Predicting Ion Channels Genes and Their Types With Machine Learning Techniques

Abstract

Motivation: The number of ion channels is increasing rapidly. As many of them are associated with diseases, they are the targets of more than 700 drugs. The discovery of new ion channels is facilitated by computational methods that predict ion channels and their types from protein sequences. Methods: We used the SVMProt and the k-skip-n-gram methods to extract the feature vectors of ion channels, and obtained 188- and 400-dimensional features, respectively. The 188- and 400-dimensional features were combined to obtain 588-dimensional features. We then employed the maximum-relevance-maximum-distance method to reduce the dimensions of the 588-dimensional features. Finally, the support vector machine and random forest methods were used to build the prediction models to evaluate the classification effect. Results: Different methods were employed to extract various feature vectors, and after effective dimensionality reduction, different classifiers were used to classify the ion channels. We extracted the ion channel data from the Universal Protein Resource (UniProt, http://www.uniprot.org/ ) and Ligand-Gated Ion Channel databases ( http://www.ebi.ac.uk/compneur-srv/LGICdb/LGICdb.php ), and then verified the performance of the classifiers after screening. The findings of this study could inform the research and development of drugs.

PSIONplusm Server for Accurate Multi-Label Prediction of Ion Channels and Their Types

“…Third, the channel subtypes should be predicted for the putative voltage-gated channels that were predicted in the second step. The third step is currently limited to the prediction of the subtypes for the voltage-gated channels because this is how the current predictors operate [ 13 , 19 , 20 , 21 , 22 , 23 ] ( Figure 1 B). The approach implemented by PSIONplus m extends the sequential prediction to cover the subtypes of the ligand-gated channels ( Figure 1 C).…”

“…Moreover, in contrast to the current tools that are limited to the prediction of a single channel subtype [ 13 , 19 , 20 , 21 , 22 , 23 ], PSIONplus m predicts multiple subtypes of the ion channels. This type of prediction can be accomplished in three ways: (1) by combining results produced by multiple current ion channel predictors; (2) by combining multiple new predictors where each individual new predictor targets a specific subtype of channels; and (3) by developing a new single multi-label predictor.…”

“…The assessment of predictive performance of the predictors of ion channels and their types and subtypes typically relies on several metrics that include accuracy, sensitivity (also called recall and true positive rate), precision and F 1 [ 13 , 19 , 20 , 21 , 22 , 23 ].…”

“…This predictor was recently upgraded to a new version, IonchanPred 2.0, which boasts improved predictive performance and which focuses on the same two types of ion channels and the four subtypes of the voltage-gated ion channels [ 20 ]. The other three predictors, which include the method by Tiwari and Srivastava [ 21 ], the method by Han et al [ 22 ], and PSIONplus [ 23 ], address the prediction of ion channels, their types (voltage- vs. ligand-gated) and the four subtypes of the voltage-gated channels (potassium, sodium, calcium, and anions). These predictors use sophisticated machine learning algorithms, such as support vector machines [ 18 , 19 , 20 , 22 , 23 ] and random forest [ 21 , 22 ], and secure relatively good predictive performance [ 13 ].…”

“…Moreover, the fourth problem of the current approaches, which was not covered in the survey, is that they do not consider the subtypes of the ligand-gated channels. They categorize the ion channels into two types: voltage- vs. ligand-gated, and further subdivide the voltage-gated channels into four subtypes: sodium, calcium, potassium and anion carrying [ 13 , 19 , 20 , 21 , 22 , 23 ]. However, they do not annotate the subtypes of the ligand-gated channels.…”