Broadly neutralizing antibodies (bNAbs) play a crucial role in fighting HIV-1, functioning by targeting the virus’s envelope proteins which shows promise in reducing viral loads and preventing infection. However, identifying these antibodies is a complex process due to the virus’s rapid mutation and evasion from the immune system. Only 255 bNAbs have been discovered, therefore necessitating assistance from AI tools to detect bNAbs from large immune datasets.
Researchers from Lausanne University Hospital, National Institutes of Health, and other institutions, developed a computational method known as RAIN. Unconventional in its technique, RAIN uses selected sequence-based features and machine learning, as opposed to traditional methods that rely on amino acid sequences or structural alignment. Experiments executed on BCR repertoires demonstrated RAIN’s 100% accuracy in predicting HIV-1 bNAbs, which were backed up by a number of validation studies such as in vitro neutralization assays and cryo-EM structural analysis.
This groundbreaking study complied with severe ethical guidelines and approvals from multiple institutional review boards. The research involved isolating serum IgG antibodies to investigate the immune response against HIV-1. Memory B cells were also isolated from peripheral blood mononuclear cells (PBMCs) and were then subjected to single-cell B-cell receptor sequencing utilizing three cutting-edge platforms: 10X Genomics, BD Rhapsody, and Singleton.
Detailed functional analysis was performed, and recombinant antibodies and Fab fragments were produced and purified. Effectiveness of the antibodies against various HIV-1 strains was assessed through neutralization assays. Advanced data processing and structural modelling tools like CryoSPARC, ChimeraX, and Phenix were used to analyze these interactions.
Identifying bNAbs against HIV-1 remains challenging due to their significant sequence diversity. Traditional methods cannot overcome this challenge due to this variability. Nevertheless, bNAbs exhibit multiple unique characteristics, which the researchers exploited by developing a machine-learning framework to automatically identify bNAbs.
The study focused on an infected donor with broad neutralization capabilities. Utilizing the RAIN pipeline, they discerned three potential bNAbs that demonstrated high-affinity binding to the HIV-1 envelope and strong neutralization activity. These findings were confirmed through biophysical and neutralization assays. The identified antibodies, particularly bNAb4251, showed broad and potent neutralization, highlighting the efficiency of this pipeline in discovering therapeutic antibodies against HIV-1. This innovative computational method could expedite the discovery and identification of bNAbs against HIV-1, potentially revolutionizing treatment options for the disease in the future.
