Abstract 1517: Impact of poly-A and ribo-depletion RNA-seq library construction protocols on transcriptomic analysis of samples from patients with haematological malignancies

Abstract

Abstract RNA sequencing (RNA-seq) is a versatile tool for characterization and quantification of transcriptomes. It has become the main approach to elucidate the transcriptomic landscape in leukemia. However, the knowledge about the advantages and disadvantages of RNA-seq in clinical decision-making and the suitability of mainstream RNA-seq library preparation methods for leukemia research are still in their infancy. Here, we have generated and sequenced polyadenylation selection (PA) and ribo-depletion (RD) RNA-seq libraries from acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL) patient samples and compared their fusion-gene, differential expression and related pathways and variant discovery abilities. Our analysis demonstrated that the RD protocol provided, on average, 5% more uniquely mapped reads than the PA protocol (81.5% of reads). Overall, 60%, 35%, and 3% of the reads in the RD libraries mapped to exonic, intronic, and intergenic regions, respectively. The corresponding numbers in the PA libraries were 76%, 21%, 3%, suggesting that PA captures more mature RNAs than the RD preparation and provides a higher read coverage for exon regions. The RD preparation was also marginally better in removing ribosomal RNAs (rRNAs) than the PA protocol. Regarding transcript quantification, both RD and PA produced fairly equivalent measures of RNA abundance. The patient-matched PA and RD libraries showed a high correlation (r = 0.96) that was slightly less than that seen between technical replicates (r = 0.98) but slightly more than that seen between library-matched sample from patients with the same disease (r = 0.93), indicating that inter-patient variance is the main source of variance in RNA-seq studies. Interestingly, based on a receiver operating characteristic (ROC) analysis using known leukemia gene signatures or gene targets of 17 clinically used drugs in AML and ALL treatment, the PA is more optimal than the RD method for disease classification (area under curve 0.69 vs. 0.65) and personalized cancer therapy guidance (AUC of 0.73 vs, 0.71). Finally, both methods equally detected known fusion genes, which were supported with high number of reads on fusion junctions. However, in the case of lowly expressed fusion transcripts, only the RD method was able to detect fusions. In conclusion, both protocols produced consistence measures and were of similar usability. However, RD preparation captured more transcriptomic features and was more suitable for fusion-gene discovery, which may be of clinical relevance for sub-typing leukemia patients. The PA method on the other hand had higher overall sensitivity and specificity in gene expression analyses and was, based on our results, more suitable than RD for clinical classification and drug treatment guidance. Citation Format: Ashwini Kumar, Matti Kankainen, Alun Parsons, Olli Kallioniemi, Pirkko Mattila, Caroline Heckman. Impact of poly-A and ribo-depletion RNA-seq library construction protocols on transcriptomic analysis of samples from patients with haematological malignancies. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 1517.