Table 2.
Summary of methods reviewed based on their input data type and approach.
| Method | Method | Input Data Type | Input Data Description | Performance (GO Biological Process Terms) | Limitations |
|---|---|---|---|---|---|
| isoPred | MIL with support vector machine (SVM) as a base learner | RNA-seq; GO | 19,209 genes and 24,274 mRNA isoforms from mouse | Area Under the Receiver Operating Curve (AUROC): 0.68–0.76 (multiple mRNA isoform genes) AUROC: 0.62–0.68 (single mRNA isoform genes) | Only RNA-Seq input; Random unannotated genes as negative set; no tissue, cell, sex, or age specificity |
| iMILP | MIL with label propagation | RNA-seq; GO | 31,454 human mRNAs | AUROC: 0.67 | Only RNA-Seq input; Genes annotated to sibling GO terms used as negative set; no tissue, cell, sex, or age specificity |
| IsoFunc | MIL with SVM as base learner | RNA-seq; GO | 11,946 genes and 59,297 mRNA isoforms from human | AUROC: 0.64 | Only RNA-Seq input; Random unannotated genes as negative set; no tissue, cell, sex, or age specificity |
| WLRM | MIL with weighted logistic regression | RNA-seq; GO | 11,946 genes and 59,297 mRNA isoforms from human | AUROC: 0.6–0.85 | Only RNA-Seq input; Random unannotated genes as negative set; no tissue, cell, sex, or age specificity |
| IIIDB | Network-based | RNA-seq; domain–domain interactions; GO; protein–protein interaction (PPI) | 31,454 mRNA isoforms from human | Data not available | Only RNA-Seq input; Subcellular localization as negative set; no tissue, cell, sex, or age specificity; limited to existing PPIs |
| Mouse Splice Isoform Network | Network-based; MIL with Bayesian network | RNA-Seq; Exon array; Protein docking; pseudo-amino acid composition; GO; Pathways | Data not available | AUROC: 0.62 | Random unannotated genes as negative set; no tissue, cell, sex, or age specificity |
| TENSION | Network- based; Random Forest | RNA-Seq; mRNA Sequence; Protein Sequence; PPI; GO; Pathways | 21,813 genes and 75,826 mRNA isoforms from mouse | AUROC: 0.94 | No cell, sex, or age specificity |
| DeepIsoFun | Deep learning | RNA-Seq; GO | 19,532 genes and 47,393 mRNA isoforms from human | AUROC: 0.74 | Only RNA-Seq input; Random unannotated genes as negative set; no tissue, cell, sex, or age specificity |
| DIFFUSE | Deep learning | RNA-Seq; mRNA sequence; Protein sequence; GO | 19,303 genes and 39,375 mRNA isoforms from human | AUROC: 0.84 | Random unannotated genes as negative set; no tissue, cell, sex, or age specificity |
| mFRecSys | Recommendation system | RNA-Seq; mRNA sequence; Protein sequence; PPI; GO; Pathways | 21,813 genes and 75,826 mRNA isoforms from mouse | AUROC: 0.99 | Limited tissue-specificity; No cell, sex, or age specificity |
| DisoFun | Recommendation system | RNA-Seq; PPI; GO | 11,868 genes and 25,939 mRNA isoforms from human | AUROC: 0.71 | Only RNA-Seq input; Random unannotated genes as negative set; no tissue, cell, sex, or age specificity |