Wavelet Based Whole Genome Doubling Aware Single Cell Copy Number Calling
Advances in single cell whole genome sequencing enable profiling of the copy number state of thousands of cells with minimal sequencing bias across the genome. The Direct Library Preparation + technique is an whole genome amplification-free single cell whole genome sequencing method that achieves high throughput by fragmenting each cells genome and ligating sequencing adapters using a modified Tn5 transposase, and sequencing to less than 0.1x coverage. Despite recent advances in experimental approaches, data analysis of single cell whole genome sequencing lags behind and the existing methods are not optimized for the analysis of frozen samples with variable DNA preservation. Furthermore, existing tools predominantly rely on read depth ratio in predefined genomic bins to call copy number, making whole genome duplication unidentifiable. To address this, we introduce Songbird, a single cell whole genome sequencing copy number caller that is whole genome duplication sensitive, and outperforms existing tools both in breakpoints identification and true copy number detection. We demonstrate that Songbird is robust down to extremely low coverage, adaptable to a variety of genome versions (hg19, hg38, hs.1), and is extensible to other single cell whole genome sequencing methods that rely on Tn5 tagmentation to fragment the genome.