Geometry-Preserving Supervised Biological Sequence Design

Abstract

Design of functional biological sequences such as DNA, RNA and peptides has wide-ranging applications in nanomaterials, bio-sensing and medicine. One common challenge across applications is the need to optimize complex high-dimensional properties such as target emission spectra of DNA-mediated fluorescent nanoclusters, photo and chemical stability, and antimicrobial activity of peptides across target microbes. Existing models optimize simple binary labels (e.g., binding/non-binding) as opposed to high-dimensional complex properties. To address this gap, we propose a geometry-preserving variational autoencoder framework, called PrIVAE, which learns latent sequence embeddings that respect the geometry of their property space. Specifically, we model the property space as a high-dimensional manifold that can be locally approximated by a nearest neighbor graph. We employ the property graph to guide the latent sequence representations through 1) GNN encoder layers and 2) an isometric regularizer. PrIVAE learns a property-organized latent space that allows rational design of new sequences with desired properties. We evaluate the utility of our framework for two generative tasks: 1) design of DNA sequences that template fluorescent metal nanoclusters and 2) design of anti-microbial peptides. The trained models retain high reconstruction accuracy while organizing the latent space according to properties. Beyond in silico experiments, our interdisciplinary team employed the designed sequences for wet lab synthesis of DNA-stabilized silver nanoclusters, resulting in up to 16.1-fold enrichment of rare-property nanoclusters compared to their abundance in training data and demonstrating the practical utility of our framework.