• Attention
• Deep Learning
• Gene Expression
• Methylation
• OS Prediction

Predicting overall survival in medical research has been a significant challenge, often tackled by leveraging multi-modal data with advanced Machine Learning (ML) and Deep Learning (DL) techniques. Despite notable progress, persistent difficulties in model
accuracy and interpretability remain, prompting the search for alternative and more effective approaches.

In response to this, I developed MethWayOS, a novel model that integrates an attention mechanism within cross-modal analysis. This model focuses on the relationship between two critical omics datasets: RNA gene expression and DNA methylation.
By incorporating an attention-driven framework, MethWayOS is capable of dynamically weighing the contributions of each modality, enabling the model to better capture complex biological interactions and improve both its performance and interpretability.

MethWayOS has been validated using case study data from The Cancer Genome Atlas (TCGA), achieving promising results in overall survival prediction. The model not only targets state-of-the-art performance but also seeks to enhance the interpretability of the
relationship between gene expression and DNA methylation, two key omics data in cancer research. Additionally, the learned attention maps provide insights into the biological processes driving the survival predictions, which could aid in clinical decision-making.

Predicting overall survival in medical research has been a significant challenge, often tackled by leveraging multi-modal data with advanced Machine Learning (ML) and Deep Learning (DL) techniques. Despite notable progress, persistent difficulties in model accuracy and interpretability remain, prompting the search for alternative and more effective approaches. In response to this, I developed MethWayOS, a novel model that integrates an attention mechanism within cross-modal analysis. This model focuses on the relationship between two critical omics datasets: RNA gene expression and DNA methylation. By incorporating an attention-driven framework, MethWayOS is capable of dynamically weighing the contributions of each modality, enabling the model to better capture complex biological interactions and improve both its performance and interpretability. MethWayOS has been validated using case study data from The Cancer Genome Atlas (TCGA), achieving promising results in overall survival prediction. The model not only targets state-of-the-art performance but also seeks to enhance the interpretability of the relationship between gene expression and DNA methylation, two key omics data in cancer research. Additionally, the learned attention maps provide insights into the biological processes driving the survival predictions, which could aid in clinical decision-making.