Synthesizing Microbiome-Disease Association Data using GANs
Anushka Naik, Ishan Patwardhan, Amit Joshi · 2023
This study evaluates the proficiency of four Generative Adversarial Network models in synthesizing datasets related to microbiome-disease associations. Utilizing the comprehensive Human Microbe Disease Association Database as a foundational resource, a multifaceted evaluative approach is employed, incorporating statistical rigor and visual analytics. While CTGAN and Copula GAN are originally designed for tabular data generation, this study aims to extend the capabilities of WGAN-GP and Vanilla GAN to explore their applicability in the same context. Results reveal that while specialized tabular GANs are adept at encapsulating intricate data patterns, the adaptively employed GANs also manifest commendable promise. The study serves as a robust analytical framework, furnishing valuable insights for ongoing and future research in the realm of microbiome-disease data synthesis.