Table 2.
Type classification tools.
| Method | Source | Year | Algorithm | Category | Feature | Subclass | Citation |
|---|---|---|---|---|---|---|---|
| ProtLock | - | 1997 | Distance-based algorithms | Statistics | ACC | Yes | [131] |
| MemType-2 L | http://www.csbio.sjtu.edu.cn/bioinf/MemType | 2007 | Ensemble of the optimized evidence-theoretic KNN algorithms | ML | Evolution, Pse‑PSSM | Yes | [130] |
| Ali’s work | - | 2015 | Multiple-stage training with KNN, PNN, SVM, Naïve bayes multinomial, and voting feature interval | ML | Pse-AAC | Yes | [138] |
| Butt’s work | - | 2016 | Multilayer neural networks | ML | Feature extraction with statistical moments | No | [134] |
| MBPpred | http://bioinformatics.biol.uoa.gr/MBPpred | 2016 | Profile hidden Markov models | ML | Membrane binding domains | No | [139] |
| Guo’s work1 | - | 2017 | Stacked generalization (ensemble) of SVM, KNN, RFs, neural networks, multiple logistic regression |
ML | Pse-AAC | Yes | [140] |
| iMem-2LSAAC | - | 2018 | Multiple-stage training with the KNN, PNN, SVM, generalize regression neural network, and random forest algorithms | ML | Pse-AAC | Yes | [132] |
| MKSVM-HSIC | https://github.com/hzwh6910/Identification-of-Membrane-Protein-Types-via-Multivariate-Information-Fusion-with-Hilbert-Schmidt | 2019 | Multiple kernel SVM | ML | Pse-PSSM | Yes | [141] |
| Guo’s work2 | https://github.com/DragonKnightss/MembraneProteinTypePrediction | 2019 | Convolutional and bidirectional long short term-memory neural networks |
DL | Pse-PSSM and other PSSM-based features | Yes | [142] |
| TooT-M | https://github.com/bioinformatics-group/TooT-M | 2020 | Selective voting ensemble classifiers | ML | Pse‑PSSM, Pse-AAC | No | [133] |
| Zhang’s work | - | 2021 | SVM, RF, simple logistic, Naive bayes, nearest neighbors, and decision trees | ML | PseAAC | Yes | [143] |
| iMPT-FDNPL | https://github.com/mufei111/iMPT-FDNPL | 2021 | word2vector, random k-labelsets ensemble (RAkEL), and RFs | ML | Sequence | Yes | [144] |
Note: SVM, support vector machine; KNN, K-nearest neighbour; RF, random forest; PNN, probabilistic neural network; Pse‑PSSM, pseudo position‑specific scoring matrix; Pse-AAC, pseudo amino acid composition. The subclass column represents whether the listed programs can be used to classify the subclasses of membrane proteins, including type I, type II, type III, type IV, multi-pass, lipid-chain-anchored, GPI-anchored, and/or peripheral proteins. If no, the program(s) can only be used to distinguish between membrane and globular proteins.