PAR-CLIP is a high-throughput method to identify binding sites of RNA binding proteins. This is done by immunoprecipitation (IP) of the protein of interest and deep sequencing the RNA crosslinked to proteins prior to IP. PAR-CLIP also has extensively been used to determine target sites of microRNAs by isolating their binding protein AGO. In an AGO-PAR-CLIP experiment, the identity of the microRNA responsible for a target site is a priori not clear and must be revealed by matching the microRNA seed sequences to the target site sequence, which is not a trivial task. PAR-CLIP data has several specific characteristics, most notably, frequent T to C conversions that are indicative for crosslinking sites. We utilize these and other features to accurately determine the seed site.
The final PAR-CLIP model is in agreement with known binding mechanisms of microRNAs and with structural knowledge of AGO and many active k-mers correspond to seeds of expressed microRNAs. The final seed assignment has two properties: Each active seed sequence explains several clusters with high likelihood and the seed positions match the model of PAR-CLIP data learned from all target sites. Based on evaluations using differential PAR-CLIP data from both a publicly available dataset as well as from a new dataset, we show that PARma is more accurate than existing approaches in terms of correct seed assignments.Download Changelog Contact References