PHALCON: phylogeny-aware variant calling from large-scale single-cell panel sequencing datasets

Single-cell sequencing (SCS) technologies provide cellular resolution data for inferring variants and reconstructing tumor phylogeny for resolving intra-tumor heterogeneity (ITH), which causes drug resistance and relapse in cancer. Moreover, recently emerged panel sequencing methods can parallely sequence thousands of cells by targeting disease-specific genes. Current variant callers and SCS-specific phylogenetic methods fail to handle such large-scale datasets or address the amplification biases in panel-based sequencing protocols. Here, we present a statistical variant caller, PHAL-CON, which enables scalable mutation detection from large-scale single-cell panel sequencing data by modeling their evolutionary history under a finite-sites model along a clonal phylogeny. PHAL-CON infers the underlying cellular sub-populations based on mutation likelihoods of candidate sites and reconstructs a clonal phylogeny and the most likely mutation history using a likelihood-based probabilistic framework. Using numerous simulated datasets across varied experimental settings, we showed that PHALCON outperforms state-of-the-art methods in terms of variant call-ing accuracy (7.29-51.67% improvement), accuracy of tumor phylogeny inference (410.43-32931.8% improvement) and runtime (40-50 × faster). Furthermore, when applied on datasets from triple negative breast cancer patients, PHALCON detected novel somatic mutations with high functional impact in important oncogenes and tumor suppressor genes, and better resolved ITH by identifying rare clones, and elucidating lineages harboring deleterious mutations in key oncogenes and tumor suppressor genes.