Improving Analysis of Herbal Medicine Data with Deep Learning Techniques

Jenn Hoskins
23rd August, 2024

Improving Analysis of Herbal Medicine Data with Deep Learning Techniques

Image Source: Natural Science News, 2024

Key Findings

  • Researchers from Macao Polytechnic University developed HerbMet, an AI system to accurately identify Chinese herbal medicines using metabolomics data
  • HerbMet achieved 95.71% accuracy and 95.42% F1-score in distinguishing seven similar Panax ginseng species
  • After applying feature selection techniques, HerbMet reached 100% accuracy and F1-score for discriminating P. ginseng species
Chinese herbal medicines have been utilized for thousands of years to prevent and treat diseases. Accurate identification of these herbs is crucial, as their medicinal effects can vary significantly between species and varieties. Traditional methods for identifying these herbs often fall short due to the complexity and subtle differences between species. In response to this challenge, researchers from Macao Polytechnic University have developed HerbMet, a high-performance artificial intelligence (AI) system designed to accurately identify Chinese herbal medicines using metabolomics data[1]. Metabolomics is a promising approach for distinguishing herbs by analyzing their unique chemical fingerprints. However, current metabolomics data analysis and modeling face limitations such as small sample sizes, high dimensionality, and the risk of overfitting. Overfitting occurs when a model learns the training data too well, including noise and outliers, which reduces its ability to generalize to new data. This study aims to overcome these challenges with HerbMet, which employs advanced AI techniques to improve the accuracy and robustness of herbal identification. HerbMet uses a 1D-ResNet architecture to extract discriminative features from metabolomics data. ResNet, or Residual Network, is a type of deep learning model known for its ability to train very deep networks without the vanishing gradient problem, which often hampers the performance of deep learning models. Additionally, HerbMet incorporates a multilayer perceptron for classification tasks and features a double dropout regularization module to mitigate overfitting. Dropout is a technique where random neurons are ignored during training, which helps prevent overfitting by making the network less sensitive to the specific weights of individual neurons. The performance of HerbMet was evaluated against 10 commonly used machine learning and deep learning methods. The results were impressive, with HerbMet achieving an accuracy of 95.71% and an F1-score of 95.42% for distinguishing seven similar Panax ginseng species. This is particularly noteworthy given the historical significance and medicinal use of Panax species, such as P. ginseng, P. quinquefolius, and P. notoginseng, which have been well-documented for their therapeutic properties[2]. The study further demonstrated that after applying feature selection techniques and incorporating prior knowledge, HerbMet achieved 100% accuracy and F1-score for discriminating P. ginseng species. The success of HerbMet highlights the potential of AI in advancing the field of traditional Chinese medicine (TCM). Previous studies have shown the therapeutic potential of natural products in TCM for treating inflammatory diseases and cancer by leveraging modern technologies like single-cell multiomics and network pharmacology[3]. HerbMet's ability to accurately identify herbal species can facilitate more precise use of these natural products, ensuring that the correct species with the desired therapeutic properties are used in treatments. Moreover, the study on ginsenosides, the main bioactive ingredients of Panax species, underscores the importance of accurate species identification. Ginsenosides have been found to possess antitumor activities by regulating cancer cell metabolism, among other beneficial effects[4]. By accurately identifying the specific Panax species, HerbMet can help ensure that the correct type of ginseng is used, thereby maximizing the therapeutic benefits. One of the key advantages of HerbMet is its user-friendliness in real-world scenarios. Unlike classical machine learning-based methods that require extensive feature ranking and selection, HerbMet simplifies the process, making it more accessible for practical applications. This ease of use, combined with its superior accuracy and robustness, positions HerbMet as a valuable tool for the identification of Chinese herbal medicines. In conclusion, HerbMet represents a significant advancement in the field of TCM by leveraging AI to address the challenges of herbal identification. Its development aligns with the ongoing efforts to modernize and internationalize traditional remedies, ensuring their accurate and effective use in contemporary medicine.

HerbsMedicineBiotech

References

Main Study

1) HerbMet: Enhancing metabolomics data analysis for accurate identification of Chinese herbal medicines using deep learning.

Published 21st August, 2024

https://doi.org/10.1002/pca.3437

Related Studies

2) Advances and challenges in ginseng research from 2011 to 2020: the phytochemistry, quality control, metabolism, and biosynthesis.

https://doi.org/10.1039/d1np00071c

3) New opportunities and challenges of natural products research: When target identification meets single-cell multiomics.

https://doi.org/10.1016/j.apsb.2022.08.022

4) Ginsenosides in cancer: A focus on the regulation of cell metabolism.

https://doi.org/10.1016/j.biopha.2022.113756


Related Articles

An unhandled error has occurred. Reload 🗙