Walk-through Example? about mobster HOT 6 CLOSED

Hi @brucemoran,
Have you checked the github.io? https://caravagn.github.io/mobster/

from mobster.

Hi @luca-dex,

thanks for getting back to me.

Yes, I've read through the available resources which are very nice.

I should have been more specific in my question, apologies: it is not obvious how to input data for multiple samples.

The general approach per sample I have used is below.

Therefore do I run this across all samples individually to fit, then interpret results from multiple output?

I would like to compare (sub)clones between multiple samples from a single patient.

Thanks,

Bruce

ds <- read_tsv("data/in/correct/format.tsv")
ds_f <- dplyr::filter(.data = ds, VAF > 0)
m_fit <- mobster_fit(ds_f)
cm_best <- Clusters(m_fit$best)

##ctrees requires driver annotation (arbitrary)
drivers_rows <- c(1, 2, 3)
cm_best$is_driver <- FALSE
cm_best$driver_label <- NA
cm_best$is_driver[drivers_rows] <- TRUE
cm_best$driver_label[drivers_rows] <- c("DR1", "DR2", "DR3")

##fit again
b_fit <- mobster_fit(cm_best)

##ctrees
m_trees <- get_clone_trees(b_fit$best)
m_top_rank <- m_trees[[1]]

##plot
ggpubr::ggarrange(
  ctree::plot.ctree(m_top_rank),
  ctree::plot_CCF_clusters(m_top_rank),
  ctree::plot_clone_size(m_top_rank),
  nrow = 1,
  ncol = 3)

from mobster.

Oh now I see - there is no multivariate mobster analysis (tails can be in principle identified from each biopsy sample independently), bu rather there is a pipeline that

• performs mobster deconvolution from each one of K samples independently;
• determines tail mutations from all samples (i.e., "Tail" in at least one of the K samples);
• removes tail mutations, and run a multivariate Binomial read-counts analysis on all samples together by using with my other tool (VIBER).

This is what we do for the CRC multi-region case study in the paper - you can have a look at the CRC analysis in this repo

https://github.com/caravagn/mobster_supp_data

PS - VAF >0 is not a reasonable filter; the Pareto density requires a sharp cut, I'd suggest VAF > 5%.

from mobster.

Just to be even more clear, the clone tree construction is to be done AFTER the multivariate analysis (ie., from all clusters found in all biopsies).

from mobster.

Hi Giulio,

thanks for the reply.

I didn't see that supp_data repo before, very nice.

In the multi-region CRC analysis, you run two functions that are not in namespace of any libraries you require:

mobster_fits = fit_mobsters(Set7_mutations, Set7_samples)
...
non_tail_mutations = get_nontail_mutations(mutations = Set7_mutations, mobster_fits)

Therefore how do I find non-tail per sample?

Data in that repo is combined from the start, I presume I need to make a single dataset of non-tail mutations to VIBER::variational_fit(?)

Then from that output, what is input to ctree::get_clone_trees (no examples of this in any repos)?

Thanks,

Bruce.

from mobster.

Hi Bruce,

the repo should be attached as SM to the main paper, but is does not matter how you found it. If you see this you will also see

# Source a bunch of auxiliary functions
source('auxiliary_functions.R', verbose = FALSE)

which means that the functions are defined here.

So if you prepare your data in the exact same format then you can use exactly those ones. I suggest you to run it line by line if you do this analysis for the first time, so you understand exactly what is happing there, and how you get to run VIBER as well.

Cheers

G

from mobster.