Assessing the performance of methods for copy number aberration detection from single-cell DNA sequencing data

Abstract

Single-cell DNA sequencing technologies are enabling the study of mutations and their evolutionary trajectories in cancer. Somatic copy number aberrations (CNAs) have been implicated in the development and progression of various types of cancer. A wide array of methods for CNA detection has been either developed specifically for or adapted to singlecell DNA sequencing data. Understanding the strengths and limitations that are unique to each of these methods is very important for obtaining accurate copy number profiles from single-cell DNA sequencing data. We benchmarked three widely used methods-Ginkgo, HMMcopy, and CopyNumber-on simulated as well as real datasets. To facilitate this, we developed a novel simulator of single-cell genome evolution in the presence of CNAs. Furthermore, to assess performance on empirical data where the ground truth is unknown, we introduce a phylogeny-based measure for identifying potentially erroneous inferences. While single-cell DNA sequencing is very promising for elucidating and understanding CNAs, our findings show that even the best existing method does not exceed 80% accuracy. New methods that significantly improve upon the accuracy of these three methods are needed. Furthermore, with the large datasets being generated, the methods must be computationally efficient.

Methods for copy number aberration detection from single-cell DNA-sequencing data

“…Since HMM-based methods produce both a segmentation and absolute copy numbers at the same time, the selection of ploidy is essential in determining the copy number profile genome-wide. We observed that HMMcopy does not always select the right ploidy [54], leading to wrong copy number inference for the entire genome. Visualization of the CNA profiles of a group of cells is usually done by heatmaps, one example of which can be seen in [27].…”

“…One advantage of HMMcopy and AneuFinder is their ability to simultaneously infer segmentation and absolute copy numbers. As shown in [54], despite the possibility of selecting the wrong ploidy, HMMcopy achieved high precision and recall (around 0.8 for recall and 0.7 for precision) in both segmentation and copy number inference in a simulation study of single cells. Another advantage of HMMcopy is that its computational burden is relatively light.…”

“…1, the method applies binning (Y) or not (N); 2, the method applies GC correction (Y) or not (N); 3, the method applies mappability correction (Y) or not (N); 4, the method removes outlier bins (Y) or not (N); 5, the method removes outlier cells (Y) or not (N); 6, the method applies sliding window (1), objective function fitting (2), or HMM (3) for segmentation; and 7, the method calls the absolute copy number (Y) or not (N). "NA" denotes that the step is not applicable to the method 1 The model is automatically calibrated given the identified normal cells copy number estimation, as demonstrated in [54]. Furthermore, in the same study, it was observed that AneuFinder's recall dropped dramatically when the ploidy increased, and its precision was overall low for all ploidies (< 0.6).…”

“…The segmentation strategy is objective function based, and the dynamic programming algorithm implementation makes it computationally light. It takes ∼ 2.78 h and < 1 Gb memory to process 1000 cells at 0.04× coverage on a Intel(R) CPU E5-2650 v2, 2.60 GHz [54]. However, CopyNumber is limited in deciphering ITH because it does not pursue the absolute copy number inference after segmentation [54].…”

“…It takes ∼ 2.78 h and < 1 Gb memory to process 1000 cells at 0.04× coverage on a Intel(R) CPU E5-2650 v2, 2.60 GHz [54]. However, CopyNumber is limited in deciphering ITH because it does not pursue the absolute copy number inference after segmentation [54]. Neither does it identify which cell contains a breakpoint it infers.…”