HybridDLDR: A hybrid deep learning-based drug resistance prediction system of Glioblastoma (GBM) using molecular descriptors and gene expression data

Abstract

Background and Objective:

Glioblastoma (GBM), a very aggressive type of brain tumor, sometimes creates a chemoresistant state that compromises the effectiveness of chemotherapy and leads to serious clinical complications in treatment. Predicting drug resistance is crucial for the improvement of medication effect during cancer treatment. Assessing drug resistance is difficult due to the pricey chemotherapeutic trails and prolonged laboratory investigations. Deep learning plays a significant role in drug resistance prediction nowadays.

Methods:

This paper presents a novel deep learning model that combines Convolutional Neural Networks (CNN), Long Short Term Memory Networks (LSTM), and transformer architectures to predict drug resistance. The proposed application acts as a system that estimate the resistance of drugs based on gene expression details and chemical properties.

Results:

As compared with existing model for drug resistance prediction, proposed model achieved lower Mean Squared Error (MSE) of 0.4109 and Mean Absolute Error (MAE) of 0.5040, along with higher $R^2$ of 0.9635 and pearson correlation of 0.9999.

Conclusions:

This work significantly advances the fields of pharmacogenomics and personalized medicine through an in-depth evaluation that includes complex performance metrics and visualizations.