No trillion-sequence library needed: MobGen designs monobodies from scratch that bind SARS-CoV-2 RBD and neutralize B.1
Synopsis
Monobodies, small binding proteins derived from a human fibronectin type III domain, previously required screening libraries of 3×1013 random sequences by phage display to find high-affinity binders against a viral antigen. MobGen chains RFdiffusion, AlphaFold2 backpropagation and ProteinMPNN into one computational pipeline; targeting the SARS-CoV-2 B.1 spike receptor-binding domain it generated 11,450 sequences, yielded 7 binders, and one of them neutralized B.1 pseudovirus at micromolar potency. Adding a C-terminal tail raised production about 7-fold on average and binding affinities reached 22.1–424.8 nM, but neutralization stayed around 2–25 µM and no monobody covered all tested variants.
Interpretation
Monobody binders can now be generated computationally from a target structure instead of being fished out of a random library. High-affinity monobodies had mainly come from phage display screens of large randomized in vitro libraries, including one work that screened a library of 3×1013 randomized sequences. Against the SARS-CoV-2 B.1 spike receptor-binding domain, the pipeline generated 11,450 sequences; 970 self-consistent designs entered yeast surface display, which returned 7 distinct binding sequences.
The pipeline first uses RFdiffusion to generate scaffolds and loop conformations, then optimizes interface sequence by AlphaFold2 backpropagation, and finally redesigns the binding surface with ProteinMPNN. Design and prediction stages were scored mainly by interface predicted alignment error (iPAE), and only designs with binder RMSD below 2.5 Å between design and prediction and pLDDT above 0.80 were accepted as self-consistent. Among three ProteinMPNN weight sets, SolubleMPNN gave the highest in silico success rate and was therefore chosen for interface refinement.
The two optimization steps improved yield and thermal stability, but ProteinMPNN redesign had little impact on binding affinity and gains in antiviral activity were limited. Most binders from the initial yeast display screen expressed at modest yields and had melting temperatures in the 42.8–53.5 °C range, with a single monobody reaching 66.2 °C. Fusing a 7-residue C-terminal tail raised production 7-fold on average, ranging from 2.6-fold to 11.4-fold; surface plasmon resonance on 15 tail-bearing binders, 4 from the initial library and 11 ProteinMPNN-refined, gave KD values of 22.1–424.8 nM, and refined Mob-9007T, 9030T and 9044T exceeded 70 °C melting temperature.
Perspective
The demonstrated scope is a single target, the SARS-CoV-2 B.1 spike receptor-binding domain, with binding and neutralization data from yeast surface display screening, bacterially expressed purified protein, pseudovirus assays and B.1 authentic-virus assays; the authors propose next evaluating immunogenicity, cytotoxicity and off-target binding directly, and redesigning against antigenically divergent variants or combining binders that recognize non-overlapping epitopes.
The authors report micromolar neutralization potency, no obvious relationship between surface-plasmon-resonance affinity and neutralization, possibly from relatively fast dissociation rates and epitopes that only partially overlap the ACE2-binding footprint; the final neutralizing monobodies retain only about 26% sequence identity to the parental fibronectin type III domain, weakening the low-immunogenicity assumption; and no design neutralized all variants tested.
