Predicting neutralization susceptibility to combination HIV-1 monoclonal broadly neutralizing antibody regimens

Area of Science:

• Immunology
• Virology
• Computational Biology

Background:

• Broadly neutralizing antibodies (bnAbs) are key for HIV-1 prevention strategies.
• Predicting in vitro neutralization potency of bnAbs and combinations against HIV-1 is an active research area.
• Current prediction models often combine individual bnAb neutralization data.

Purpose of the Study:

• To compare two distinct modeling approaches for predicting combination bnAb neutralization potency against HIV-1.
• To evaluate the performance of 'combine-then-predict' (CP) versus 'predict-then-combine' (PC) strategies.
• To inform the selection of antibody combinations for HIV-1 prevention efficacy trials.

Main Methods:

• Explored CP and PC approaches using simulated data.
• Utilized data from the Los Alamos National Laboratory's NAb Panels repository.
• Assessed prediction performance for both binary and continuous neutralization outcomes.

Main Results:

• The CP approach demonstrated superiority over the PC approach for binary neutralization outcomes (e.g., susceptibility).
• For continuous outcomes, CP performed comparably to PC with strong individual bnAb prediction algorithms.
• CP outperformed PC when individual bnAb prediction algorithms exhibited weaker performance.

Conclusions:

• The CP approach is a more reliable method for predicting combination bnAb neutralization potency, especially for binary outcomes.
• Findings support the use of the CP approach when in vitro data for antibody combinations is unavailable.
• This research aids in the evaluation and selection of bnAb combinations for HIV-1 prevention.