De novo sequence-based method for ncRPI prediction using structural information

Improving knowledge of RNA-binding protein targets is focusing the attention towards non-coding RNAs (ncRNAs), i.e., transcripts not translated into a protein; they are associated with a wide range of biological functions through different molecular mechanisms, usually concerning the interaction with one or more protein partners. Recent studies confirmed that the alteration of ncRNA-protein interactions (ncRPIs) may be linked to various pathologies, including autoimmune and metabolic diseases, neurological and muscular disorders and cancer. Unfortunately, the limited number of structurally characterized RNA-protein complexes available does not allow to accurately establish their role in cellular processes and diseases. Experimental analyses to identify ncRNA-protein interactions are providing a large amount of valuable data, but these experiments are expensive and time consuming. For these reasons, computational approaches based on machine learning techniques appear very useful to predict ncRPIs. Yet, there are still few studies regarding the prediction of ncRPIs, especially including the use of higher-order structures, which are of vital importance for the ncRPI functions. In this work, a new computational method for non-coding RNA-protein interaction prediction is developed; from sequence data it derives more accurate information about the secondary structure of the molecules involved in such interactions, which it then uses in the prediction. Obtained results suggest that the use of machine learning techniques, together with considering also information on higher-order structures of ncRNAs and proteins, can be useful to better predict ncRPIs.