Multimodal functional deep learning for multiomics data.

Area of Science:

• Genomics
• Biostatistics
• Computational Biology

Background:

• High-throughput technologies enable large-scale multimodal omics data collection.
• Analyzing complex interactions within multiomics data for disease prediction presents significant challenges due to high dimensionality and noise.

Purpose of the Study:

• To propose a novel analytical method, multimodal functional deep learning (MFDL), for high-dimensional multiomics data analysis.
• To address the challenges in understanding complex biological mechanisms and predicting disease phenotypes from multiomics data.

Main Methods:

• Developed a multimodal functional deep learning (MFDL) method.
• MFDL utilizes deep neural networks to model relationships between multiomics variants and disease phenotypes.
• Incorporates functional data analysis to handle high-dimensional omics data and captures inter-omics interactions.

Main Results:

• MFDL demonstrates superior prediction accuracy compared to existing methods.
• The proposed method shows robustness in handling high-dimensional and noisy omics data.
• Simulation studies and real-data applications validate the effectiveness of MFDL.

Conclusions:

• MFDL offers a powerful approach for analyzing high-dimensional multiomics data.
• The method effectively models complex biological relationships and improves disease phenotype prediction.
• MFDL facilitates a more comprehensive understanding of disease mechanisms through integrated omics analysis.