The microbiome is a group of microorganisms that live inside the human
gut. Imbalances in this intestinal microbiome can lead to the development of
Inflammatory Bowel Disease (IBD), because of which it is essential to detect the
anomalies in the microbiome metagenomics, which can help in better diagnosis of
diseases like Crohn’s disease. Various anomaly and feature selection techniques are
available, such as statistical methods (Z-score & IQR, Principal Component Analysis
(PCA)) and machine learning techniques (Random Forest, Support Vector Machine
(SVM), and K-means clustering). However, due to the high dimensionality of the
genomics data, it becomes difficult for these techniques to capture the non-linear
relationships of the data, as there is a lot of probabilistic uncertainty associated with it.
In this research, we have tried to suggest a study for feature extraction and anomaly
detection methods using Variational Autoencoders (VAE) and Generative Adversarial
Networks (GAN) to tackle the problem of high dimensionality, proposing a better way
to capture non-linear relationships and handle probabilistic uncertainty. The expected
outcome for using the VAE and GAN approach is that it can help in better diagnosing
and predicting the diseases that are caused by genetic factors, such as Crohn’s disease.
Keywords: Crohn’s disease, Generative Adversarial Network (GAN), Metagenomics, Microbiome, Variational Autoencoders (VAE).
