Artificial intelligence-guided precision treatment of chronic kidney disease-mineral bone disorder

Area of Science:

• Nephrology
• Pharmacology
• Computational Biology

Background:

• CKD-MBD is a serious complication of chronic kidney disease, impacting cardiovascular and skeletal health.
• Current treatments for CKD-MBD face challenges in achieving therapeutic goals and reducing patient risks.
• Effective management of CKD-MBD is crucial for improving patient outcomes and preventing severe complications.

Purpose of the Study:

• To investigate the application of quantitative systems pharmacology (QSP) modeling and artificial intelligence (AI) for improved CKD-MBD management.
• To develop and implement a reinforcement learning (RL) approach for optimizing treatment strategies in CKD-MBD.
• To enhance patient outcomes by achieving target levels of serum calcium, phosphorus, and parathyroid hormone.

Main Methods:

• Development of a QSP model integrated with AI techniques, specifically reinforcement learning (RL).
• Simulation of treatment strategies involving concurrent dosing of phosphate binders, vitamin D analogs, and calcimimetics.
• Testing the RL approach in a simulated cohort of 80 subjects using Matlab.

Main Results:

• The combined QSP and RL model demonstrated superior efficacy in achieving treatment goals compared to standard approaches.
• The system achieved therapeutic targets more quickly and effectively, even under conditions of simulated poor patient compliance.
• Key decision variables for therapeutic recommendations were identified through the in silico simulations.

Conclusions:

• QSP modeling coupled with RL offers a powerful tool for optimizing CKD-MBD treatment strategies.
• This computational approach can improve the management of CKD-MBD, leading to better patient outcomes.
• The identified decision variables can guide clinical recommendations for personalized CKD-MBD therapy.