Predicting ATP-Binding Cassette Transporters Using the Random Forest Method.

Area of Science:

• Biochemistry
• Computational Biology
• Molecular Biology

Background:

• ATP-binding cassette (ABC) proteins are essential across species, involved in nutrient uptake, toxin export, and drug resistance.
• Accurate identification of ABC transporters is critical for understanding cellular functions and developing therapeutic strategies.

Purpose of the Study:

• To develop and validate an efficient computational model for identifying ATP-binding cassette (ABC) transporters.
• To leverage sequence information and physicochemical properties for feature extraction and classification.

Main Methods:

• Utilized the 188D method for feature extraction based on protein sequence information and physicochemical properties.
• Employed t-Distributed Stochastic Neighbor Embedding (t-SNE) for feature visualization.
• Applied random forest (RF) as a machine learning classifier for protein identification.

Main Results:

• The 188D feature extraction method showed potential for sample separation.
• The random forest model achieved an average accuracy of 89.54% on the training set with 10-fold cross-validation.
• Achieved high performance metrics: 0.87 specificity, 0.92 sensitivity, and 0.79 MCC on the training set.
• The model demonstrated 89% accuracy on an independent testing set.

Conclusions:

• The combined approach of 188D feature extraction and random forest classification provides an optimal computational tool for identifying ABC transporters.
• This model offers a promising method for accelerating the discovery and characterization of ABC proteins in various biological contexts.