Pith. sign in

REVIEW 4 major objections 7 minor 30 references

Learning from B Cell Evolution: Adaptive Multi-Expert Diffusion for Antibody Design via Online Optimization

T0 review · 4 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Adaptive multi-expert physics guidance with online per-target tuning improves antibody design success rates.

desk verdict Per-target Bayesian optimization tunes the exact metrics the paper reports as gains, so the headline success-rate improvement is largely a selection artifact. read the letter →

arxiv 2508.02834 v2 pith:IGHVN2VE submitted 2025-07-25 cs.LG cs.AI

Hanqi Feng , Peng Qiu , Mengchun Zhang , Yiran Tao , You Fan , Jingtao Xu , Barnabas Poczos This is my paper · ORCID
classification cs.LGcs.AI
keywords antibodydesigndiffusionmodelsSE(3)-equivariantgenerationmulti-objectiveoptimizationphysics-basedguidanceonlineBayesianhotspotcoverageaffinitymaturation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Antibody design by diffusion models fails when one metric lags far behind the rest, so the paper argues that generation should be steered per antigen rather than by a fixed protocol. It adds four physics-based guidance experts—van der Waals balance, epitope-hotspot recognition, contact-density balance, and interface geometry—to an SE(3)-equivariant backbone diffusion pipeline, and an online Bayesian optimizer learns each expert's temporal activation profile for each target. The paper reports that this adaptive guidance lifts the fraction of designs passing its success thresholds from 36.7% to 41.1%, with gains in CDR-H3 RMSD, hotspot coverage, interface pAE, and shape complementarity, while reducing variance. If correct, the practical consequence is that better antibody designs can be obtained by tuning guidance per target without retraining or replacing the base generative model. The biological inspiration is B cell affinity maturation, where antibodies are refined by competing selection pressures acting over iterative rounds.

What carries the argument

The load-bearing object is the guidance gradient $g(T_t,t)=\sum_{i=1}^4 w_i(t)\,\lambda_i(t)\,\nabla \mathcal{L}_i(T_t)$ injected into every reverse-diffusion step. Each $\mathcal{L}_i$ is one expert's loss: $L_{\text{vdw}}$ penalizes atom overlaps, $L_{\text{hotspot}}$ pulls CDR residues toward annotated epitope hotspots, $L_{\text{contact}}$ keeps interface contact counts in a target range, and $L_{\text{geom}}$ enforces distance uniformity and penalizes cavities. The temporal strength $\lambda_i(t)=\lambda_{\mathrm{base},i}\,f_{\text{temporal}}(t,\alpha_i,\beta_i)$ is a Beta-distribution profile whose shape parameters $(\alpha,\beta)$ are learned per antigen by Bayesian optimization; small values of $\alpha$ give early-peaking guidance for global structure, large $\alpha$ give late-peaking atomic refinement. Because all gradients are built from pairwise distances and relative orientations, the combined guidance remains SE(3)-equivariant.

What would settle it

Run the same pipeline on one target with three different hotspot sets—the paper's curated five, five high-buried-surface-area residues chosen independently, and five random residues—and compare success rates. If the curated set beats the random set by less than the 4.4-point advantage over the undirected baseline, or if random selection closes that gap, then hotspot annotation, not adaptive guidance, explains the result.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that the “weakest link” problem in computational antibody design can be attacked during generation. The method wraps an existing SE(3) diffusion backbone in a multi-expert guidance layer: each expert computes a gradient from a physical loss—steric clashes, hotspot proximity, contact density, interface uniformity—a routing module weights experts by current structural severity, and a Gaussian-process Bayesian optimizer chooses the shape parameters of each expert's Beta-distributed activation schedule for the target at hand. The reported outcome is a balanced improvement across all evaluation metrics, with success rate 41.1% versus 36.7% for the state-of-the-art undirected baseline, a 7% reduction in CDR-H3 RMSD, 9% higher hotspot coverage, 12% better interface pAE, and 5% higher shape complementarity. The paper further claims that fixed expert guidance alone reaches only 38.9% success, so the online adaptation step, not merely the physics losses, drives the gain.

Load-bearing premise

The load-bearing premise is that the five manually chosen hotspot residues per target are the right ones: they steer the recognition expert and are also the metric used to score designs, and the paper reports that swapping hotspot sets can change success rates by up to 15 percentage points, so noisy annotation could produce the observed gain.

Editorial extensions

If this is right

  • Per-target adaptation is the main driver of improvement: fixed physics guidance alone raises success from 36.7% to 38.9%, and online learning adds a further 2.2 points while shrinking standard deviations.
  • Different antigen classes need different temporal schedules, so a universal guidance protocol leaves performance on the table; the paper finds early-peaking activation for small epitopes and late-peaking activation for large interfaces.
  • The reported success-rate gain means fewer designs have to be generated and validated per viable candidate, which can offset the roughly 15% extra compute per design in a full campaign.
  • The same guided-diffusion pipeline generalizes across six therapeutically relevant targets spanning small epitopes and large protein interfaces, suggesting the method can be transferred to new antigens without retraining the base model.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural test is to use the learned Beta schedules as a prior for a new antigen: if schedules cluster by epitope type, they could seed Bayesian optimization at a good starting point instead of the neutral (2,2) profile.
  • Because hotspot coverage is both the steering signal and one of the headline metrics, removing the recognition expert or replacing manual hotspots with an automated predictor would reveal how much of the gain is adaptive scheduling versus annotation information.
  • The online-optimization loop is generic: the same Gaussian-process search over schedule parameters could tune guidance for other generative structure models, including flow-matching or sequence co-design pipelines, rather than only diffusion backbones.
  • The current adaptation happens between batches; a truly B-cell-like version would update the schedule within a single denoising run based on real-time structural metrics, a variation the paper's routing mechanism only partially implements.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. The paper proposes an adaptive multi-expert guidance framework for SE(3)-equivariant diffusion-based antibody design. Four physics-inspired experts (van der Waals, molecular recognition, energy balance, and interface geometry) provide guidance gradients during reverse diffusion, with per-target temporal schedules parameterized by Beta distributions whose shape parameters (α, β) are selected by Bayesian optimization with Gaussian processes. The method is evaluated on six antibody-antigen targets against DiffAb and RFAntibody, reporting a 41.1% success rate versus 36.7% for RFAntibody, along with improvements in CDR-H3 RMSD, hotspot coverage, interface pAE, and shape complementarity. The central claim is that per-target online adaptation of guidance improves multi-objective antibody design without retraining the base diffusion model.

Significance. If the reported results were established by a sound evaluation protocol, the paper would make a useful contribution: it extends diffusion-model guidance with target-specific online adaptation, provides an explicit equivariance argument for the guidance gradients, includes an ablation separating fixed expert guidance from online learning, and broadens the benchmark to six targets. The appendix discussion of hotspot sensitivity is unusually candid and is a credit to the authors. However, the current evidence for the central claim is undermined by the circular evaluation protocol: the Bayesian optimizer is run on the same metrics that define the success criterion, on the same targets that are then reported in the headline tables, with no validation split. The hotspot-coverage result is also partially self-referential because hotspot coverage is both a guided objective and an evaluation metric. The significance of the contribution is therefore contingent on a re-evaluation that separates parameter selection from performance reporting.

major comments (4)
  1. [Online Parameter Learning (Eq. 25), Appendix I, Tables 2 and 3] The per-target Bayesian optimization selects (α, β) by maximizing L(θ) = Σ_m ω_m·metric_m/ν_m, where the metrics are CDR-H3 backbone RMSD, pAE, and ipAE, and the paper's success criterion is thresholds on exactly these three metrics. Appendix I states that during online optimization the success rate rises from 36.7% to 39–40% and then exceeds 41%, so the headline "Ours 41.1%" in Table 2 is the value obtained after per-target selection using the evaluation metrics on the tested targets. No held-out validation or nested evaluation is described anywhere in the paper. Because RFAntibody receives no equivalent per-target fitting, the comparison conflates method quality with test-set selection; any optimizer that can query the evaluation score would appear to improve that score. This is load-bearing for the central claim and must be addressed by reporting fixed-hyperparameter results, a validation split, or a statistically disciplined selection protocol.
  2. [Appendix "Impact of Hotspot Residue Selection" and Eqs. (14)–(15)] The appendix states that varying hotspot selections on the same target can change success rates by up to 15%. Since the Molecular Recognition Expert actively attracts CDR residues toward the manually annotated hotspots and Hotspot Coverage is both a guided objective and an evaluation metric, the reported gains (e.g., hotspot coverage increasing from 48.5% to 57.5%, and the overall success improvement) may be largely determined by the choice of hotspot annotation rather than by the guidance framework itself. The main text gives no robustness analysis over hotspot sets, so the transferability of the headline results to new targets is not established.
  3. [Tables 6 and 7, per-target results] The pooled improvement is not statistically established. For 5J13, Ours (29.4% important pass rate) is below RFAntibody (30.8%), and for several targets the per-target differences are within the reported spreads; the pooled success rates 41.1±10.5 vs 36.7±14.0 overlap at one standard deviation. The paper should provide per-target significance tests or effect sizes, and should report the number of independent design batches used to compute the standard deviations.
  4. [Tables 6 and 7, DiffAb baseline] The DiffAb baseline numbers are implausible as structural RMSD values: CDR-H3 RMSD = 25.56±0.22 Å for 5NGV, 20.57±0.58 Å for 6PPG, and 17.03±0.62 Å for 5J13, with near-zero standard deviations, while DiffAb CDR Int. Pass is 0.0 for every target. This suggests either a different evaluation protocol (e.g., alignment to a different reference or inclusion of failed designs) or a computational error. Because DiffAb is one of only two baselines, this undermines the comparative claim; please clarify the exact protocol and report success rates separately for designs that produced valid structures.
minor comments (7)
  1. [Related Work and Experiments] RFAntibody is cited inconsistently as "Adolf-Bryfogle, Toth, and Bahl 2024" in Related Work and as "Bennett et al. 2024" in the Experiments section; please unify the citation and specify the exact model version used.
  2. [Table 3] Table 3 uses "RFdiffusion" as the label for the no-guidance baseline while the text calls the baseline "RFAntibody"; clarify whether the baseline is RFdiffusion alone or the full RFAntibody pipeline.
  3. [Figure 3 caption] The Figure 3 caption says "across four antibody targets," but the experiments report six targets; please update the caption to be consistent with the main text.
  4. [Table 1 and Eq. (25)] The normalizers ν_m for RMSD, pAE, and ipAE are listed as 1.5, 7.0, and 10.0, but the success thresholds are 3.0, 10.0, and 10.0; the relationship between the normalizers and the thresholds should be explained.
  5. [Abstract] The claim of being the "first biologically-motivated framework" is too strong given existing affinity-maturation-inspired design work cited in Appendix A; please soften or qualify the novelty claim.
  6. [Appendix E, Table 1] The Beta parameter ranges are listed as [0.5, 10.0] for both α and β, but the initial value is (2.0, 2.0) and Appendix I reports convergence to α∈[1.5,3.5], β∈[1.5,4.0]; clarify whether the ranges are search bounds or expected ranges.
  7. [General] No data or code availability statement is provided; to support reproducibility, please release the hotspot annotations, evaluation code, and per-target parameter traces.

Circularity Check

2 steps flagged · score 6.0 of 10

Per-target Bayesian optimization maximizes a score built from the same CDR-H3 RMSD, pAE, and ipAE metrics used to define success and report Table 2; the reported gains are selection artifacts, not independent predictions.

  1. fitted input called prediction [Online Parameter Learning, Eq. (25); Evaluation Metrics / Overall Success Criteria; Table 2; Appendix I]
    "After each design evaluation, the GP posterior is updated using the observed loss: L(θ) = P M m=1 ωm· metricm νm (25) where the metrics include CDR-H3 backbone RMSD, predicted aligned error (pAE), and interaction pAE (ipAE), each normalized by appropriate constantsνm. Overall Success Criteria: Successful designs must satisfy: CDR-H3 RMSD< 3.0 ˚A, mean pAE< 10.0, and mean interaction pAE (ipAE)< 10.0."

    The Bayesian optimizer selects per-target Beta parameters by maximizing a weighted objective composed of CDR-H3 RMSD, pAE, and ipAE, while the paper's success criterion thresholds exactly these three metrics and Table 2 reports success rate plus these metrics as the method's outcome. Appendix I confirms the headline 41.1% is reached only during BO on the same targets (success rate improves from baseline 36.7% to approximately 39-40%, then exceeds 41%). No held-out split is described, and RFAntibody receives no equivalent per-target fitting. The reported success-rate and RMSD/pAE/ipAE improvements are therefore values of the optimizer's own objective on the fitted targets, not independent predictions of a generalizable gain.

  2. self definitional [Equations (14)-(15); Evaluation Metrics, Hotspot Coverage; Appendix 'Impact of Hotspot Residue Selection']
    "L_hotspot = sum_{h∈H} min_{c∈CDR} ||x_h − x_c||^2 (14) The gradient attracts the nearest CDR residue c* toward uncovered hotspots: ∇L_hotspot = 2 sum_{h∈H_uncovered}(x_{c*}−x_h) (15). Hotspot Coverage: Percentage of critical epitope residues forming close contacts with CDR residues."

    The Molecular Recognition Expert is defined as a loss pulling CDR residues into contact with the annotated hotspot set H, and Hotspot Coverage is defined as the percentage of those same epitope residues forming close contacts with CDR residues. Thus the reported 9% hotspot-coverage improvement is the guided loss itself rather than an emergent property. The appendix makes the dependency explicit: varying hotspot selections on the same target can lead to success-rate differences of up to 15%, so the guided-and-evaluated metric is heavily determined by the manual annotation that also drives the guidance.

full rationale

The method's construction—multi-expert equivariant gradients, Beta-schedule parameterization, GP surrogate, and routing logic—is internally consistent and is not justified by a self-citation chain; the one shared-author citation (Shen et al. 2024) is used only for a general limitations remark. However, the headline quantitative comparison is partly circular by evaluation design. Eq. (25) makes the GP acquisition objective a weighted sum of CDR-H3 RMSD, pAE, and ipAE; the Overall Success Criteria threshold exactly these three metrics; and Appendix I states that the 41.1% success rate is obtained after online optimization on the same targets. Since no validation split is described and the RFAntibody baseline receives no equivalent per-target tuning, the pooled improvement in success rate and in the three optimized metrics is a selection artifact rather than an out-of-sample prediction. Per-target Table 8 additionally shows 5J13 with Ours (29.4%) below RFdiffusion (30.8%), and standard deviations overlap broadly, so the pooled gain is not statistically established. A second, narrower circularity is that hotspot coverage is both a guided loss (Eq. 14-15) and a reported evaluation metric, with the appendix reporting up to 15% sensitivity to the manual hotspot annotation. Other reported metrics (SC, BSA, VDW, pLDDT) are not in the BO objective and retain independent content, which prevents the entire derivation from collapsing to definitional circularity. Overall score 6: one or more central reported predictions reduce by construction to the optimizer's objective.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The central claim rests on manually curated hotspots, hand-set guidance strengths and thresholds, and per-target schedule parameters fitted to the evaluation metrics. The information-theoretic and SO(3) derivations are standard background. No new physical entities are postulated; the four experts are engineered loss functions.

free parameters (8)
  • Beta schedule parameters alpha, beta per target = e.g., small epitopes: alpha=3.2, beta=1.8; large interfaces: alpha=2.1, beta=3.5
    Tuned by Gaussian-process Bayesian optimization to maximize the reported evaluation metrics (Eq. 25).
  • Base guidance strengths lambda_base,i = VDW 0.5, Recognition 2.0, Energy adaptive, Interface 1.0
    Hand-set in Appendix E; not learned.
  • VDW clash threshold r_clash = 2.8 Angstroms
    Hand-set threshold in the VDW expert loss (Eq. 12).
  • Activation threshold theta_min = 0.1
    Hand-set threshold in expert routing (Eq. 21).
  • Peak factor lambda_peak = 5.0
    Hand-set scaling for temporal modulation (Eq. 11).
  • Hotspot combined score weights = S = 0.4*S_energy + 0.3*S_BSA + 0.2*S_distance + 0.1*S_conservation
    Hand-chosen weights for selecting the 5 hotspot residues per target (Appendix G).
  • Metric normalizers nu_m = RMSD 1.5, pAE 7.0, ipAE 10.0
    Hand-set in Eq. 25 and Table 1.
  • GP kernel lengthscale and noise = length scale 2.0, GP noise 0.3
    Hand-set in Table 1.
assumptions (5)
  • standard math IGSO3 score function on SO(3) (Appendix B)
    Standard Riemannian geometry on SO(3); accepted background.
  • standard math Mutual information between noisy and clean states is a strictly increasing function of SNR, with low-SNR slope d/2 (Appendix C)
    Uses Guo-Shamai-Verdú identity and Gaussian channel capacity; mathematically standard.
  • domain assumption Adding heuristic physics-based gradient guidance to the reverse diffusion update preserves sample validity (Eq. 8, Appendix J)
    Classifier-guidance style perturbation is assumed not to break the denoising process; the paper provides equivariance arguments but no distributional guarantee.
  • domain assumption The four expert losses capture the physics relevant to binding (VDW, hotspot coverage, contact density, interface geometry)
    These are coarse backbone-level heuristics; side chains are not present during backbone diffusion, so the physics is approximate.
  • ad hoc to paper Hotspot annotations are reliable and transferable
    The appendix reports up to 15% success-rate variation with different hotspot sets, making this a fragile, manually-curated input.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning from B Cell Evolution: Adaptive Multi-Expert Diffusion for Antibody Design via Online Optimization." pith.science (2026). https://pith.science/paper/IGHVN2VE

@misc{pith2026250802834,
  author       = {Pith},
  title        = {Pith review of: Learning from B Cell Evolution: Adaptive Multi-Expert Diffusion for Antibody Design via Online Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IGHVN2VE}},
  note         = {Machine review of arXiv:2508.02834}
}
read the original abstract

Recent advances in diffusion models have shown remarkable potential for antibody design, yet existing approaches apply uniform generation strategies that cannot adapt to each antigen's unique requirements. Inspired by B cell affinity maturation, where antibodies evolve through multi-objective optimization balancing affinity, stability, and self-avoidance, we propose the first biologically-motivated framework that leverages physics-based domain knowledge within an online meta-learning system. Our method employs multiple specialized experts (van der Waals, molecular recognition, energy balance, and interface geometry) whose parameters evolve during generation based on iterative feedback, mimicking natural antibody refinement cycles. Instead of fixed protocols, this adaptive guidance discovers personalized optimization strategies for each target. Our experiments demonstrate that this approach: (1) discovers optimal SE(3)-equivariant guidance strategies for different antigen classes without pre-training, preserving molecular symmetries throughout optimization; (2) significantly enhances hotspot coverage and interface quality through target-specific adaptation, achieving balanced multi-objective optimization characteristic of therapeutic antibodies; (3) establishes a paradigm for iterative refinement where each antibody-antigen system learns its unique optimization profile through online evaluation; (4) generalizes effectively across diverse design challenges, from small epitopes to large protein interfaces, enabling precision-focused campaigns for individual targets.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 26 canonical work pages

  1. [1]

    Gradients are computed in the model’s working coordi- nates

    work page

  2. [2]

    PLoS Computational Biology, 18(6): e1010271

    IG-V AE: Generative modeling of protein structure by direct 3D coordinate generation. PLoS Computational Biology, 18(6): e1010271. Evans, R.; O’Neill, M.; Pritzel, A.; Antropova, N.; Senior, A.; Green, T.; ˇZ´ıdek, A.; Bates, R.; Blackwell, S.; Yim, J.; et al. 2021. Protein complex prediction with AlphaFold- Multimer. biorxiv, 2021–10. Fischer, E. 1894. E...

    work page 2021

  3. [3]

    All geometric calculations use invariant features

    work page

  4. [4]

    In International Conference on Machine Learning , 8867–

    Equivariant diffusion for molecule generation in 3D. In International Conference on Machine Learning , 8867–

    work page

  5. [5]

    Iterative refinement using learned parameters This integration preserves the strengths of RFdiffusion while addressing its limitations in antibody-specific applications. Downstream Compatibility The improved structural quality from our approach benefits downstream tools: • ProteinMPNN achieves higher sequence recovery rates • AlphaFold2 validation shows i...

    work page

  6. [7]

    Advances in Neural Information Processing Systems, 35: 9754–9767

    Antigen-specific antibody design and optimization with diffusion-based generative models for protein struc- tures. Advances in Neural Information Processing Systems, 35: 9754–9767. Mason, D. M.; Friedensohn, S.; Weber, C. R.; Jordi, C.; Wagner, B.; Meng, S. M.; Ehling, R. A.; Bonati, L.; Dahin- den, J.; Gainza, P.; et al. 2021. Optimization of therapeutic...

    arXiv 2021

  7. [8]

    Chemistry-Inspired Diffusion with Non-Differentiable Guidance

    DLAB: deep learning methods for structure-based virtual screening of antibodies. Bioinformatics, 38(2): 377– 383. Shanehsazzadeh, A.; McPartlon, M.; Kasun, G.; Steiger, A. K.; Sutton, J. M.; Yassine, E.; Shuai, R.; Kohnert, C.; Osborn, M.; Mustafa, S.; et al. 2023. Unlocking de novo an- tibody design with generative artificial intelligence.bioRxiv, 2023–0...

    work page Pith review arXiv 2023

  8. [9]

    arXiv preprint arXiv:2203.02923

    GeoDiff: A geometric diffusion model for molecular conformation generation. arXiv preprint arXiv:2203.02923. Xue, L. C.; Dobbs, D.; Bonvin, A. M.; and Honavar, V . 2015. Computational prediction of protein interfaces: A review of data driven methods. FEBS Letters, 589(23): 3516–3526. Yanover, C.; Rosin, C. D.; Zheng, W.-Y .; M ¨osch, A.; and Bradley, P. 2...

    arXiv 2015

Show all 30 references
  1. [12]

    In the reinforcement learning domain, similar ideas have been explored for molecular optimization (Ben- gio et al

    proposed time-dependent guidance schedules for pro- tein design. In the reinforcement learning domain, similar ideas have been explored for molecular optimization (Ben- gio et al. 2021), though application to structure generation remains limited. Multi-Agent and Ensemble Appro...

    2021

  2. [15]

    No absolute position references are introduced

  3. [17]

    The guidance strength modulation is scalar (invariant) Limitations and Failure Analysis Common Failure Modes Despite overall improvements, our adaptive guidance frame- work exhibits specific failure patterns that provide insights for future development: Shallow Epitope Challen...

  4. [18]

    Sequence-Structure Coupling : The current approach optimizes structure before sequence, potentially missing co-evolutionary patterns

  5. [19]

    Dynamic Considerations: Static structure optimization may not capture important binding dynamics

  6. [20]

    Evaluation Metrics: Computational metrics imperfectly predict experimental outcomes

  7. [21]

    The combination of physics-informed expertise, intelligent routing, and contin- uous learning creates a robust framework for practical an- tibody design

    Computational Cost: Full pipeline evaluation remains expensive for large-scale screening Proposed Improvements Future work should address these limitations through: • Joint Optimization : Simultaneous sequence-structure optimization using coupled diffusion processes • Ensemble...

  8. [22]

    Initial structure generation using base RFdiffusion

  9. [23]

    Multi-expert guidance injection during reverse diffusion

  10. [24]

    Adaptive strength modulation based on timestep and metrics

  11. [25]

    Batch evaluation and Bayesian optimization-based pa- rameter updates

  12. [27]

    Annotation Bias : Expert selection introduces human bias and requires extensive domain knowledge, limiting scalability to new targets

  13. [28]

    Cytokines often have concentrated binding sites, while cell surface receptors may have distributed epitopes

    Target Variability : Optimal hotspot characteristics vary significantly across antigen types. Cytokines often have concentrated binding sites, while cell surface receptors may have distributed epitopes

  14. [29]

    Dynamic Interfaces : Many therapeutically relevant epitopes involve conformational changes not captured by static hotspot selection

  15. [30]

    Limited Transferability: Hotspot patterns learned on one target class may not generalize to structurally distinct antigens. Toward Hotspot-Free Design Recognizing these limitations, we are developing next- generation approaches to reduce hotspot dependency: Learnable Hotspot D...

  16. [2003]

    to machine learning frameworks: • Epitope-focused design: Methods targeting specific sur- face patches (Adolf-Bryfogle et al. 2018) • Affinity maturation simulation : Computational mimicry of somatic hypermutation (Tiller and Tessier 2017) • Developability optimization: Incorp...

    2018

  17. [2016]

    MAbs, 8(7): 1259–1268

    ABodyBuilder: Automated antibody structure pre- diction with data–driven accuracy estimation. MAbs, 8(7): 1259–1268. Lepore, R.; Olimpieri, P. P.; Messih, M. A.; and Tramontano, A. 2017. PIGSPro: prediction of immunoGlobulin structures v2. Nucleic acids research, 45(W1): W17–W...

    2017

  18. [2017]

    Machine Learning Transition The transition to machine learning methods began with the introduction of statisti- cal potentials learned from structural databases

    introduced antibody-specific features such as H3 kink prediction and VH-VL orientation sampling, but remained limited by the accuracy of their underlying energy functions. Machine Learning Transition The transition to machine learning methods began with the introduction of sta...

    2008

  19. [2020]

    Applications to Biomolecules The adaptation of diffu- sion models to biomolecular design required addressing the unique geometric constraints of molecular structures

    formalized the connection between diffusion models and score-based generative modeling, providing a theoreti- cal framework for continuous-time diffusion processes. Applications to Biomolecules The adaptation of diffu- sion models to biomolecular design required addressing the...

    2022

  20. [2022]

    Bioinformatics, 38(7): 1877–1880

    ABlooper: Fast accurate antibody CDR loop structure prediction with accuracy estimation. Bioinformatics, 38(7): 1877–1880. Adolf-Bryfogle, J.; Kalyuzhniy, O.; Kubitz, M.; Weitzner, B. D.; Hu, X.; Adachi, Y .; Schief, W. R.; and Dunbrack Jr, R. L. 2018. Affinity maturation of a...

    2018

  21. [2023]

    arXiv preprint arXiv:2305.14784

    Unified Guidance for Geometry-Conditioned Molec- ular Generation. arXiv preprint arXiv:2305.14784. Ho, J.; Jain, A.; and Abbeel, P. 2020. Denoising diffusion probabilistic models. Advances in neural information pro- cessing systems, 33: 6840–6851. Ho, J.; and Salimans, T. 2022...

    2020 arXiv

  22. [8887]

    Hsu, C.; Verkuil, R.; Liu, J.; Lin, Z.; Hie, B.; Sercu, T.; Lerer, A.; and Rives, A

    PMLR. Hsu, C.; Verkuil, R.; Liu, J.; Lin, Z.; Hie, B.; Sercu, T.; Lerer, A.; and Rives, A. 2022. Learning inverse folding from mil- lions of predicted structures. In International Conference on Machine Learning, 8946–8970. PMLR. Ingraham, J.; Baranov, M.; Costello, Z.; Barber,...

    2022 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.