Table 2.
Overview of datasets used by methods considered in this review. In bold are the datasets selected for the benchmark
| Dataset | Year | First introduced by | Description and numbers | Ref |
|---|---|---|---|---|
| RNAcompete | 2013 | RNAcompete | 207 RBPs, across 231 experiments. Total of 241 000 sequences each 38–41 nt length, split into 2 sets: set A (120,326) and set B (121,031). Each sequence has a score from the log odds ratio of intensities of pull-down vs input. |
[44] |
| RBP-24 | 2014 | GraphProt | Sites derived from CLIP-seq experiments. 24 experiments (23 from Dorina) for 21 RBPs. 16 PAR-CLIP, 4 HITS-CLIP, 4 iCLIP. |
[20] |
| AURA2 | 2014 | AURA2 | Database of post-transcriptional regulatory interactions in UTR, including binding sites of RBPs. |
[84] |
| RBP-31 | 2016 | iONMF | Mix of PAR-CLIP, iCLIP and HITS-CLIP. 31 CLIP experiments for 19 RBPs. Positive controls: derived from positions with high read-counts. Negative controls: positions sampled from genes without interactions from any of the 19 RBPs. |
[61] |
| RBP-67 | 2016 | RNAcommender | 67 distinct RBPs, 72 226 UTRs. Total of 502 178 interactions curated from the AURA2 database. |
[85] |
| Mukherjee_PARCLIP | 2019 | DeepRiPe | PAR-CLIP experiments for 59 RBPs profiled in the HEK293 cell line. |
[39] |
| VanNostrand_ENCODE | 2020 | Van Nostrand et al. | 150 RBPs assessed in two cell-types (HepG2, K562), for a total of 223 experiments. |
[7] |