An Algorithm with Base-Pair Resolution for Identifying Cancer Heterogeneity by Estimating Multiple Clonal Haplotypes

An algorithm for identifying haplotype heterogeneity in cancer genomes is proposed to consider somatic mutational events carried by multiple sub-clones. The algorithm is based on the genomic sequencing data with multiple libraries of tumor tissue and extracts the features from both the multi-library and the constraints of paired-end reads. A priori number of sub-clones is roughly estimated by clustering the allelic variant frequency of each somatic loci. A contig-and-extension algorithm is designed, and the haplotype sequences are assembled by traversing the reads mapping to the loci. Thus, the contigs present an identification resolution on base-pair level. The number and proportion of sub-clones and the evolution relationships among them are further estimated by maximizing the likelihood of the posterior probabilities. Simulation results show that the algorithm reaches 99% in accuracy when the sequencing based library satisfies some coverage. The proposed algorithm outperforms the existing two-stage pipeline, which is widely used in data analysis now.