An approved Ebola antibody may miss its mark against Bundibugyo virus

By Hugo Francisco de SouzaReviewed by Susha Cheriyedath, M.Sc.Sep 28 2026

Antibodies developed for one Ebola virus species face a different viral target in the 2026 outbreak. Structural models reveal why testing each treatment against Bundibugyo virus matters.

Study: Early insights into predicted efficacy of Ebola monoclonal antibodies for the 2026 Bundibugyo virus disease outbreak. Image Credit: Corona Borealis Studio / Shutterstock

In a recent study published in the journal Nature Communications, researchers modeled the binding behavior of approved and experimental Ebola monoclonal antibodies (mAbs) against the virus causing the emerging 2026 Bundibugyo virus disease (BVD) outbreak.

The research team first gathered 44 Bundibugyo virus (BDBV) genomes collected in 2007, 2012, and 2026, including 12 from the 2026 outbreak in the Democratic Republic of the Congo (DRC) and Uganda. They then identified mutations in these sequences and used structural modeling to map changes in the viral glycoprotein (GP) and to computationally predict each mAb's binding stability. The authors also checked whether the identified mutations persisted in additional available sequences.

The modeling predicted that the licensed monotherapy mAb114 (Ebanga) would exhibit displaced binding and a sharply reduced affinity for BDBV, an observation the researchers attributed to key epitope mutations. For Inmazeb, while two individual components of the triple-antibody cocktail were predicted to miss their intended sites on their own, the model predicted that the three antibodies bound at their intended sites together.

The most notable finding was that both components of the experimental pan-ebolavirus cocktail MBP134 targeted entirely conserved epitopes and were predicted to bind without structural displacement. Together, these computational predictions indicate that therapeutic efficacy cannot be assumed when moving across divergent viral species, offering evidence to help prioritize experimental testing and guide outbreak response in the DRC and Uganda.

Background

Ebolavirus outbreaks carry devastating health consequences, with fatality rates reaching as high as 90% in some past outbreaks. Treatments currently approved by the United States Food and Drug Administration (FDA; specifically, mAb114 and REGN-EB3 [Inmazeb]) were developed for Ebola virus (EBOV, Orthoebolavirus zairense).

The 2026 outbreak in Central and East Africa is driven by Bundibugyo virus (BDBV, Orthoebolavirus bundibugyoense), which is itself an orthoebolavirus. Genetic analyses have demonstrated that O. zairense and O. bundibugyoense share only 63.9% genome-wide identity and roughly 65% amino acid similarity in their glycoprotein composition. The international health community has therefore faced uncertainty about whether approved EBOV therapeutics could neutralize BDBV or whether the substantial genetic divergence between the two species might limit their effectiveness.

About the study

To address these knowledge gaps, the authors analyzed 44 BDBV whole-genome sequences spanning 2007, 2012, and 2026, including 12 collected from the 2026 outbreak (eight from the DRC and four from Uganda). The researchers checked whether identified mutations persisted in additional available sequences.

The study investigated predictions for three primary therapeutic candidates: 1. mAb114 (commercially named 'Ebanga'), 2. Inmazeb (a trimeric cocktail composed of Atoltivimab, Odesivimab, and Maftivimab), and 3. An experimental cocktail called MBP134 (synthesized by combining ADI-15878 and ADI-15946).

The computational work used AlphaFold 3 (AF3) deep-learning structural predictions plus Rosetta-based in silico mutagenesis to quantify changes in binding free energy (ΔΔG in kilocalories per mole [kcal/mol]). The predictions were further used to compute how tightly antibody surfaces packed against the viral protein and the steric impact of site-specific glycosylation at Asn563.

Study findings

Structural analyses found that BDBV has evolved 10 shared glycoprotein substitutions in the 2012 and 2026 sequences relative to the 2007 reference. Two novel mutations, Y387H and R506S, were present in all 12 analyzed 2026 sequences and absent from the earlier sequences compared.

Antibody binding predictions showed that, for mAb114, the critical recognition region (residues 111–119) harbored E112D and P116A mutations, which the model indicated would cause the antibody to dock off-target. Subsequent mutagenesis simulations estimated that E112D destabilized the interface. P116A was estimated to reduce docking affinity.

The simulations indicated that, when modeled as monotherapies, two of Inmazeb's components, Atoltivimab and Maftivimab, would fail to engage their intended epitopes. The third component, Odesivimab, was predicted to bind its intended site independently.

Maftivimab alone was predicted to be severely disrupted by natural glycan shielding at Asn563. When the three antibodies were modeled together, the structural prediction showed glycan accommodation and binding at the intended sites.

Modeling also showed that both components of MBP134 (ADI-15878 and ADI-15946) targeted epitopes that remained 100% conserved in all 44 BDBV genomes and were predicted to bind at their intended target sites without structural displacement.

Conclusions

The findings suggest that epitope conservation and interactions among cocktail components may influence binding. The authors note that their static computational models cannot capture the dynamic physiological movements of the viral glycoprotein or the full complexity of real-world antibody binding, and they caution that these findings should not be used alone to alter clinical practice. The predictions provide a rationale for comparing Maftivimab with the complete Inmazeb cocktail through experimental testing.

These computational findings need urgent validation through binding measurements and live-virus or pseudovirus neutralization assays. The results identify MBP134 and the full Inmazeb cocktail as candidates for experimental evaluation; their effectiveness against BDBV remains unknown.