An Ovarian Cancer Susceptible Gene Prediction Method Based on Deep Learning Methods

Area of Science:

• Oncology
• Genetics
• Bioinformatics

Background:

• Ovarian cancer (OC) is a leading cause of cancer death in women globally.
• Late-stage diagnosis and high recurrence rates contribute to poor survival outcomes.
• Identifying novel OC-associated genes is crucial for understanding disease mechanisms.

Purpose of the Study:

• To develop and validate a novel computational method for predicting ovarian cancer causal genes.
• To leverage multi-omics data and deep learning for enhanced gene discovery.
• To identify new therapeutic targets and understand OC pathogenesis.

Main Methods:

• Utilized graph attention network (GAT) for compact gene feature representation.
• Employed a deep neural network (DNN) for predicting OC-related genes.
• Integrated diverse omics data for comprehensive gene analysis.

Main Results:

• The proposed model achieved a high Area Under the Curve (AUC) of 0.761 and Area Under the Precision-Recall Curve (AUPR) of 0.788.
• Successfully predicted 245 novel ovarian cancer causal genes.
• Identified 10 top Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways associated with OC.

Conclusions:

• The novel deep learning approach demonstrates high accuracy and effectiveness in identifying OC causal genes.
• The predicted genes and pathways offer new insights into ovarian cancer mechanisms.
• This method holds potential for advancing OC research and therapeutic strategies.