Active Learning for Rapid Design of Vaccines and Antibodies

Project Overview

The LDRD ER "Active Learning for Rapid Design of Vaccines and Antibodies" was conceived to address a need for rapid, scalable, computational design of therapeutic or prophylactic antibodies and vaccine antigens, two important classes of protein medical countermeasure (MCM). The state of the art at the beginning of the program, shortly prior to the COVID-19 pandemic, was that vaccine antigens and therapeutic antibodies were difficult and costly to develop from pathogens, in the case of vaccines, or discover from human or animal immune systems, in the case of antibodies. In both cases, the COVID-19 pandemic has seen some process acceleration. However, MCM proteins can be evaded by an evolving pathogen, as has happened repeatedly during the pandemic. Intentional, pre-emptive design, enabled by DOE computing, could result in significant acceleration of MCM design and delivery. Further advantages might accrue in terms of robustness to ongoing pathogen evasion when the designed MCMs are deployed. We approached this problem by building and expanding an autonomous, closed-loop, active learning software system meeting the requirements for deployment on Livermore Computing (LC) resources, deploying a suite of binding prediction codes and other protein design evaluation tools onto LC resources, and repeatedly exercising these methods for design of antigens (pre-pandemic) and antibodies (during the pandemic).

With the advent of COVID-19, significant additional resources were provided by LLNL, including an affiliated LDRD program for laboratory work (20-ERD-064). Starting from early in the pandemic, we targeted SARS-CoV-2, the causative agent of COVID-19, by re-purposing neutralizing antibodies against SARS-CoV-1 that had been identified in the wake of that outbreak in the early 2000's. We successfully re-targeted three different anti-SARS-CoV-1 antibodies to neutralize SARS-CoV-2 in vitro. As antibodies were identified from the blood of humans infected with SARS-CoV-2, we shifted to collaborate with academic partners to develop improved versions of their human-derived antibodies. The capabilities developed by 20-ERD-032 were employed in rapid response to the emergence of the Omicron variant of concern (VOC) in late 2021. In a matter of weeks, we computationally designed derivative antibodies of COV2-2130, one of two antibodies from Vanderbilt that form the basis of the AstraZeneca Evusheld prophylactic drug product. This drug product suffers a serious loss of efficacy against Omicron BA.1 and BA.1.1, the first Omicron strains. Our designs were successful, including a pair of designs which provide potent neutralization of not only Omicron BA.1 and BA.1.1, but also the earlier Delta variant, and subsequent Omicron strains including BA.2, BA.4, BA.5, and BA.2.75, demonstrating that our multi-target design process can, by its nature, produce robust antibody designs that strictly improve over the parental antibody. These results, recognized by a 2022 Director's Science and Technology award, have enabled the follow-on GUIDE program, to commence in FY23. We intend to publish shortly.

Mission Impact

20-ERD-032 and its internally- and externally-funded affiliates have provided a significant mission impact. They have developed science and technology tools and capabilities to meet future national security challenges, where this work will immediately continue under the GUIDE program. 20-ERD-032 has enabled rapid response to biological emergencies and, as the GUIDE program proceeds, will provide a deterrent to any attempt to engineer pathogens. As the COVID-19 pandemic has demonstrated, these capabilities may improve the security and resilience of the nation, not only directly from biological agents, but also from the significant disruptions to daily life incurred during a pandemic. These impacts are particularly relevant to DOD missions, where significant disruptions to force readiness or deployment can result from even public health pathogens.

Publications, Presentations, and Patents

Zhu, F. et al., 2022. "Large-Scale Application of Free Energy Perturbation Calculations for Antibody Design." Scientific Reports 12, 12489 (2022). https://doi.org/10.1038/s41598-022-14443-z.

Zhang, F. "Learning Region of Interest for Bayesian Optimization with Adaptive Level-Set Estimation." ICML2022 Workshop on Adaptive Experimental Design and Active Learning in the Real World, Baltimore, MD. 2022. LLNL-CONF-835910.

Arrildt, K. et al. "Using Antibody-Antigen Interactions to Design Better Vaccine Antigens." Keystone Symposia-Antibodies as Drugs (B1), February 2020, Santa Fe, New Mexico. 2020.