A Turing-Style Test for In-Silico Antibodies: How Sampling Mode Makes WGAN-GP Beat VAE in the Wet Lab

Abstract

Abstract Recently, it has become feasible to generate antibodies in silico using AI-based approaches such as deep learning, natural language processing, and diffusion models. This opens the door to computational antibody design as a complement to laboratory-based methods (animal immunization, hybridoma technology, and molecular display) for biologic drug discovery. However, when proposing human antibody sequences, or libraries thereof, it remains essential to determine whether they can be expressed, purified, and biophysically characterized using standardized laboratory assays typically applied early in discovery campaigns. In this work, we devised a Turing imitation game inspired experiment to compare two deep learning methods, Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), for generating experimentally viable de novo antibody sequences. A Wasserstein GAN with gradient penalty produced sequences that were successfully validated at a markedly higher rate (92/93; 99%) than those from a VAE (2/40; 5%). These contrasting outcomes can be rationalized by differences in model architecture and sampling strategies. Our findings highlight the promise of computationally driven discovery of antibody-based biotherapeutics.