TSCytoPred: a deep learning framework for inferring cytokine expression trajectories from irregular longitudinal gene...

Area of Science:

• Immunology and Computational Biology
• Biomarker Discovery and Predictive Modeling

Background:

• Cytokines are critical immune regulators, but their dysregulation is linked to severe diseases like cancer and COVID-19.
• Longitudinal cytokine profiling is essential for predicting disease severity and patient outcomes, yet data is often limited.
• Computational inference of cytokine expression from gene expression data offers a promising approach to overcome data scarcity.

Purpose of the Study:

• To develop and validate TSCytoPred, a deep learning model for inferring time-series cytokine expression trajectories from gene expression data.
• To assess the model's performance in predicting cytokine dynamics and its utility in clinical disease prediction.
• To address limitations in current time-series cytokine data availability for robust predictive modeling.

Main Methods:

• TSCytoPred utilizes deep learning on time-series gene expression data, identifying key genes through interaction relationships and correlations.
• A neural network with an interpolation block estimates cytokine expression trajectories between observed time points.
• The model was evaluated using a COVID-19 dataset, comparing its performance against baseline regression methods.

Main Results:

• TSCytoPred significantly outperformed baseline methods, achieving high R² and low MAE in cytokine expression trajectory inference.
• Inferred cytokine data from TSCytoPred improved the prediction of COVID-19 patient severity risk.
• The model demonstrated effectiveness on datasets with limited time points and irregular temporal gaps.

Conclusions:

• TSCytoPred provides a robust method for inferring cytokine expression dynamics, addressing limitations of sparse time-series data.
• The model enhances the clinical utility of multi-omics data for disease outcome analysis, particularly in contexts like COVID-19.
• TSCytoPred facilitates early disease detection, treatment response prediction, and expands the application of multi-omics in rare diseases.