Reliable detection of subclonal single-nucleotide variants in tumour cell populations

According to the clonal evolution model, tumour growth is driven by competing subclones in somatically evolving cancer cell populations, which gives rise to genetically heterogeneous tumours. Here we present a comparative targeted deep-sequencing approach combined with a customised statistical algorithm, called deepSNV, for detecting and quantifying subclonal single-nucleotide variants in mixed populations. We show in a rigorous experimental assessment that our approach is capable of detecting variants with frequencies as low as 1/10,000 alleles. In selected genomic loci of the TP53 and VHL genes isolated from matched tumour and normal samples of four renal cell carcinoma patients, we detect 24 variants at allele frequencies ranging from 0.0002 to 0.34. Moreover, we demonstrate how the allele frequencies of known single-nucleotide polymorphisms can be exploited to detect loss of heterozygosity. Our findings demonstrate that genomic diversity is common in renal cell carcinomas and provide quantitative evidence for the clonal evolution model. The detection of subclonal variants in heterogeneous cancer specimens is a challenge due to errors that occur during sequencing. In this study, a statistical algorithm and a sequencing strategy are reported that circumvent this issue and can accurately detect variants at a frequency as low as 1/10,000.