Transcriptome-based Deep Learning Model for Predicting Gemcitabine and Cisplatin Chemotherapy Response in Urothelial...

Area of Science:

• Oncology
• Genomics
• Computational Biology

Background:

• Advanced urothelial carcinoma (UC) treatment relies on gemcitabine and cisplatin chemotherapy.
• Patient response to chemotherapy varies, necessitating predictive biomarkers.
• Accurate prediction of treatment response is vital for optimizing therapy and minimizing toxicity.

Purpose of the Study:

• To develop a deep learning model for predicting chemotherapy response in UC patients.
• To utilize RNA sequencing gene expression data for predictive modeling.
• To enhance personalized treatment strategies for urothelial carcinoma.

Main Methods:

• A deep learning model was constructed using RNA sequencing data from TCGA and GEO.
• External validation was performed on an independent patient cohort.
• Gene ontology and survival analyses were conducted for model interpretation.

Main Results:

• The deep learning model achieved 94.7% accuracy in training and 90.0% in external validation.
• Key biological pathways associated with response include DNA damage response and cell cycle regulation.
• Predicted responders demonstrated significantly better survival outcomes (p=0.019).

Conclusions:

• A transcriptome-based deep learning model shows promise for predicting chemotherapy response in UC.
• This approach integrates high-dimensional data and machine learning for clinical decision support.
• The model offers a potential tool for personalized urothelial carcinoma treatment planning.