Prediction of protein structural class based on symmetrical recurrence quantification analysis

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Biology

Background:

• Protein structural class prediction is vital for understanding protein function and facilitating drug design.
• Predicting structural class for sequences with low similarity presents a significant computational challenge.
• Existing methods often struggle with accuracy and generalization for diverse protein families.

Purpose of the Study:

• To develop and evaluate a novel computational method for predicting protein structural class, particularly for low similarity sequences.
• To assess the efficacy of symmetrical recurrence quantification analysis as a feature extraction technique.
• To compare the performance of different machine learning classifiers for this prediction task.

Main Methods:

• Protein sequences were transformed into DNA and subsequently into binary sequences.
• Symmetrical recurrence quantification analysis was applied to extract 8 features per symmetry plot.
• Machine learning algorithms including Linear Discriminant Analysis (LDA), Random Forest (RF), and Support Vector Machine (SVM) were employed and compared.

Main Results:

• The proposed method, combining sequence transformation and symmetrical recurrence quantification analysis, demonstrated high predictive performance.
• The Random Forest classifier, when used with the extracted features, achieved an outstanding overall accuracy of 100%.
• The approach proved effective in avoiding overfitting, indicating robust generalization capabilities.

Conclusions:

• Symmetrical recurrence quantification analysis is a powerful feature extraction method for protein sequence data.
• The Random Forest classifier, coupled with this feature extraction technique, offers a superior solution for protein structural class prediction.
• This method holds significant promise for advancing applications in drug design and protein functional analysis.