Machine learning-based monosaccharide profiling for tissue-specific classification of Wolfiporia extensa samples.

Area of Science:

• Bioinformatics
• Mycology
• Computational Biology

Background:

• Machine learning (ML) is increasingly applied in bioinformatics for clinical decisions and diagnostics.
• Accurate identification of fungal tissue types is crucial for research and application.

Purpose of the Study:

• To evaluate the efficacy of eight machine learning algorithms for classifying four tissue types of Wolfiporia extensa based on monosaccharide composition.
• To compare the performance of ML models against traditional methods like linear discriminant analysis (LDA).

Main Methods:

• Eight machine learning algorithms were tested: linear discriminant analysis (LDA), logistic regression (LR), k-nearest neighbor (KNN), random forest (RF), gradient boosting machine (GBM), support vector machine (SVM), Naïve Bayes classifier (NB), and artificial neural network (ANN).
• Monosaccharide composition profiles of Wolfiporia extensa tissues were used as input features.
• Classification and prediction capabilities were assessed using Area Under the Curve (AUC) metrics.

Main Results:

• All eight ML models demonstrated exemplary performance with AUC > 0.8 for tissue classification.
• Five models (LDA, KNN, RF, GBM, ANN) achieved excellent classification accuracy (AUC > 0.9) for four tissue types.
• All eight models showed good predictive performance (AUC > 0.8) for three tissue types.
• ML-based methods significantly outperformed the traditional LDA plotting method.

Conclusions:

• Machine learning algorithms provide robust and accurate methods for classifying Wolfiporia extensa tissue types using monosaccharide composition.
• ML approaches offer superior performance compared to traditional regression techniques, especially for large datasets.
• These findings highlight the potential of ML to enhance the accuracy of fungal tissue identification in bioinformatics.