Bryan Briney
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Bryan Briney has a total of 468 co-authors. Here are the top co-authors by number of shared publications:
- Dennis R. Burton
Department of Immunology and Microbiology, The Scripps Research Institute, La Jolla, CA, USA.
21 shared publications Jonathan Hurtado
Department of Immunology and Microbiology, The Scripps Research Institute, La Jolla, CA 92037, USA; Consortium for HIV/AIDS Vaccine Development (CHAVD), The Scripps Research Institute, La Jolla, CA 92037, USA
13 shared publications- Andrew B. Ward
Scripps Research Institute
10 shared publications - Devin Sok
Consortium for HIV/AIDS Vaccine Development (CHAVD), Scripps Research Institute, La Jolla, CA 92037, USA
8 shared publications - Collin Joyce
Department of Immunology and Microbiology, The Scripps Research Institute, La Jolla, CA, USA
7 shared publications - Ian A. Wilson
Department of Integrative Structural and Computational Biology, The Scripps Research Institute
7 shared publications
Latest Publications
- Open Access
- 19 August 2026
- Published by Springer Nature in Nature Biotechnology
- p. 1-13
A blinded, prospective benchmark of in silico antibody discovery anchored to experimental affinity and developability
Experimentally validated prospective, blinded benchmarks are needed to separate durable advances from hype in computational antibody design. Here AIntibody, a challenge inspired by the Critical Assessment of Structure Prediction, tests 511 artificial intelligence (AI)-designed or predicted antibodies from 29 organizations on three tasks: in silico affinity maturation from phase 1 sequencing outputs, affinity ranking within heavy-chain complementarity-determining region 3 (HCDR3) clusters of a selection output and CDR design of proteins not included in a selection output. Validated with diverse experimental assays, several groups produced developable antibodies with affinities <100 pM. However, these successes were exceptions that did not transfer across tasks. Affinity-matured antibodies were modeled effectively. Except for one model, predicting high-affinity clones from clustered HCDR3 datasets was worse than random clone picking. Out-of-library design was highly variable for most method submissions, with many failing to outperform standard selections. The AIntibody challenge shows that AI can optimize antibodies in defined, biologically grounded regimes, in addition to highlighting critical gaps including affinity prediction and library-inspired antibody design and cross-task generalization.
- 19 August 2026
- Published by Springer Nature in Nature Biotechnology
- p. 1-13
Open AccessA blinded, prospective benchmark of in silico antibody discovery anchored to experimental affinity and developability
Experimentally validated prospective, blinded benchmarks are needed to separate durable advances from hype in computational antibody design. Here AIntibody, a challenge inspired by the Critical Assessment of Structure Prediction, tests 511 artificial intelligence (AI)-designed or predicted antibodies from 29 organizations on three tasks: in silico affinity maturation from phase 1 sequencing outputs, affinity ranking within heavy-chain complementarity-determining region 3 (HCDR3) clusters of a selection output and CDR design of proteins not included in a selection output. Validated with diverse experimental assays, several groups produced developable antibodies with affinities <100 pM. However, these successes were exceptions that did not transfer across tasks. Affinity-matured antibodies were modeled effectively. Except for one model, predicting high-affinity clones from clustered HCDR3 datasets was worse than random clone picking. Out-of-library design was highly variable for most method submissions, with many failing to outperform standard selections. The AIntibody challenge shows that AI can optimize antibodies in defined, biologically grounded regimes, in addition to highlighting critical gaps including affinity prediction and library-inspired antibody design and cross-task generalization.
- Open Access
- 2 July 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Longitudinal analysis of drift in the circulating human antibody repertoire over four years
The human immune system is a sophisticated network of cells and molecules that plays a vital role in safeguarding the body against a broad range of foreign agents. One of the key components of the immune system is the circulating antibody repertoire, a vast collection of different antibodies that recognize and eliminate foreign agents specifically. The diversity and specificity of the circulating antibody repertoire are essential for effective immunity. In a study of ten healthy donors, we previously demonstrated the accessible human antibody repertoire to be on the order of 1015-1018 distinct members, although an individual will only sample a fraction of that repertoire at a given time point. The mode of generation of the naïve antibody repertoire through random recombination before antigen contact suggests that the composition of the circulating antibody repertoire may drift over time, but the magnitude of this drift is poorly understood. Here, we used high-throughput sequencing to analyze two donors from our original study, approximately four years later. Our results reveal conservation of the size, overall diversity, and gross features of the circulating antibody repertoire, such as V- and J-gene usage and CDRH3 length distribution, over time but suggest substantial changes to fine features of the repertoire, such as combinations of V, J, D gene segments and P and N sequences, which may influence primary responses to newly encountered pathogens as well as to immunization procedures.
- 2 July 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessLongitudinal analysis of drift in the circulating human antibody repertoire over four years
The human immune system is a sophisticated network of cells and molecules that plays a vital role in safeguarding the body against a broad range of foreign agents. One of the key components of the immune system is the circulating antibody repertoire, a vast collection of different antibodies that recognize and eliminate foreign agents specifically. The diversity and specificity of the circulating antibody repertoire are essential for effective immunity. In a study of ten healthy donors, we previously demonstrated the accessible human antibody repertoire to be on the order of 1015-1018 distinct members, although an individual will only sample a fraction of that repertoire at a given time point. The mode of generation of the naïve antibody repertoire through random recombination before antigen contact suggests that the composition of the circulating antibody repertoire may drift over time, but the magnitude of this drift is poorly understood. Here, we used high-throughput sequencing to analyze two donors from our original study, approximately four years later. Our results reveal conservation of the size, overall diversity, and gross features of the circulating antibody repertoire, such as V- and J-gene usage and CDRH3 length distribution, over time but suggest substantial changes to fine features of the repertoire, such as combinations of V, J, D gene segments and P and N sequences, which may influence primary responses to newly encountered pathogens as well as to immunization procedures.
- Open Access
- 3 June 2026
- Published by Elsevier in Cell Reports
- Vol. 45 (6) , 117439
Heterologous betacoronavirus spike immunization in non-human primates elicits antibodies that neutralize both sarbeco- and merbecoviruses
- 3 June 2026
- Published by Elsevier in Cell Reports
- Vol. 45 (6) , 117439
Open AccessHeterologous betacoronavirus spike immunization in non-human primates elicits antibodies that neutralize both sarbeco- and merbecoviruses
- Open Access
Evaluating Expert Specialization in Mixture-of-Experts Antibody Language Models
A bstract: Antibody language models (AbLMs) show an impressive aptitude for learning antibody features, but tend to struggle learning the highly diverse, non-templated regions of antibodies. Existing AbLMs use dense architectures, where all model parameters attend to each amino acid token. We hypothesized that the modular nature of antibodies could benefit from a sparse mixture-of-experts (MoE) architecture, allowing specific parameters (referred to as ‘experts’) to specialize in distinct antibody features. While MoE architectures are widely adopted and optimized in natural language processing domains, they are less common in biological modeling. To this end, we assess existing MoE routing strategies and find that token-choice routing strategies outperform expert-choice routing, presumably due to their specialization in CDRH3 residues. We further optimized the token-choice router for AbLMs, by minimizing the routing of padding tokens to enable pre-training with varying sequence lengths. Finally, we show that a large-scale baseline antibody language model with a Top-2 MoE architecture (BALM-MoE), trained on a mixture of unpaired and paired antibody sequences, outperforms its dense counterpart with the same number of active parameters.
Open AccessEvaluating Expert Specialization in Mixture-of-Experts Antibody Language Models
A bstract: Antibody language models (AbLMs) show an impressive aptitude for learning antibody features, but tend to struggle learning the highly diverse, non-templated regions of antibodies. Existing AbLMs use dense architectures, where all model parameters attend to each amino acid token. We hypothesized that the modular nature of antibodies could benefit from a sparse mixture-of-experts (MoE) architecture, allowing specific parameters (referred to as ‘experts’) to specialize in distinct antibody features. While MoE architectures are widely adopted and optimized in natural language processing domains, they are less common in biological modeling. To this end, we assess existing MoE routing strategies and find that token-choice routing strategies outperform expert-choice routing, presumably due to their specialization in CDRH3 residues. We further optimized the token-choice router for AbLMs, by minimizing the routing of padding tokens to enable pre-training with varying sequence lengths. Finally, we show that a large-scale baseline antibody language model with a Top-2 MoE architecture (BALM-MoE), trained on a mixture of unpaired and paired antibody sequences, outperforms its dense counterpart with the same number of active parameters.
- Top Cited PapersOpen Access
Germline-targeting HIV immunogen induces cross-neutralizing antibodies in outbred macaques
Top Cited PapersOpen AccessGermline-targeting HIV immunogen induces cross-neutralizing antibodies in outbred macaques
- Open Access
Heterologous betacoronavirus spike immunization in nonhuman primates elicits cross-reactive antibodies that neutralize both sarbeco- and merbecoviruses
Summary: In anticipation of future coronavirus (CoV) pandemics, developing vaccines that elicit broadly neutralizing antibodies (bnAbs) against diverse CoVs is critical. Here, we vaccinated rhesus macaques with SARS-CoV-2 spike (S)-protein, then boosted with heterologous β-CoV S-proteins to focus responses to common conserved S2 bnAb epitopes. Initial SARS-CoV-2 priming elicited receptor-binding domain (RBD)-focused responses, while MERS-CoV boosting redirected responses toward the S2 region, including the stem-helix bnAb site. Although S2-directed serum cross-neutralization was undetectable and most isolated cross-reactive monoclonal antibodies (mAbs) targeted non-neutralizing epitopes, two S2 stem-helix mAbs were identified from memory B cells. These bnAbs neutralized diverse sarbeco- and merbecoviruses, including MERS-CoV, and conferred robust in vivo protection against SARS-CoV-2 challenge. Structural studies revealed that these macaque bnAbs closely mimic human S2-stem bnAbs induced by infection. These findings provide proof-of-principle for vaccination strategies that elicit broadly protective β-coronavirus responses and highlight non-human primates as a translational model for evaluating S2-targeted immunogens.
Open AccessHeterologous betacoronavirus spike immunization in nonhuman primates elicits cross-reactive antibodies that neutralize both sarbeco- and merbecoviruses
Summary: In anticipation of future coronavirus (CoV) pandemics, developing vaccines that elicit broadly neutralizing antibodies (bnAbs) against diverse CoVs is critical. Here, we vaccinated rhesus macaques with SARS-CoV-2 spike (S)-protein, then boosted with heterologous β-CoV S-proteins to focus responses to common conserved S2 bnAb epitopes. Initial SARS-CoV-2 priming elicited receptor-binding domain (RBD)-focused responses, while MERS-CoV boosting redirected responses toward the S2 region, including the stem-helix bnAb site. Although S2-directed serum cross-neutralization was undetectable and most isolated cross-reactive monoclonal antibodies (mAbs) targeted non-neutralizing epitopes, two S2 stem-helix mAbs were identified from memory B cells. These bnAbs neutralized diverse sarbeco- and merbecoviruses, including MERS-CoV, and conferred robust in vivo protection against SARS-CoV-2 challenge. Structural studies revealed that these macaque bnAbs closely mimic human S2-stem bnAbs induced by infection. These findings provide proof-of-principle for vaccination strategies that elicit broadly protective β-coronavirus responses and highlight non-human primates as a translational model for evaluating S2-targeted immunogens.
- Open Access
GhostFold: Accurate protein structure prediction using structure-constrained synthetic coevolutionary signals
The accuracy of protein structure prediction models such as AlphaFold2 is tightly coupled to the depth and quality of multiple sequence alignments ( MSAs ), posing a persistent challenge for proteins with few or no identifiable homologs. We present GhostFold, a method for conjuring structure-constrained synthetic MSAs from a single amino acid sequence, bypassing the need for traditional homology searches. Leveraging the ProstT5 protein language model and the 3Di structural alphabet, GhostFold projects a query sequence into a tokenized structural representation and iteratively back-translates to generate an ensemble of diverse, fold-consistent sequences. These synthetic alignments ( pseudoMSAs ) encode emergent coevolutionary constraints that are sufficient for high-accuracy structure prediction of difficult targets such as orphan proteins and hypervariable antibody loops. GhostFold consistently matches or exceeds the performance of MSA-based and language model-based structure predictors while being computationally lightweight and independent of large sequence databases. Notably, we observe a decoupling of confidence metrics (e.g., pLDDT) from prediction accuracy when using pseudoMSAs, suggesting that AlphaFold2’s internal confidence calibration is strongly influenced by the statistical properties of natural sequence alignments. These results establish that structure-guided synthetic MSAs can functionally substitute for evolutionary data, offering a scalable and generalizable solution to one of the central limitations in computational structural biology. GhostFold represents a shift from passive data mining to intelligent sequence synthesis, redefining how structural priors are encoded in deep learning-based protein folding.
Open AccessGhostFold: Accurate protein structure prediction using structure-constrained synthetic coevolutionary signals
The accuracy of protein structure prediction models such as AlphaFold2 is tightly coupled to the depth and quality of multiple sequence alignments ( MSAs ), posing a persistent challenge for proteins with few or no identifiable homologs. We present GhostFold, a method for conjuring structure-constrained synthetic MSAs from a single amino acid sequence, bypassing the need for traditional homology searches. Leveraging the ProstT5 protein language model and the 3Di structural alphabet, GhostFold projects a query sequence into a tokenized structural representation and iteratively back-translates to generate an ensemble of diverse, fold-consistent sequences. These synthetic alignments ( pseudoMSAs ) encode emergent coevolutionary constraints that are sufficient for high-accuracy structure prediction of difficult targets such as orphan proteins and hypervariable antibody loops. GhostFold consistently matches or exceeds the performance of MSA-based and language model-based structure predictors while being computationally lightweight and independent of large sequence databases. Notably, we observe a decoupling of confidence metrics (e.g., pLDDT) from prediction accuracy when using pseudoMSAs, suggesting that AlphaFold2’s internal confidence calibration is strongly influenced by the statistical properties of natural sequence alignments. These results establish that structure-guided synthetic MSAs can functionally substitute for evolutionary data, offering a scalable and generalizable solution to one of the central limitations in computational structural biology. GhostFold represents a shift from passive data mining to intelligent sequence synthesis, redefining how structural priors are encoded in deep learning-based protein folding.
- Open Access
A Single-Cell Atlas of Transcriptional and Immunoglobulin Repertoire Evolution in Early B Cell Development
The development of human B cells in the bone marrow can be separated into functionally and transcriptionally distinct subsets, which while canonically described have yet to be profiled deeply and individually and with NGS and bioinformatic techniques. In this study, single-cell RNA sequencing was performed on 65,110 B cells from six healthy donors to observe the foundation of the diversity observed in mature immune repertoires. Following the established narrative of development, committed B cells undergo heavy and light chain recombination, which together form the functional BCR. However, a granular per-subset phenotypic analysis reveals proliferative bursts following each recombination event, an aspect of BCR recombination previously undescribed in the pro-B phase of development. Heavy and light chain pairing becomes more similar to that of mature, circulating B cells with progress through lymphopoiesis, deleting features associated with autoreactivity. Prominent among these repertoire alterations is a substantial shortening of heavy chain CDR3s, and changes in V, D, and J gene usage. This study provides a detailed snapshot of the transcriptomic and repertoire genetics landscape of developing B cells, providing a deeper understanding of the repertoire-shaping influence of early selection processes, and introducing new avenues for repertoire development research.
Open AccessA Single-Cell Atlas of Transcriptional and Immunoglobulin Repertoire Evolution in Early B Cell Development
The development of human B cells in the bone marrow can be separated into functionally and transcriptionally distinct subsets, which while canonically described have yet to be profiled deeply and individually and with NGS and bioinformatic techniques. In this study, single-cell RNA sequencing was performed on 65,110 B cells from six healthy donors to observe the foundation of the diversity observed in mature immune repertoires. Following the established narrative of development, committed B cells undergo heavy and light chain recombination, which together form the functional BCR. However, a granular per-subset phenotypic analysis reveals proliferative bursts following each recombination event, an aspect of BCR recombination previously undescribed in the pro-B phase of development. Heavy and light chain pairing becomes more similar to that of mature, circulating B cells with progress through lymphopoiesis, deleting features associated with autoreactivity. Prominent among these repertoire alterations is a substantial shortening of heavy chain CDR3s, and changes in V, D, and J gene usage. This study provides a detailed snapshot of the transcriptomic and repertoire genetics landscape of developing B cells, providing a deeper understanding of the repertoire-shaping influence of early selection processes, and introducing new avenues for repertoire development research.
- Open Access
Structural Hijacking of FcRn by Human Astrovirus Spikes Reveals Conserved Epitopes for Broad-Spectrum Antivirals
Human astroviruses (HAstVs) are a leading cause of pediatric gastroenteritis and emerging systemic infections, yet no targeted therapies exist. A critical barrier to intervention has been the lack of molecular insights into viral entry, particularly the interaction between the HAstV capsid spike and its receptor, the neonatal Fc receptor (FcRn). Here, we report high-resolution crystal structures of the HAstV spike from classical serotypes 2 and 6 in complex with human FcRn, defining the conserved receptor-binding interface at atomic resolution. These structures reveal serotype-specific variations dictating receptor affinity and that reported neutralizing antibodies inhibit infection primarily by sterically blocking the receptor-binding site. Mapping conserved epitopes across classical HAstV serotypes provides a blueprint for design of broad-spectrum antivirals that disrupt viral entry. Notably, our structural data rationalize the potential repurposing of clinical FcRn inhibitors, such as Nipocalimab, to block HAstV infection, bridging critical gaps in astrovirus biology and antiviral development.
Open AccessStructural Hijacking of FcRn by Human Astrovirus Spikes Reveals Conserved Epitopes for Broad-Spectrum Antivirals
Human astroviruses (HAstVs) are a leading cause of pediatric gastroenteritis and emerging systemic infections, yet no targeted therapies exist. A critical barrier to intervention has been the lack of molecular insights into viral entry, particularly the interaction between the HAstV capsid spike and its receptor, the neonatal Fc receptor (FcRn). Here, we report high-resolution crystal structures of the HAstV spike from classical serotypes 2 and 6 in complex with human FcRn, defining the conserved receptor-binding interface at atomic resolution. These structures reveal serotype-specific variations dictating receptor affinity and that reported neutralizing antibodies inhibit infection primarily by sterically blocking the receptor-binding site. Mapping conserved epitopes across classical HAstV serotypes provides a blueprint for design of broad-spectrum antivirals that disrupt viral entry. Notably, our structural data rationalize the potential repurposing of clinical FcRn inhibitors, such as Nipocalimab, to block HAstV infection, bridging critical gaps in astrovirus biology and antiviral development.
- Open Access
- 11 September 2025
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 21 (9) , e1013473
A curriculum learning approach to training antibody language models
There is growing interest in pre-training antibody language models (AbLMs) with a mixture of unpaired and natively paired sequences, seeking to combine the proven benefits of training with natively paired sequences with the massive scale of unpaired antibody sequence datasets. However, given the novelty of this strategy, the field lacks a systematic evaluation of data processing methods and training strategies that maximize the benefits of mixed training data while accommodating the significant imbalance in the size of existing paired and unpaired datasets. Here, we introduce a method of curriculum learning for AbLMs, which facilitates a gradual transition from unpaired to paired sequences during training. We optimize this method and compare it to other data sampling strategies for AbLMs, including a constant mix and a fine-tuning approach. We observe that the curriculum and constant approaches show improved performance compared to the fine-tuning approach in large-scale models, likely due to their ability to prevent catastrophic forgetting and slow overfitting. Finally, we show that a 650M-parameter curriculum model, CurrAb, outperforms existing mixed AbLMs in downstream residue prediction and classification tasks.
- 11 September 2025
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 21 (9) , e1013473
Open AccessA curriculum learning approach to training antibody language models
There is growing interest in pre-training antibody language models (AbLMs) with a mixture of unpaired and natively paired sequences, seeking to combine the proven benefits of training with natively paired sequences with the massive scale of unpaired antibody sequence datasets. However, given the novelty of this strategy, the field lacks a systematic evaluation of data processing methods and training strategies that maximize the benefits of mixed training data while accommodating the significant imbalance in the size of existing paired and unpaired datasets. Here, we introduce a method of curriculum learning for AbLMs, which facilitates a gradual transition from unpaired to paired sequences during training. We optimize this method and compare it to other data sampling strategies for AbLMs, including a constant mix and a fine-tuning approach. We observe that the curriculum and constant approaches show improved performance compared to the fine-tuning approach in large-scale models, likely due to their ability to prevent catastrophic forgetting and slow overfitting. Finally, we show that a 650M-parameter curriculum model, CurrAb, outperforms existing mixed AbLMs in downstream residue prediction and classification tasks.