Learn about LLNL’s innovative platform for rapid computational design and experimental validation of antibody therapeutics and vaccines.
LLNL biologists, computer scientists and engineers are developing a machine-learning enabled computational/experimental framework that enables unprecedented acceleration of the drug development process.
Our platform integrates world-class high-performance computing and simulation, machine learning, structural bioinformatics, and multi-objective optimization to perform in silico antibody and antigen design and optimization. These designs are then experimentally validated at LLNL and at our partner facilities, and results are either pursed as candidate medical countermeasures (MCMs) or provided as critical feedback for computational redesign.
We are partnering with academia, and industry to further develop and demonstrate this capability. When fully implemented, our system will enable both a proactive and rapid reactive approach for designing MCMs.
If you are interested in joining our team or have questions, please contact us at CPDB [at] llnl.gov.
Following reports in December 2021 that SARS-CoV-2 variants Omicron BA.1 and BA.1.1 escaped the clinical AZD1061 antibody (COV2-2130), we launched our platform to design COV2-2130 mutants that simultaneously restored binding to both BA.1 and BA.1.1, while maintaining strong binding to the co-circulating Delta variant.
In under 3 weeks, we used our platform to propose 376 antibody variants. Experimental evaluation confirmed enrichment across our design goals, with our top candidate potently neutralizing wildtype SARS-CoV-2 and the Delta, BA.1, BA.1.1, and BA.2 variants.
AbBERT: Learning Antibody Humanness Via Masked Language Modeling
One component of our computational pipeline is a model to understand the degree of similarity of novel, engineered antibody sequences to typical human antibody sequences. Accurate understanding of the degree of humanness can help reduce or manage the risk of failure modes such as immunogenicity or poor manufacturability. AbBERT was developed to fill this need as a transformer-based language model derived from ProtBert, which was trained on up to 20 million unpaired sequences from the Observed Antibody Space database.
We validated AbBERT using a novel “multi-mask'' scoring procedure to demonstrate high accuracy in predicting antibody complementary determining regions—including the challenging hypervariable H3 region. We then demonstrated several uses of AbBERT on downstream antibody design and prediction tasks. AbBERT enhances in silico antibody optimization via deep reinforcement learning by utilizing its learned embeddings as additional observations during optimization.
We have career opportunities to develop a next generation computational antibody and antigen design platform driven by simulation and machine learning models trained on large experimental datasets.
- Applied Machine Learning - Lead
- ChemBio Data Scientist
- Full Stack Software Developer
- Protein Structure Bioinformaticist
- Project Manager
- Microbiologist/Molecular Biologist
- Protein Chemist
- Biologist - Lab Automation Specialist for Molecular Biology/Biochemistry
- Staff Scientist - Molecular Biology, Protein Biochemistry, or Biophysics
- Biologist - Antibody Discovery, Developability, and Formulation
Lawrence Livermore National Laboratory has been ranked among the top 10 employers in the San Francisco Bay Area by Forbes Magazine, won multiple Glassdoor Best Places to Work awards, and is among the top 12 in government services nationwide. The Laboratory offers employees outstanding careers with a first-class benefits package in a beautiful location. See our Careers page for more information about our culture, the benefits and perks, our history, and job opportunities.
Dan Faissol, PhD
faissol1 [at] llnl.gov (Email)
Shankar Sundaram, PhD
sundaram1 [at] llnl.gov (Email)
Tom Bates, PhD
bates13 [at] llnl.gov (Email)
whalen3 [at] llnl.gov (Email)
Kate Arrildt, PhD
Virology, Vaccinology, LLNL
Rapid Response Laboratory Lead
Tom Desautels, PhD
Multi-fidelity simulation and machine learning for efficacy prediction
Ed Lau, PhD
Molecular dynamics simulation
Brenden Petersen, PhD
Deep reinforcement learning, multi-objective optimization, system integration
Brent Segelke, PhD
Molecular microbiologist and protein crystallographer
Fangqiang Zhu, PhD
Molecular dynamics simulation, structural antibody modeling
Adam Zemla, PhD
Protein structure bioinformatics
Drew Bennett, PhD
Molecular dynamics simulation, multi-scale modeling
Jose Cadena Pico, PhD
Matt Coleman, PhD
Vaccinologist, biochemical engineer
Jeff Drocco, PhD
Andre Goncalves, PhD
Machine learning, multi-task learning, multi-objective optimization
Software development lead
Biostatistics and data science, software development
Software development and database engineering
Tek Hyung Lee, PhD
Synthetic and systems biologist and microbial engineer
Jay Jayaraman Thiagarajan, PhD
Xander Ladd, PhD
Machine learning, Bayesian Optimization
Mikel Landajuela, PhD
Deep reinforcement learning
Leno Da Silva, PhD
Deep reinforcement learning
Biochemical assay development specialist
Nate Mundhenk, PhD
Deep learning, Convolutional Neural Networks
Priyadip Ray, PhD
Dante Ricci, PhD
Molecular genetics, systems biology
Cellular, molecular, micro - biologist/biochemist
Ed Saada, PhD
Pathogen interactions and medical countermeasures
Bayesian Statistics and Data Science
Biochemical and analytical scientist
Camilo Valdes, PhD
Deep learning language modeling, software development
Dina Weilhammer, PhD
Cellular, molecular, and viral biologist
Jiachen Yang, PhD
Reinforcement learning, machine learning
Boya Zhang, PhD
Machine learning, surrogate-based modeling
Interested in joining the cutting edge of computational design and bioengineering?
If you’re interested in joining our team or have questions, contact us at CPDB [at] llnl.gov.