Abstract

In this webinar, we discuss a sequence-based protein language model (PLM) termed MAGE (monoclonal antibody generator) that is fine-tuned for the task of generating paired human variable heavy- and light-chain antibody sequences against targets of interest. Since the traditional process of antibody discovery is limited by inefficiency, high costs, and low success rates, approaches to utilize this form of artificial intelligence (AI) have been developed to optimize existing antibodies and generate antibody sequences in a target-agnostic manner. MAGE's training relies in part on antibody sequences identified using LIBRA-seq, a high-throughput single-cell sequencing technology that pairs B-cell receptor sequences with their cognate antigen specificities. We discuss the advantages of combining LIBRA-seq and MAGE into a unified antibody discovery engine, pairing high-throughput experimental identification of natural antibodies with AI-driven generation of novel candidates to accelerate and expand the reach of antibody discovery. In the work presented here, we show that MAGE can generate novel and diverse antibody sequences with experimentally validated binding specificity, as shown in proof-of-concept work for SARS-CoV-2, an emerging avian influenza H5N1, and respiratory syncytial virus A (RSV-A). MAGE represents a first-in-class model capable of designing human antibodies against multiple targets with no starting template.

Highlights:

• Protein language model (PLM) fine-tuned to generate antigen-specific human antibody sequences.

• Experimental validation of binding specificity.

• Proof-of-concept antibody generation for SARS-CoV-2, RSV-A, and H5N1.

• Combining LIBRA-seq and MAGE for accelerated antibody discovery.