TelomereHunter - in silico estimation of telomere content and composition from cancer genomes

Abstract

Background: Establishment of telomere maintenance mechanisms is a universal step in tumor development to achieve replicative immortality. These processes leave molecular footprints in cancer genomes in the form of altered telomere content and aberrations in telomere composition. To retrieve these telomere characteristics from high-throughput sequencing data the available computational approaches need to be extended and optimized to fully exploit the information provided by large scale cancer genome data sets.

Results: We here present TelomereHunter, a software for the detailed characterization of telomere maintenance mechanism footprints in the genome. The tool is implemented for the analysis of large cancer genome cohorts and provides a variety of diagnostic diagrams as well as machine-readable output for subsequent analysis. A novel key feature is the extraction of singleton telomere variant repeats, which improves the identification and subclassification of the alternative lengthening of telomeres phenotype. We find that whole genome sequencing-derived telomere content estimates strongly correlate with telomere qPCR measurements (r = 0.94). For the first time, we determine the correlation of in silico telomere content quantification from whole genome sequencing and whole genome bisulfite sequencing data derived from the same tumor sample (r = 0.78). An analogous comparison of whole exome sequencing data and whole genome sequencing data measured slightly lower correlation (r = 0.79). However, this is considerably improved by normalization with matched controls (r = 0.91).

Conclusions: TelomereHunter provides new functionality for the analysis of the footprints of telomere maintenance mechanisms in cancer genomes. Besides whole genome sequencing, whole exome sequencing and whole genome bisulfite sequencing are suited for in silico telomere content quantification, especially if matched control samples are available. The software runs under a GPL license and is available at https://www.dkfz.de/en/applied-bioinformatics/telomerehunter/telomerehunter.html .

Fig. 1

TelomereHunter workflow. a TelomereHunter extracts reads containing telomere repeats from an input BAM file. The reads are sorted by mapping position into intrachromosomal, subtelomeric, junction spanning and intratelomeric reads. From the intratelomeric reads, telomere content, telomere variant repeats and singletons are obtained. b Examples of a typical intratelomeric read containing a TGAGGG singleton and a typical subtelomeric read containing multiple telomere variant repeats in arbitrary context

Fig. 2

Exemplary ALT-positive glioblastoma patient. a GC corrected telomere content: Telomere content of tumor and control sample in reads per million GC-matched reads. Contributions of TVRs are indicated by color code. b Telomere repeats in intratelomeric reads: Histogram of repeat units per read for all extracted intratelomeric reads. c Telomere variant repeats (arbitrary context): Overview on log₂ T/C of TVRs in arbitrary context. d Singleton telomere variant repeats: Overview on log₂ T/C of TVRs in singleton context

Fig. 3

Exemplary ALT-negative medulloblastoma patient. a GC corrected telomere context: Telomere content of tumor and control sample in reads per million GC-matched reads (RGCM). Contribution of TVR is indicated by color code. b Telomere repeats in intratelomeric reads: Histogram of repeat units per read for all extracted intratelomeric reads. c Telomere variant repeats (arbitrary context): Overview on log₂ T/C of TVRs in arbitrary context. d Singleton telomere variant repeats: Overview on log₂ T/C of TVRs in singleton context

Fig. 4

Telomere content estimation from WES data. a Correlation of telomere content log₂ T/C determined by telomere qPCR and TelomereHunter for WES samples aligned with bwa-mem from 49 leiomyosarcoma patients. b Correlation of qPCR and TelomereHunter results for individual tumor and control samples of the same patients (shown on a logarithmic scale). The Spearman correlation coefficients are indicated

Fig. 5

Telomere content estimation from WGBS data. Correlation of telomere content estimation from WGBS and WGS data of 34 medulloblastoma samples using TelomereHunter. The Spearman correlation coefficient is indicated