JAM-2: Fully computational design of drug-like antibodies with high success rates
Nabla Bio

Abstract
We present JAM-2, a general-purpose de novo protein design system that for the first timeyields VHH-Fc and full-length mAb antibodies with drug-like affinities and developability, whileachieving double-digit success rates with unprecedented target and epitope breadth. Across 16unseen targets, JAM-2 produced binders for 100% of them, with average success rates of 39%(VHH-Fcs) and 18% (mAbs). Using only 45 designs per format, JAM-2 delivered picomolar orsingle-digit nanomolar binders on half of these targets. Across another 10 targets, each with 20user-specified epitopes, JAM-2 generated VHH-Fc binders to 30-70% of epitopes for half thetargets. Remarkably, JAM-2 also produced antibody binders directly to the GPCRs CXCR4 andCXCR7 in their native cellular contexts with 11.7% and 3.8% success rates, respectively, withtop designs reaching single-digit nanomolar affinities. Developability profiling of hundreds of denovo designs- the largest developability dataset for computationally designed biologics-showed more than half met core industry criteria, with many top designs exhibiting lead-qualityprofiles that may not reguire further optimization of any kind. These results position JAM-2 asthe state-of-the-art de novo antibody design system, and the first ready for front-line use in drugdiscovery, matching or surpassing traditional discovery approaches.
Build AI with AI
From idea to launch — accelerate your AI development with free AI co-coding, out-of-the-box environment and best price of GPUs.