Optimizing multi-omics data imputation with NMF and GAN synergy.

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Multi-omics studies offer deep insights into disease mechanisms and treatment responses.
• Disparities in sample sizes across omics datasets pose challenges, leading to bias and reduced statistical power.
• Integrating diverse omics data is crucial for advancing precision medicine.

Purpose of the Study:

• To introduce OmicsNMF, a novel framework for imputing missing omics data and improving disease phenotype prediction.
• To address the challenge of sample size disparity in multi-omics datasets.
• To enhance the accuracy of data integration and predictive modeling in omics research.

Main Methods:

• OmicsNMF integrates Generative Adversarial Networks (GANs) for data generation with Non-Negative Matrix Factorization (NMF) for pattern discovery.
• The framework imputes missing omics data to create more complete and balanced datasets.
• NMF identifies underlying patterns, while GANs generate realistic synthetic data samples.

Main Results:

• OmicsNMF demonstrated superior performance in predicting breast cancer subtypes compared to baseline methods.
• Survival analysis using imputed omics profiles showed significant prognostic power for overall survival and disease-free status.
• The framework effectively imputes missing samples while preserving critical biological features.

Conclusions:

• OmicsNMF successfully leverages GANs and NMF to overcome sample size disparities in multi-omics data.
• The imputed data enhances the predictive accuracy for disease subtypes and patient outcomes.
• This approach holds significant potential for advancing precision oncology through improved data integration and analysis.