Phylogeny on merged samples

[freddie090]

Hi,

I have multiple samples from an experiment that have ~6000 cells of 10X scRNA data each. If I were to try and run NUMBAT on the entire merged experiment the BAMs would be too big.

Is it possible to run NUMBAT on the independent samples, but then merge the samples for the phylogeny part of the analysis? (for example, as was done in figure 5a of the NUMBAT paper).

Additionally, if I were to subset the BAMs given some high quality cells, merge the subsetted samples and then run NUMBAT on these merged BAMs/expression matrices, would this improve the robustness of the analysis? ie is there any advantage to the samples being processed simultaneously vs independently?

Thanks

[teng-gao]

Hi @freddie090 ,

You can genotype the samples (from the same individual) using the multi-sample mode of pileup_and_phase (you can provide a list of BAMs), and provide the combined count_mat and alelle_df in a single numbat run. Numbat should be able to handle 6000 cells fine.

The advantage is that you get consistent CNV and clone calls across samples and get an integrated phylogeny. Genotyping using multiple samples can also improve phasing accuracy.

Best,
Teng

[freddie090]

Hi Teng,

Ah great, okay - my hunch was the inference would be more robust if it had access to information from all samples at once. I'll give it a go!

Thanks -
Freddie

[freddie090]

Hi @teng-gao - sorry, just to clarify:

After running pileup and phase where I provide a list of BAM files and corresponding sample names (as comma separated values as a single argument, e.g.: -- samples samp_1,samp_2,samp_3 \ --bams samp_1.bam,samp_2.bam,samp_3.bam the script produces an 'allele_counts.tsv' file for each sample.

Has the multi sample mode worked? I wasn't sure whether I should expect a single combined allele counts table for all samples. If not, then do you suggest manually merging the expression matrix and allele counts for each sample before running Numbat?

Best
Freddie

[teng-gao]

Yes you should get a separate allele count df for each sample. You can then concatenate them (ditto for expression count matrix) before feeding to run_numbat.

[freddie090]

Okay - and sorry final Q @teng-gao - are the sample identities preserved somewhere for distinguishing in the phylogeny plots later?

[teng-gao]

Okay - and sorry final Q @teng-gao - are the sample identities preserved somewhere for distinguishing in the phylogeny plots later?

You can plot the sample identities associated with cell barcodes on a sidebar using the annot = option in plot_phylo_heatmap: