bioRxiv2018
MotivationRecently, next-generation sequencing techniques have been applied for the detection of RNA secondary structures called high-throughput RNA structural (HTS) analy- sis, and dozens of different protocols were used to detect comprehensive RNA structures at single-nucleotide resolution. However, the existing computational analyses heavily depend on experimental data generation methodology, which results in many difficulties associated with statistically sound comparisons or combining the results obtained using different HTS methods.\n\nResultsHere, we introduced a statistical framework, reactIDR, which is applicable to the experimental data obtained using multiple HTS methodologies, and it classifies the nucleotides into three structural categories, stem, loop, and unmapped. reactIDR uses the irreproducible discovery rate (IDR) with a hidden Markov model (HMM) to discriminate accurately between the true and spurious signals obtained in the replicated HTS experiments. In reactIDR, IDR and HMM parameters are efficiently optimized by using an expectation-maximization algorithm. Furthermore, if known reference structures are given, a supervised learning can be applicable in a semi-supervised manner. The results of our analyses for real HTS data showed that reactIDR achieved the highest accuracy in the classification problem of stem/loop structures of rRNA using both individual and integrated HTS datasets as well as the best correspondence with the three-dimensional structure. Because reactIDR is the first method to compare HTS datasets obtained from multiple sources in a single unified model, it has a great potential to increase the accuracy of RNA secondary structure prediction at transcriptome-wide level with further experiments performed.\n\nAvailabilityreactIDR is implemented in Python. Source code is publicly available at https://github.com/carushi/reactIDRhttps://github.com/carushi/reactIDR.\n\nContactkawaguchi-rs@aist.go.jp\n\nSupplementary informationSupplementary data are available at online.