AbstractSomatic copy number variations (CNVs) play a crucial role in development of many human cancers. The broad availability of next-generation sequencing data has enabled the development of algorithms to computationally infer CNV profiles from a variety of data types including exome and targeted sequence data; currently the most prevalent types of cancer genomics data. However, systemic evaluation and comparison of these tools remains challenging due to a lack of ground truth reference sets. To address this need, we have developed Bamgineer, a tool written in Python to introduce user-defined haplotype-phased allele-specific copy number events into an existing Binary Alignment Mapping (BAM) file, with a focus on targeted and exome sequencing experiments. As input, this tool requires a read alignment file (BAM format), lists of non-overlapping genome coordinates for introduction of gains and losses (bed file), and an optional file defining known haplotypes (vcf format). To improve runtime performance, Bamgineer introduces the desired CNVs in parallel using queuing and parallel processing on a local machine or on a high-performance computing cluster. As proof-of-principle, we applied Bamgineer to a single high-coverage (mean: 220X) exome sequence file from a blood sample to simulate copy number profiles of 3 exemplar tumors from each of 10 tumor types at 5 tumor cellularity levels (20-100%, 150 BAM files in total). To demonstrate feasibility beyond exome data, we introduced read alignments to a targeted 5-gene cell-free DNA sequencing library to simulate EGFR amplifications at frequencies consistent with circulating tumor DNA (10, 1, 0.1 and 0.01%) while retaining the multimodal insert size distribution of the original data. We expect Bamgineer to be of use for development and systematic benchmarking of CNV calling algorithms by users using locally-generated data for a variety of applications. The source code is freely available at http://github.com/pughlab/bamgineer.Author summaryWe present Bamgineer, a software program to introduce user-defined, haplotype-specific copy number variants (CNVs) at any frequency into standard Binary Alignment Mapping (BAM) files. Copy number gains are simulated by introducing new DNA sequencing read pairs sampled from existing reads and modified to contain SNPs of the haplotype of interest. This approach retains biases of the original data such as local coverage, strand bias, and insert size. Deletions are simulated by removing reads corresponding to one or both haplotypes. In our proof-of-principle study, we simulated copy number profiles from 10 cancer types at varying cellularity levels typically encountered in clinical samples. We also demonstrated introduction of low frequency CNVs into cell-free DNA sequencing data that retained the bimodal fragment size distribution characteristic of these data. Bamgineer is flexible and enables users to simulate CNVs that reflect characteristics of locally-generated sequence files and can be used for many applications including development and benchmarking of CNV inference tools for a variety of data types.
Bamgineer: Introduction of simulated allele-specific copy number variants into exome and targeted sequence data sets