools::read.maf(maf = system.file(package = "katdetectr", "extdata/APL_primary.maf"), verbose = FALSE, vc_nonSyn = "Frame_Shift_Del")

# This contains only Frame_Shift_Del non-synonymous variants
maf@data

# This contains all other (synonymous) variants
maf@maf.silent

I prefer not incorporating the vc_nonSyn argument as we want to minimise the preprocessing options in katdetectr. If you want to analyse preprocessed data please do the preprocessing yourself, save this processed data as a .maf file (or vcf or VRanges) and than use katdetectr to analyse your data.

I hope this explanation is useful for you. If you have any follow-up or other questions or comments feel free to reach out!

Kind regards, Daan Hazelaar

[Joannavonberg]

Hi @daanhazelaar , thanks for the elaborate response! Fully agree with not using the vc_nonSyn argument, I was just trying to see what could be a quick fix but didn't know about the separately saved non-coding and synonymous variants. I'm trying out the new version at the moment, but obviously it's a bit slower with so many more variants so I might update you next week

[daanhazelaar]

Hi @Joannavonberg,

No problem and sounds good! I'm looking forward to your update!

Have a nice weekend!

Kind regards, Daan Hazelaar