MINTIE: Identifying cryptic variants in cancer transcriptomes using RNA-seq data

Short talk at Victorian Cancer Bioinformatics Symposium 2019.

Transcript

1. @ Murdoch Children’s Research Institute, 2019 Identifying cryptic variants in cancer transcriptomes using RNA-seq data 16.08.2019 Victorian Cancer Bioinformatics Symposium Marek Cmero
2. 2 Motivation Image 1: Szczepański, T., Harrison, C. J., & van Dongen, J. J. M. (2010). Genetic aberrations in paediatric acute leukaemias and implications for management of patients. The Lancet Oncology, 11(9), 880–889. https://doi.org/10.1016/S1470-2045(09)70369-9 Image 2: Tsapogas, P., Mooney, C. J., Brown, G., & Rolink, A. (2017). The cytokine Flt3-ligand in normal and malignant hematopoiesis. International Journal of Molecular Sciences, 18(6). https://doi.org/10.3390/ijms18061115 - Gene fusions and transcriptomic variants can modify gene function in cancer, e.g.: - BCR-ABL1 fusion - FLT3 internal tandem duplication - RNA-seq can effectively identify and characterise gene fusions - Beyond fusions, other variants are difficult to detect: - Non-canonical fusions - Transcribed structural variants - Novel splice variants - No method exists to detect, annotate and visualise all types of cryptic variants in RNA- seq data cryptic variants
3. 3 Fusions Detected by fusion callers Not detected by fusion callers Typically detected in RNA
4. 4 Transcribed structural variants (TSVs) Can be detected by specialised callers (or novel intron) (SVs) typically detected in DNA, harder to detect in RNA
5. 5 Novel splice variants Not present in the DNA Detected by transcript assemblers
6. 6 Complex/combined variants • Difficult to detect for fusion callers • May be detected by some TSV callers
8. 8 pipeline Basic idea Find and annotate transcripts containing cryptic variants - Assemble transcripts in case sample - Not biased by reference genome - Quantify assembled transcripts using fast pseudo- alignment - Interested in novel contigs not present in controls (rare variants) - Perform DE on assembled transcripts - 1 case vs. N controls - Identify up-regulated novel transcripts - Align DE contigs to genome - Annotate variants not matching reference - Visualise variants
9. 9 Applying MINTIE to 1500 simulated cryptic variants (100 per category)
10. 10 Comparison of cryptic variants detected using existing tools (targeted)
11. 11 Cryptic variants called in a real B-ALL sample Assemble Annotate DE N = 584146 (contigs) Filter Merge N = 22567 (variants) N = 278 (variants) N = 176 (variants) N = 108 (variants)
12. 12 RB1 unpartnered fusion – genome view (RNA-seq) RNA-seq coverage (case) Fusion contig Putative deletion Fusion boundaries Novel contigs
13. 13 Another example in a different sample IKZF1 partial tandem duplication RNA-seq coverage (case) Novel contigs Variant contig
14. 14 Summary • MINTIE Detects all kinds of cryptic variants in RNA-seq cancer samples: • Canonical, non-canonical and unpartnered fusions • Novel splice variants • Transcribed structural variants • Method • De novo assemble > quantify > DE > annotate > visualise • Detects more variants than any other tool • Detected RB1 unpartnered fusion and IKZF1 PTD in B-ALL samples • We are hard at work on the visualisation component!
15. Acknowledgements MCRI Bioinformatics • Alicia Oshlack* • Nadia Davidson* • Breon Schmidt • + Whole team! MCRI Cell biology • Paul Ekert WEHI • Ian Majewski https://github.com/Oshlack/MINTIE @marekcmero *supervised this work equally