How deep mutational scanning of viral proteins might aid do novo design of antibody countermeasures

(brainstorming slides for Bloom and Baker labs)

 

 

 

 

These slides: https://slides.com/jbloom/bloom-baker-dms-for-ab-design

Overview and goals

The Bloom lab has measured how all mutations to viral entry proteins affect the function of these proteins and their neutralization by antibodies. These data:

  • provide large datasets that could parameterize protein-protein interaction models
  • can identify highly constrained regions of viral proteins for de novo antibody design targeting
  • might be used to engineer antibodies to be more resilient to viral escape
  • might be used to engineer more potent antibodies
  • could be applied to assess if engineered antibodies are indeed more potent and escape resilient

 

These slides first provide some brief background on the topic and prior work by the Bloom lab. They then provide details on specific datasets that might help with the above goals.

Background on neutralizing antibodies to viruses

Viruses have one or more entry proteins that bind receptor and mediate membrane fusion

viral membrane

cell membrane

SARS-CoV-2 spike

spike conformational change

Image adapted from here

ACE2

Neutralizing antibodies can bind to these proteins and inhibit their cell entry functions

antibody

Image adapted from here

Viral proteins can acquire resistance mutations: eg, most early SARS-CoV-2 antibodies escaped by viral mutations

Cox et al (2022)

Example use of anti-viral antibodies: respiratory syncytial virus (RSV)

Especially in infants, RSV can cause airway inflammation and difficulty breathing

In USA, RSV is the leading cause of infant hospitalization. Infants hospitalized with RSV receive supportive care (eg, oxygen, ventilation, fluids) and usually recover (~0.1% in-hospital case-fatality rate)

 

In developing world where supportive care not available, RSV is a leading cause of infant mortality (~100,000 infant deaths per year)

Li et al (2022)

Image from ABC News

https://biologicmodels.com/project/rsv-fusion-f-glycoprotein/

RSV F protein fuses viral and cell membranes

Antibodies can bind to F and neutralize viral infection of cells

https://pdb101.rcsb.org/motm/286

The antibody nirsevimab bound to prefusion F at epitope that includes glycan. Note F is a trimer, and structure shows one antibody Fab bound to each monomer.

Antibodies can bind to F and neutralize viral infection of cells

https://pdb101.rcsb.org/motm/286

The antibody nirsevimab bound to prefusion F at epitope that includes glycan.

Simonich et al (2025)

Economic challenges of monoclonal antibodies against respiratory viruses

Although population disease burden is substantial, annual risk to any individual is low, and impossible to prospectively identify who will become severely ill at any given time.

Monoclonal antibodies are expensive to produce, and repeated dosing is required for sustained protection.

 

Viruses can evolve to become resistant. 

Although population disease burden is substantial, annual risk to any individual is low, and impossible to prospectively identify who will become severely ill at any given time.

Monoclonal antibodies are expensive to produce, and repeated dosing is required for sustained protection.

 

Viruses can evolve to become resistant.

Why anti-RSV antibody prophylaxis for infants is economically feasible

Severe disease concentrated in an easily identifiable population (infants).

Although population disease burden is substantial, annual risk to any individual is low, and impossible to prospectively identify who will become severely ill at any given time.

Monoclonal antibodies are expensive to produce, and repeated dosing is required for sustained protection.

 

Viruses can evolve to become resistant.

Severe disease concentrated in an easily identifiable population (infants).

Infants require lower dose (they're smaller), most need protection only for first year, and progress has been made in engineering more potent and long-lived antibodies.

Why anti-RSV antibody prophylaxis for infants is economically feasible

Although population disease burden is substantial, annual risk to any individual is low, and impossible to prospectively identify who will become severely ill at any given time.

Monoclonal antibodies are expensive to produce, and repeated dosing is required for sustained protection.

 

Viruses can evolve to become resistant.

Severe disease concentrated in an easily identifiable population (infants).

Infants require lower dose (they're smaller), most need protection only for first year, and progress has been made in engineering more potent and long-lived antibodies.

This is a still a concern, and is topic of this talk.

Why anti-RSV antibody prophylaxis for infants is economically feasible

Brief history of anti-F antibodies for RSV prevention in infants

Palivizumab

Lower potency than subsequent antibodies.

 

First approved primarily for prophylaxis of high-risk infants (eg, born prematurely at <36 weeks), with dosing of 15 mg/kg each month for five months.

 

Due in part to cost, recommendation progressively narrowed: by 2014 only for infants born <29 weeks gestational age or <32 weeks with chronic lung disease

Neutralization curves from Simonich et al (2025)

Suptavumab

Developed by Regeneron: much more potent against some strains than palivizumab.

 

Failed Phase 3 clinical trial from 2015-2017 due to lack of efficacy against subtype B; coincided with evolution of new variants with mutations at F sites 172 and 173.

Neutralization curves from Simonich et al (2025)

Developed by AstraZeneca and Sanofi; much more potent than palivizumab and has extended half life.

 

Recommended in 2023 for all infants <8 months old entering their first RSV season. Dosing just one injection of 50 mg for infants <5 kg.

 

~80% effectiveness in preventing RSV hospitalization.

Nirsevimab

Neutralization curves from Simonich et al (2025)

Clesrovimab

Developed by Merck, and has high potency and extended half life similar to nirsevimab but targets different region of F.

 

Similar recommendations for use as nirsevimab, only recently approved (in 2025)

Neutralization curves from Simonich et al (2025)

Pseudoviruses to study how mutations to F affect antibody neutralization

A pseudovirus system for measuring RSV neutralization

Simonich et al (2025)

Transfection of RSV F (and G) along with plasmids expressing lentiviral proteins creates pseudotyped viral particles, which can only undergo a single round of cell entry and are not pathogens.

This system makes it possible to measure neutralization of any F mutant or variant

Simonich et al (2025)

We can synthesize any variant of F and make pseudoviruses for neutralization assays, providing a safe way to study the effects of viral mutations on antibody neutralization.

Resistant strains have been identified in clinical and lab-passaging studies, plotted data from Simonich et al (2025)

Neutralization assays can characterize strains with resistance mutations

Pseudovirus deep mutational scanning to measure effects of all viral protein mutations on antibody neutralization

We used deep mutational scanning to quantify how all F mutations affect neutralization

Library of pseudoviruses expressing all single amino-acid mutants of RSV F.

 

Pseudoviruses can only undergo single round of cell entry, and so provide safe way to study effect of F mutations.

Simonich et al (2026)

Workflow for measuring how all F mutations affect antibody neutralization

We can also measure how all mutations affect pseudovirus cell entry in absence of antibody, providing a measure of functional constraint.

Simonich et al (2026)

How F mutations affect nirsevimab neutralization

Measurements of how mutations affect viral entry protein function quantify constraint on epitopes

Simonich et al (2026)

Letter heights indicate reduction in antibody neutralization, color indicates impact on F's cell entry function. These visualizations help quantify how constraint limits escape from different antibodies.

Nirsevimab

Clesrovimab

reduction in neutralization

site

cell entry function

De novo antibody design opens possibility of rationally choosing epitopes to minimize potential for viral escape

Historically, anti-viral antibodies isolated from infected or vaccinated humans. Those antibodies target whatever regions of a viral protein happen to be immunodominant in human immune response.

 

In many cases, those regions of a viral protein are also mutationally tolerant and can evolve rapidly. For instance, Jian et al (2022) showed how only 2 of 141 SARS-CoV-2 antibodies isolated from individuals who were exposed to an early SARS-CoV-2 variant still neutralized the viral variants present in 2023. This is because naturally elicited antibodies mostly target mutationally tolerant sites.

 

Our deep mutational scanning allows us to identify regions of viral proteins that cannot change without disrupting function. If antibodies could be rationally targeted to those epitopes there would be reduced potential for escape, and our deep mutational scanning could quantify this.

Rational targeting of constrained epitopes

For a given antibody, higher Fab affinity provides greater resilience to escape 

Two subtypes of RSV: A & B. Their F proteins have ~90% sequence identity & similar structures

Joyce et al (2019)

RSV B F structure

RSV A F structure

The structures have a RMSD deviation of only 1.8 angstroms.

Nirsevimab IgG has very similar neutralization activity against F from RSV A and RSV B

plotted data from Simonich et al (2025)

But resistance is more common for RSV B than A in infections of nirsevimab-dosed infants

studyresistance in RSV-Aresistance in RSV-B
Fourati et al (2025a)2/195 = 1%23/184 = 13%
Fourati et al (2025b)0/236 = 0%2/24 = 8%
Ahani et al (2023)0/11 = 0%2/14 = 14%

Rates of resistance to nirsevimab neutralization in RSV breakthrough infections of infants who received nirsevimab.

Experiments show some F mutations reduce nirsevimab neutralization of subtype B but not A

Simonich et al (2026)

IgG antibodies (like nirsevimab) are bivalent, with two connected Fabs

The apparent affinity (avidity) of bivalent IgG to antigen can be very high 

Klein and Bjorkman (2010)

In high avidity regime, viral mutations can reduce Fab but not IgG neutralization

See https://jbloomlab.github.io/IgG-vs-Fab-neutralization/notebook.html for full mathematical model

Model predicts viral mutations only reduce IgG neutralization of strain with lower Fab affinity

Simonich et al (2026)

Some RSV F mutations reduce IgG neutralization of subtype B, but Fab neutralization of A & B

Simonich et al (2026)

Therefore, engineering higher potency antibodies will reduce escape caused by viral mutations

Re-engineering existing antibodies to target new viral variants is valuable

The only SARS-CoV-2 antibody still approved in USA (VYD222) is a version of an earlier antibody that was re-engineered to target new viral variants (Yuan et al, 2025).

 

The company producing this has now also reported another generation of engineered antibody (VYD2311) even more potent against new antibodies (Mellis et al, 2026).

Real-world value of re-engineering existing antibodies to target new viral variants

Understanding our deep mutational scanning data in detail

What exactly do we measure?

For each mutation relative to a wildtype viral protein sequence, we measure:

  • Impact of mutation on viral protein cell entry function. This is reported as a cell entry score that is 0 if the mutation has no effect on cell entry function, >0 if it improves cell entry function, and <0 if it impairs cell entry function. Most mutations are either neutral (~0) or deleterious (<<0)
  • Impact of mutation on antibody neutralization (escape). This is reported as an escape score that is 0 if mutation has no effect on antibody neutralization, >0 if it reduces (escapes) neutralization, and <0 if it increases neutralization (negative escape). Most mutations typically have no effect, with mutations at a small handful of sites increasing neutralization. Usually these escape values will be roughly proportional to the log decrease in binding affinity although they have a limited dynamic range.
  • Only look at escape scores for mutants with at least moderate function. If a mutation strongly impairs cell entry function (eg, misfolds protein), its antibody escape score will not be reliable / meaningful.
  • The measurements have some noise. In our plotted data and default numerical CSVs we apply some reasonable filters, but for datasets you are especially interested in we can give more detailed assessments of noise / reliability.

Understanding our interactive visualizations

We like to look at the data in interactive visualizations like the one here. Elements:

  • Line plot at the top that reports summed effect of all mutations at each site on antibody escape
  • Heatmap showing effect of each mutation on escape. Light gray indicates mutation too deleterious for cell entry function to reliably measure escape; dark gray indicates missing from dataset.
  • Heatmap showing effect of mutation on cell entry function.
  • Interactive options including what threshold on cell entry function is used to call a mutation as "too deleterious to measure escape."

Understanding our numerical data

CSVs of numerical data are on GitHub, as per here. These report:

  • Site in protein in standard numbering scheme (usually what is in PDB)
  • Wildtype amino acid in protein used in our experiments
  • Mutant amino acid
  • Effect on antibody escape
  • Effect on cell entry function
  • Sequential number in protein sequence used in our experiment

RSV F data

RSV F is an important target, and several major companies (eg, Sanofi, Merck) have antibodies on market, and more are being developed:

SARS-CoV-2 spike data

SARS-CoV-2 antibodies remain important for vulnerable populations, and two antibodies (VYD222 and VYD2311) are in clinical use or development in USA:

Nipah RBP data

Nipah is a virus of pandemic concern, and antibodies are being developed as countermeasures. RBP is the Nipah receptor-binding protein:

Nipah F data

Nipah is a virus of pandemic concern, and antibodies are being developed as countermeasures. F is the Nipah fusion protein:

H5 influenza hemagglutinin data

Avian influenza is considered to be one of the priority pathogens of pandemic concern. There have been 5 influenza pandemics over the last 120 years and most pandemic strains had avian origins. No clinically approved antibody is available for use against influenza viruses.

These slides

These slides: https://slides.com/jbloom/bloom-baker-dms-for-ab-design

bloom-baker-dms-for-ab-design

By Jesse Bloom

bloom-baker-dms-for-ab-design

How deep mutational scanning of viral proteins might aid do novo design of antibody countermeasures (brainstorming slides for Bloom and Baker labs)

  • 50

More from Jesse Bloom