Neuromuscular disease (NMDs) comprises a variety of diseases that affect the nervous system, which eventually lead to muscle weakness and in severe cases paralysis. Among all of the NMDs, facioscapulohumeral dystrophy (FSHD) is a topic of concern because of its unique genetic basis and its wide impact. The early diagnosis of FSHD is crucial for designing a disease-specific treatment plan. It can be diagnosed by genetic testing using microarray technology which can analyze large-scale genetic data. However, the high dimension of microarray data poses problem which is termed as “curse of dimensionality” in machine learning and requires advanced methods and techniques to address the problem. The chapter demonstrates the utility of VAEs in dimensionality reduction of microarray data. VAE is a type of generative model which takes the input of training data and downscales it using an encoder to a lower-dimensional space namely latent space. After that a decoder uses the latent space vector to reconstruct the training data. We exploit this this ability of VAE to downsample the microarray data. The latent space vector is used to train the classifiers. The results demonstrate that VAE was effectively able to reduce the dimensions of the data retaining useful information. Classifiers like Naïve Bayes and Logistic Regression achieved the highest accuracy. This chapter suggests that VAEs can be a reliable tool for dimensionality reduction task.