Summary

nowellpack is a Blang library for cancer genomics. The main current features focus on Bayesian phylogenetic tree inference from single cell data. We make the approximation that change points in the copy number profiles are perfect phylogeny markers. However those markers are observed in a noisy fashion hence the name we give to this model: corrupt phylogenies.

Tree inference

Run an end-to-end pipeline (sampling, summarizing posterior, viz) with:

git clone --depth=1 https://github.com/UBC-Stat-ML/corrupt-nextflow.git
cd corrupt-nextflow
./nextflow run main.nf -resume --tipInclusionProbabilities DATA.csv

Change DATA.csv into the data you are interested in (more below). Provided you have Oracle Java 8 in your PATH variable this will create a directory called deliverables/main containing an inferred tree and a bunch of other outputs.

Input format

The input for all included phylogenetic inference methods is a csv file with the following header:

cells,loci,tipInclusionProbabilities

followed by lines of the form

myfirstcell,somelocus,1
...

Note that for the binary model, tipInclusionProbabilities are set to 0 and 1, and a false positive/false negative rates model is inferred jointly.

The loci should follow the format 1_100500001_101000000 where 1 here is the chromosome index, 100500001 is the left boundary of the bin, and 101000000 is the right boundary, both inclusive.

Cells can use arbitrary unique identifiers.

Tree growing

Let us say you have a pre-computed tree inferred from a cell/loci matrix M and you would like to either:

• add more cells that were measured with the same set of loci as M
• add more loci that were measured with the same set of cells as M

In such cases you can use the corrupt-grow utility to quickly grow the tree by using the extra data. This is done using maximum a posteriori placement which has the advantage of being very fast but does not provide measures of uncertainty.

To use the corrupt-grow utility:

1. Clone the repo
2. Build using ./setup-cli.sh
3. Add build/install/nowellpack/bin to your path
4. Invoke with corrupt-grow --matrix ReadOnlyCLMatrix TIDY_FILE --phylo NEWICK_FILE where NEWICK_FILE is the pre-computed phylogeny in newick format, and TIDY_FILE is a matrix in the format

cells,loci,tipInclusionProbabilities

followed by lines of the form

myfirstcell,somelocus,0.123
...

where the entries are the "local posterior distributions" described in the preprint. Alternatively, you can use corrupt-grow --matrix NoisyBinaryCLMatrix --matrix.binaryMatrix TIDY_FILE --matrix.fpr 0.1 --matrix.fnr 0.2 --phylo NEWICK_FILE for a binary matrix TIDY_FILE with prescribed false positive and false negative rates.