A novel artificial neural network method for biomedical prediction based on matrix pseudo-inversion.

Area of Science:

• Biomedical informatics
• Computational biology
• Machine learning in healthcare

Background:

• Biomedical prediction using clinical and genome-wide data is crucial for disease diagnosis and classification.
• Existing methods may require lengthy iterations or lack optimal accuracy.
• There is a need for efficient and accurate predictive models in clinical care.

Purpose of the Study:

• To develop a novel Artificial Neural Network (ANN) method based on Matrix Pseudo-Inversion (MPI) for biomedical prediction.
• To evaluate the performance of the MPI-ANN method against established techniques like LASSO, SVM, and logistic regression.
• To assess the utility of MPI-ANN for disease classification, prediction, and biomarker selection.

Main Methods:

• A three-layer feed-forward Artificial Neural Network (ANN) was constructed.
• Weights connecting hidden and output layers were directly determined using Matrix Pseudo-Inversion (MPI).
• The MPI-ANN method was validated using simulated Single Nucleotide Polymorphism (SNP) data and real breast cancer data with 5-fold cross-validation.

Main Results:

• The MPI-ANN method demonstrated significantly superior accuracy in disease classification and prediction compared to the LASSO method.
• On real breast cancer data, MPI-ANN outperformed other machine learning methods, including support vector machine (SVM), logistic regression (LR), and iterative ANNs.
• The study confirmed the potential of MPI-ANN for effective biomarker selection.

Conclusions:

• The developed MPI-ANN method offers an effective and accurate approach for biomedical prediction and disease classification.
• MPI-ANN presents a significant advancement over existing machine learning techniques in terms of predictive accuracy.
• The method shows promise for clinical applications, including biomarker discovery and improved patient care.