Generative Adversarial Networks and Its Applications in Biomedical Informatics

Abstract

The basic Generative Adversarial Networks (GAN) model is composed of the input vector, generator, and discriminator. Among them, the generator and discriminator are implicit function expressions, usually implemented by deep neural networks. GAN can learn the generative model of any data distribution through adversarial methods with excellent performance. It has been widely applied to different areas since it was proposed in 2014. In this review, we introduced the origin, specific working principle, and development history of GAN, various applications of GAN in digital image processing, Cycle-GAN, and its application in medical imaging analysis, as well as the latest applications of GAN in medical informatics and bioinformatics.

Relevant Applications of Generative Adversarial Networks in Drug Design and Discovery: Molecular De Novo Design, Dimensionality Reduction, and De Novo Peptide and Protein Design

“…At first, Goodfellow et al [23] introduced the concept of the GAN architecture as a form of generative models in an adversarial way. Since then, the GAN architecture has turned into one of interesting and hot topics in the field of artificial intelligence and machine learning [30,31]. Especially, the GAN architecture has become state-of-the-art in the field of computer vision and image processing (such as image generation) where astonishing progresses have been achieved [30,31].…”

“…Since then, the GAN architecture has turned into one of interesting and hot topics in the field of artificial intelligence and machine learning [30,31]. Especially, the GAN architecture has become state-of-the-art in the field of computer vision and image processing (such as image generation) where astonishing progresses have been achieved [30,31]. Because there is a flood of publications on numerous variants of the GAN architecture in different branches of science and engineering, it is really challenging to follow the emerging trend [30,31].…”

“…There are a wide range of applications by using the GAN architecture, including computer vision, image processing, medical imaging, music, speech processing, and video processing, to name a few [30,31]. In addition, the GAN architecture has been applied in chemistry (such as chemoinformatics and molecular informatics) and biology (such as drug design and discovery, bioinformatics, medical informatics, and multi-omics) [30,31].…”

“…In this section, we introduce the following three variants: the Wasserstein GAN structure, the conditional GAN structure, and the deep adversarial autoencoder structure. In addition, the reader can refer to recent reviews by Alqahtani et al [30] and Lan et al [31] for other variants of the GAN architecture.…”

“…It should be noted that the following limitations may be attributed to not only the GAN architecture but also any generative approaches. As the GAN architecture employs deep neural networks, the common problems in deep neural networks may also be included in the GAN architecture [31]. For instance, the accuracy of the GAN-based structure would be rather low when the sample size is limited [31].…”