MOMA: a multi-task attention learning algorithm for multi-omics data interpretation and classification

Area of Science:

• Computational biology
• Bioinformatics
• Medical informatics

Background:

• Accurate diagnostic classification and biological interpretation are crucial in data-rich fields like biology and medicine.
• Integrating diverse data types is essential for predicting clinical phenotypes and disease prognosis.

Purpose of the Study:

• To propose a novel multi-task attention learning algorithm (MOMA) for multi-omics data analysis.
• To enhance diagnostic performance and biological interpretability by capturing key biological processes.
• To improve the prediction of complex disease prognosis.

Main Methods:

• Developed MOMA, a multi-task attention learning algorithm for multi-omics data.
• Employed a geometric approach to vectorize features and modules.
• Utilized an attention mechanism to focus on important modules within multi-omics data.

Main Results:

• MOMA demonstrated superior performance in classification tasks using public Alzheimer's disease and cancer data.
• The algorithm achieved high diagnostic performance and interpretability.
• Experimental comparisons confirmed the utility of MOMA, including the impact of its attention mechanism.

Conclusions:

• MOMA offers a powerful approach for integrating multi-omics data for enhanced disease diagnosis and prognosis.
• The algorithm's attention mechanism is key to its improved performance and interpretability.
• MOMA provides a valuable tool for biological interpretation and clinical prediction in complex diseases.