Last Updated: 10/06/2026
Leveraging Large language models for the design of monoclonal antibodies against malaria
Objectives
The project aims to develop novel human monoclonal antibodies (mAbs) against Plasmodium vivax by leveraging large language models and advanced computational techniques. This proposal is based on the co-crystal structure of PvAMA1 with a human monoclonal antibody 826827, which shows potent inhibition in both pre-erythrocytic and erythrocytic stages of the parasite.
Current treatments for Plasmodium vivax (Pv) are limited, highlighting the need for new therapies targeting both liver and blood stages. Relapsing parasites account for up to 80% of infections and disease, establishing a global reservoir that is difficult to eliminate with current treatments. Blocking liver-stage infections is the most effective strategy to prevent dormant parasites. A human monoclonal antibody (humAb826827) targets Pv apical membrane antigen1 (AMA1), blocking both liver and blood stage Pv infections. Large language models and binder designs will be used to develop novel human monoclonal antibodies targeting Pv invasion ligands, which will be tested in vitro with Pv clinical isolates. This strategy adapts methods used in enhancing anti- EGFR mAb, Cetuximab, for cancer therapy. Three approaches will be used to design mAbs: enhancing existing mAbs, designing new mAbs, and optimizing current mAbs. Enhanced and novel mAbs will be tested using a collaborative network in Cambodia. The computational approach avoids the bottleneck of isolating PBMCs from Pv-infected individuals. Developing effective mAb candidates requires optimization across multiple dimensions, including specific target binding, conformational stability, scalable production, and an acceptable immunogenicity profile. Aim 1 focuses on improving existing and developing new mAbs based on 826827 and 864865. Computational development of 1000 mAb scaffolds per target epitope will be done using RFDiffusion, PyRosetta, and MPNN. Optimal mAbs will be selected based on production efficiency, competition with mAb826827, and blocking capability. Aim 2 focuses on developing new mAbs recognizing PvCSP VK210 and VK247. Computational approaches from Aim 1 will be applied to PvCSP, using blocking murine mAbs to guide the generation of new mAbs. Optimal mAbs will be selected based on production efficiency, competition with murine 2F2, and blocking capability. The project aims to establish which target proteins are functionally relevant for blocking Pv growth and determine the best therapeutic mAbs, either alone or in combination.
Apr 2026 — Mar 2028
$282,757
Related Resources
