Generation of DNA Strands Representing Handwritten Numbers Based on GAN

Our work aims to use Generative Adversarial Networks (GANs) to generate DNA strands that represent handwritten numbers. We converted handwritten numbers as a kind of artificial information into DNA sequences, and generated new DNA sequences by training as a text generation. We compared the similarity between the output results of GANs and the real data and found that the similarity between them is very high. From this result, we believe that it is feasible to convert artificial information into DNA sequences and use neural networks to generate new sequences. This also proves the powerful development potential of synthetic biology and molecular computers in the future.