Deconvolution and Phylogeny Inference of Diverse Variant Types Integrating Bulk DNA-seq with Single-cell RNA-seq.

Motivation: Reconstructing clonal lineage trees ("tumor phylogenetics") has become a core tool of cancer genomics. Earlier approaches based on bulk DNA sequencing (DNA-seq) have largely given way to single-cell DNA-seq (scDNA-seq), which offers far greater resolution for clonal substructure. Available data has lagged behind computational theory, though. While single-cell RNA-seq (scRNA-seq) has become widely available, scDNA-seq is still sufficiently costly and technically challenging to preclude routine use on large cohorts. This forces difficult tradeoffs between the limited genome coverage of scRNA-seq, limited availability of scDNA-seq, and limited clonal resolution of bulk DNA-seq. These limitations are especially problematic for studying structural variations and focal copy number variations that are crucial to cancer progression but difficult to observe in RNA-seq. Results: We develop a method, TUSV-int, combining advantages of these various genomic technologies by integrating bulk DNA-seq and scRNA-seq data into a single deconvolution and phylogenetic inference computation while allowing for single nucleotide variant (SNV), copy number alteration (CNA) and structural variant (SV) data. We accomplish this by using integer linear programming (ILP) to deconvolve heterogeneous variant types and resolve them into a clonal lineage tree. We demonstrate improved deconvolution performance over comparative methods lacking scRNA-seq data or using more limited variant types. We further demonstrate the power of the method to better resolve clonal structure and mutational histories through application to a previously published DNA-seq/scRNA-seq breast cancer data set. Availability: The source code for TUSV-int is available at https://github.com/CMUSchwartzLab/TUSV-INT.git.