Method Dataset Features Description
Pancaldi and Bähler [15] 5,166 mRNA-protein interacting pairs from immunopurification experiments Predicted protein secondary structure, localization, protein physical properties, gene physical properties, UTR properties, genetic interactions Protein and RNA sequences encoded using > 100 features are used to train SVM and RF classifiers
Bellucci et al. (catRAPID) [17] 7,409 interacting pairs from 858 RNA-protein complexes from PDB Physicochemical properties including secondary structure propensities, hydrogen-bonding propensities, and van der Waals interaction propensities Propensities are calculated for each amino acid and ribonucleotide to generate an interaction profile (   
Muppirala et al. (RPISeq) [22] 2,241 interacting pairs from 943 RNA-protein complexes from PRIDB (RPI2241) Sequence composition of proteins, represented as conjoint triads, and RNAs, represented as tetrads Protein and RNA sequences encoded sequence-composition-based features are used to train SVM and RF classifiers (
Wang et al. [26] RPI 2241 generated by Muppirala et al. & 367 interacting pairs from NPInter Sequence composition of protein and RNA Input to NB and ENB classifiers is a combination of protein triads and RNA triad features similar to those used in RPISeq
Table 1: Computational Methods for Predicting RNA-Protein Interaction Partners.