Table 3. Comparative evaluation of APRICOT with other RBP prediction tools: catRAPID signature, SPOT-Seq-RNA and RNApred.
| RBP prediction tools | APRICOT | catRAPID signature | SPOT-Seq-RNA | RNApred |
|---|---|---|---|---|
| Main features of the tools | ||||
| Main criteria for RBP characterization | RNA-binding motifs and domain families | Physico–chemical properties | RBP structure homologs | SVM classification by composition features of proteins |
| Additional analysis | Sequence-based scoring of domain (includes physico–chemical properties) | Prediction of RNA-binding regions | RNA-binding residue prediction and binding affinity | PSSM based evolutionary information |
| Availability | Command-line and Docker image | Web server | Web server and command-line | Web server |
| Query types | Amino-acid sequences/gene names/UniProt protein/taxonomy ids | Amino-acid sequences | Amino-acid sequence | Amino-acid sequences |
| Allowed number of query proteins | Unlimited | 100 proteins or total number of submitted characters = 100 000 | One query at a time | Unlimited for composition or one query at a time for the PSSM based analysis |
| Probability scores for RBPs | Bayesian score (0-1), 1 = best score | SVM score (Threshold −0.2) | Z-score | SVM score (Threshold −0.2) |
| Main criteria for RBP characterization | RNA-binding motifs and domain families | Physico–chemical properties | RBP structure homologs | SVM classification by composition features of proteins |
| Performance assessment | ||||
| TP (proteins) | 193 | 125 | 166 | 180 |
| FP (proteins) | 44 | 150 | 6 | 102 |
| TPR (SN) | 0.79 | 0.51 | 0.67 | 0.73 |
| FPR (1-SP) | 0.12 | 0.42 | 0.02 | 0.29 |
| ACC | 0.83 | 0.54 | 0.83 | 0.72 |
| MCC | 0.66 | 0.1 | 0.69 | 0.44 |
| F-measure | 0.83 | 0.54 | 0.8 | 0.72 |
Different features of the tools are listed in the upper part of the table to highlight the main differences and advantages over other tools. The comparative performance of the tools, shown in the lower part of the table, was assessed using the positive set of RBscore_R246 containing 246 proteins (total proteins from RBscore_R130 and RBscore_R116) and the negative set of RNApred-negative containing 355 proteins. APRICOT achieved higher overall SN of 0.79 compared to the other tools.