An optimized method for dose-effect prediction of traditional Chinese medicine based on 1D-ResCNN-PLS

Area of Science:

• Computational chemistry
• Pharmacometrics
• Machine learning

Background:

• Traditional Chinese medicine (TCM) dose-effect data often exhibits complex covariance and nonlinearity, challenging conventional analytical models.
• Accurate modeling of these relationships is crucial for understanding drug efficacy and optimizing treatment protocols.

Purpose of the Study:

• To develop and validate a novel computational model, the one-dimensional residual convolutional neural network with Partial Least Squares (1D-ResCNN-PLS), for analyzing TCM dose-effect data.
• To address the limitations of traditional methods in handling nonlinearities and covariances inherent in TCM pharmacological data.

Main Methods:

• Implementation of a 1D convolutional neural network integrated with residual blocks to capture intricate nonlinear features within the data.
• Integration of Partial Least Squares (PLS) regression for robust prediction, leveraging the feature extraction capabilities of the neural network.
• Validation of the 1D-ResCNN-PLS model using the Ma Xing Shi Gan Decoction dataset, comparing its performance against established conventional models.

Main Results:

• The 1D-ResCNN-PLS model demonstrated superior performance over conventional methods on the Ma Xing Shi Gan Decoction dataset.
• Achieved significantly high accuracy, sensitivity, specificity, and Area Under the Curve (AUC) values, indicating robust predictive power.
• Observed substantial reductions in mean square error (MSE), highlighting the model's efficiency in dose-effect relationship prediction.

Conclusions:

• The 1D-ResCNN-PLS model is highly effective for processing nonlinear data, particularly in the context of traditional Chinese medicine dose-effect relationships.
• The model's success suggests its potential for broader applications in analyzing complex biological and pharmacological datasets across various public domains.
• This computational approach offers a promising advancement for quantitative pharmacology and data-driven drug development.