Noisy Epistasis Using Deep Learning

Nowadays, the analysis of the complex diseases through the epistatic interactions between single nucleotide polymorphisms (SNPs), for the detection of their statistical association with the disease is challenging due to curse of dimensionality, time complexity, absence of marginal effect and effect of the environmental factors. Studies of deep Learning (DL) techniques are shown to have more accurate results compared to other techniques such as Logistic Regression (LR), Multifactor dimensionality reduction (MDR) and associative classification-based multifactor dimensionality reduction (MDRAC). However, DL is not tested against different sources of noise. In this paper, we are concerned about studying the effect of different types of noise on a DL technique. Experiments are designed to compare the performance of the technique for different data models. The empirical results show that the DL approach gives robust and accurate results when compared to LR, MDR and MDRAC approaches.