JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

A method combining LDA and neural networks for antitumor drug efficacy prediction

  • 0University of Science and Technology of China, Hefei, Anhui, China.
Digital Health +

|

September 11, 2024

|

September 11, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

Predicting antitumor drug efficacy for personalized cancer treatment is crucial. This study combines Latent Dirichlet Allocation (LDA) and neural networks to accurately forecast drug responses from clinical data, outperforming previous methods.

Area Of Science

  • Oncology
  • Bioinformatics
  • Computational Biology

Background

  • Personalized medicine and precision cancer treatment are increasingly important due to patient genetic diversity.
  • Predicting antitumor drug efficacy in advance remains a significant challenge for effective cancer therapy.

Purpose Of The Study

  • To develop a method for predicting antitumor drug efficacy in individual cancer patients using clinical data.
  • To leverage Latent Dirichlet Allocation (LDA) and neural networks for personalized drug response prediction.

Main Methods

  • Clinical text data from cancer patients was encoded into a probability distribution vector using the Latent Dirichlet Allocation (LDA) model.
  • A neural network was designed to predict drug response by inputting the LDA representation.
  • The model was evaluated on lung and bowel cancer patient data treated with platinum drugs using Precision, Recall, F1-score, Accuracy, and Area Under the ROC Curve (AUC).

Main Results

  • The proposed method achieved high performance metrics, including an average AUC of 0.81 for cisplatin efficacy prediction in non-small cell lung cancer patients.
  • Performance was validated on an independent dataset of 266 bowel cancer patients, demonstrating generalizability.
  • The approach showed superior efficacy prediction compared to previous methods, with AUC at least 4% higher.

Conclusions

  • Combining LDA and neural networks offers a promising approach for predicting antitumor drug efficacy from clinical text.
  • The developed method demonstrates potential for enhancing precise tumor treatment strategies in clinical practice.

Keywords: