Deep Sequencing Of Lung Cancer Samples Using Different Library Preparation Methods Produces Discordant Short Non-Coding Rna Profiles

Background: RNA-seq is the new standard for profiling coding and non-coding RNAs in a cell. This study compared two methods for miRNA library preparation from NEB and Qiagen. Our goal was to evaluate overall performance in terms of coverage, known miRNA detection, novel miRNA detection, isomiR detection, and tRNA fragment (tRF) detection. Methods: We performed short RNA-seq on 4 lung cancer tumor-normal pairs. We used two library preparation kits: NEB Next Small RNA kit and Qiagen Qiaseq. All 16 preparations were sequenced on an Illumina NextSeq 550 using NextSeq High V2 chemistry and 75 bp single-end reads were obtained. After adapter removal and quality trimming reads were mapped to the hg19 human genome assembly using the Bowtie2 and SHRiMP packages. Results: On average read yield was 46 million reads. A significantly higher portion of the sequenced reads survived adapter removal and quality trimming in the NEB preparations (96.5%) compared to the Qiagen (72.9%). The portion that could be uniquely mapped to hg19 was lower for the NEB kit (41.1%) than for the Qiagen (50.8%).NEB performed better (39.6%) compared with Qiagen (36.9%). Qiagen preparations had lengths between 20 and 25 bp whereas more than half of the mapped reads in the NEB preparation were 26 bp or longer. At the miRNA arm level the Qiagen kit detected significantly more known miRNAs (792) with 10 or more reads, compared with NEB (576). The same 530 miRNA arms were identified by both kits. There was no discernible pattern in the ID’s or the sequence composition of the miRNAs that were identified by each kit. The detection of novel miRNAs was also higher with Qiagen (102), compared with NEB (82). The same 79 novel miRNAs were identified by both. The Qiagen kit detected nearly twice as many isomiRs (5,316) with 10 or more reads compared to NEB (2,958). The same 2,631 isomiRs were identified by both kits. However, isomiR detection across samples was more consistent with the NEB kit. When we computed pairwise Pearson correlations of normal samples, using the most highly expressed miRNAs in each sample, the NEB kit exhibited higher consistency (0.98) compared with the Qiagen kit (0.95). Pearson correlations of tumor samples showed even higher consistency for NEB (0.96) than Qiagen (0.89). Unsurprisingly, Pearson correlations of like samples across the NEB and Qiagen kits was very low: 0.42 for normal and 0.56 for tumor samples. RIN value did not seem to affect the overall performance of either kit. Lastly, we compared tRFs. Here, the differences were very pronounced. For multiple choices of the support threshold the NEB and Qiagen profiles agreed on approximately 33% of the reported tRFs. Conclusions: Library preparation kits give rise to both consistent and divergent results. End users interested in the detection of miRNAs, isomiRs or tRFs may derive greater utility by selecting one kit over another. Citation Format: Brid M. Ryan, Phillipe Loher, Khadijah Mitchell, Adriana Zingone, Yongmei Zhao, Jyoti Shetty, Bao Tran, Isidore Rigoutsos. Deep sequencing of lung cancer samples using different library preparation methods produces discordant short non-coding RNA profiles [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 4423. doi:10.1158/1538-7445.AM2017-4423