Prediction model of sinoatrial node field potential using high order partial least squares

Area of Science:

• Biomedical Engineering
• Computational Biology
• Data Science

Background:

• Predictive modeling for biological systems is crucial for understanding complex physiological processes.
• High glucose levels significantly impact cellular functions, including electrical activity in tissues like the sinoatrial node.
• Existing methods may have limitations in handling multi-dimensional biological data.

Purpose of the Study:

• To develop a predictive model for the relationship between sinoatrial node field potential and high glucose.
• To evaluate the efficacy of High Order Partial Least Squares (HOPLS) for this complex biological modeling task.

Main Methods:

• Utilized High Order Partial Least Squares (HOPLS), a novel tensor-based data processing technique.
• Employed three sub-signals of the sinoatrial node field potential as model inputs.
• Defined high glucose concentration and actuation duration as model outputs.

Main Results:

• HOPLS successfully built a predictive model for sinoatrial node field potential and high glucose.
• The model demonstrated comparable predictive ability to traditional Partial Least Squares (PLS) for two-dimensional variables.
• HOPLS exhibited a lower dispersion degree compared to PLS, indicating improved model stability.

Conclusions:

• High Order Partial Least Squares (HOPLS) is a robust method for predicting biological system responses with tensor inputs and outputs.
• HOPLS offers a promising alternative to PLS for modeling the intricate relationship between sinoatrial node electrophysiology and hyperglycemia.
• The findings suggest HOPLS can enhance the accuracy and reliability of predictive models in biomedical research.