Computational approach to discriminate human and mouse sequences in patient-derived tumour xenografts

• Published in: BMC genomics 2018-01, Vol.19 (1), p.19-19, Article 19
• Main Authors: Callari, Maurizio, Batra, Ankita Sati, Batra, Rajbir Nath, Sammut, Stephen-John, Greenwood, Wendy, Clifford, Harry, Hercus, Colin, Chin, Suet-Feung, Bruna, Alejandra, Rueda, Oscar M, Caldas, Carlos
• Format: Article
• Language: English
• Subjects: Alignment; Animals; Breast Neoplasms - genetics; Breast Neoplasms - metabolism; DNA sequencing; Gene expression; Gene Expression Profiling; Genetic aspects; Genomics - methods; High throughput sequencing; High-Throughput Nucleotide Sequencing - methods; House mouse; Humans; ICRG; In silico combined human-mouse reference genome; Methodology; Mice; Mouse stroma; Mutation; Sequence Alignment; Sequence Analysis, DNA; Sequence Analysis, RNA; Short-reads; Tumors; Xenograft Model Antitumor Assays; Xenotransplantation
• ISSN: 1471-2164; 1471-2164
• DOI: 10.1186/s12864-017-4414-y

Patient-Derived Tumour Xenografts (PDTXs) have emerged as the pre-clinical models that best represent clinical tumour diversity and intra-tumour heterogeneity. The molecular characterization of PDTXs using High-Throughput Sequencing (HTS) is essential; however, the presence of mouse stroma is challenging for HTS data analysis. Indeed, the high homology between the two genomes results in a proportion of mouse reads being mapped as human. In this study we generated Whole Exome Sequencing (WES), Reduced Representation Bisulfite Sequencing (RRBS) and RNA sequencing (RNA-seq) data from samples with known mixtures of mouse and human DNA or RNA and from a cohort of human breast cancers and their derived PDTXs. We show that using an In silico Combined human-mouse Reference Genome (ICRG) for alignment discriminates between human and mouse reads with up to 99.9% accuracy and decreases the number of false positive somatic mutations caused by misalignment by >99.9%. We also derived a model to estimate the human DNA content in independent PDTX samples. For RNA-seq and RRBS data analysis, the use of the ICRG allows dissecting computationally the transcriptome and methylome of human tumour cells and mouse stroma. In a direct comparison with previously reported approaches, our method showed similar or higher accuracy while requiring significantly less computing time. The computational pipeline we describe here is a valuable tool for the molecular analysis of PDTXs as well as any other mixture of DNA or RNA species.