Intercellular heterogeneity serves as both a confounding factor in studying individual clones and an information source in characterizing any heterogeneous tissues, such as blood, tumor systems. Due to inevitable sequencing errors and other sample pr
eparation artifacts such as PCR errors, systematic efforts to characterize intercellular genomic heterogeneity must effectively distinguish genuine clonal sequences from fake derivatives. We developed a novel approach (SIGH) for identifying significant genuine clonal sequences directly from mixed sequencing reads that can improve genomic analyses in many biological contexts. This method offers several attractive features: (1) it automatically estimates the error rate from raw sequence reads and identifies genuine clonal sequences; (2) it is robust to the large variety of error rate due to the various experimental conditions; (3) it is supported by a well grounded statistical framework that exploits probabilistic characteristics of sequencing errors; (4) its unbiased strategy allows detecting rare clone(s) despite that clone relative abundance; and (5) it estimates constituent proportions in each sample. Extensive realistic simulation studies show that our method can reliably estimate the error rates and faithfully distinguish the genuine clones from fake derivatives, paving the way for follow up analysis that is otherwise ruined by the often dominant fake clones.
