# Compartment Model and Neural Network-Based Analysis of Combination Medication Ratios

## TL;DR

This paper introduces a new model combining compartment analysis and neural networks to optimize drug combination ratios for better treatment outcomes.

## Contribution

A novel approach using compartment models and neural networks to determine optimal pharmacodynamic component ratios in drug combinations.

## Key findings

- The model recalculates drug combination ratios using feedback adjustment in neural networks.
- Empirical validation showed reduced Combination Index (CI) after adjusting efficacy.
- The approach was tested on epilepsy and stroke treatment combinations.

## Abstract

Background: Combination medication strategies often involve complex interactions, making determining the appropriate pharmacodynamic component ratios challenging. Methods: This study established a time–dose relationship model through the compartment model, deriving the in vivo drug quantity ratios corresponding to the blood concentrations of the pharmacodynamic components. A neural network was then employed to establish a dose–effect relationship model between the blood concentrations of the pharmacodynamic components and the overall body response. Utilizing the feedback adjustment mechanism of neural networks continuously adjusts the network to achieve the desired drug efficacy, thereby deriving the corresponding dose ratio of the pharmacodynamic components. Empirical studies were conducted on combining Cynanchum otophyllum saponins M1 and M2 with phenobarbital for epilepsy treatment, as well as the anti-ischemic stroke activity of the prototype and metabolites of Erigeron breviscapus. Results: After adjusting the efficacy, the model recalculated the new ratio proportions for each combination, validated by the reduced Combination Index (CI). Conclusions: This model provides a new approach to combination medication strategies.