CoChat
Sign Up
Cited by 221Monoclonal and Polyclonal Antibodies Researchvaccines and immunoinformatics approachesProtein Structure and Dynamics

Computational approaches to therapeutic antibody design: established methods and emerging trends

Open access
Norman, Richard A, Ambrosetti, Francesco, Bonvin, Alexandre M J J, Colwell, Lucy J, Kelm, Sebastian, Kumar, Sandeep, Krawczyk, Konrad·2020·Briefings in Bioinformatics
DOI 10.1093/bib/bbz095PMC PMC7947987
Open this paper in CoChat — free accountOpen PDF

Abstract

Antibodies are proteins that recognize the molecular surfaces of potentially noxious molecules to mount an adaptive immune response or, in the case of autoimmune diseases, molecules that are part of healthy cells and tissues. Due to their binding versatility, antibodies are currently the largest class of biotherapeutics, with five monoclonal antibodies ranked in the top 10 blockbuster drugs. Computational advances in protein modelling and design can have a tangible impact on antibody-based therapeutic development. Antibody-specific computational protocols currently benefit from an increasing volume of data provided by next generation sequencing and application to related drug modalities based on traditional antibodies, such as nanobodies. Here we present a structured overview of available databases, methods and emerging trends in computational antibody analysis and contextualize them towards the engineering of candidate antibody therapeutics.

AI snapshot

Written from the paper’s full text.

TL;DR

This review surveys databases, computational methods, and emerging trends for therapeutic antibody design, covering numbering, structural modelling, interface prediction, docking, design, immunogenicity, and developability. It argues computational tools can guide but not replace experimental antibody discovery, with NGS data and nanobody-specific benchmarking as key emerging areas.

Key findings

  • Antibodies are the largest class of biotherapeutics, with five monoclonal antibodies ranked in the top 10 blockbuster drugs: adalimumab and infliximab (anti-TNFα), rituximab (anti-CD20), bevacizumab (anti-VEGF), and trastuzumab (anti-HER2/neu).
  • Established computational methods include homology modelling, protein–protein docking, and interface prediction; in the Antibody Modelling Assessment II, overall FV modelling accuracy averaged 1.1 Å RMSD, while CDRH3 was modelled to >5 Å RMSD in some targets.
  • Antibody-specific databases and data resources are expanding: SAbDab identified about 3500 of approximately 150,000 PDB structures as containing at least one antibody or nanobody chain, and Ab-Bind holds data on 1101 mutations across 32 antibody complexes.
  • Nanobodies are an emerging therapeutic format, with caplacizumab approved in 2018; their CDRH3 loops are on average three to four residues longer than antibody counterparts, and nanobody paratopes contain on average nearly three additional residues, making direct transfer of antibody computational tools uncertain.

Methods

  • Design: narrative review providing a structured overview of available databases, algorithms, and resources for computational antibody analysis and design.
  • Data or sample: no primary experimental sample; the paper synthesizes published antibody databases, computational tools, and literature on therapeutic antibody development, including NGS, structural, and experimental data resources.
  • Main analysis: qualitative synthesis and contextualization of methods toward engineering candidate antibody therapeutics, including tables of databases and computational tools.

Limitations

  • The paper states that computational methods are unlikely to replace the entire discovery process, that docking methods still cannot rival X-ray crystallography-derived structures, that existing epitope prediction systems still do not provide reliable predictions, and that the connection between computational immune epitope predictions and anti-drug antibodies remains largely untested.
  • The paper states that it is not clear whether current antibody modelling, docking, and related methods are directly transferable to nanobodies, and that systematic benchmarking is needed.
  • Evident from the text: this is a narrative review without a systematic search strategy, quantitative synthesis, or new experimental data.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1093/bib/bbz095

Cites

241 of 241 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 177 cited sources checked · 0 supported · 0 flagged

  • Findingnone given

    AntibodyInterfacePrediction and AG-Fast-Parapred were not compared against one another.

    “To the best of our knowledge AntibodyInterfacePrediction and AG-Fast-Parapred are currently the best performing paratope prediction methods as compared to previously available methods; however, they were not compared against one another.”
    Quote in the paperNot a substantial claim
  • Premisecitation

    Computational approaches hold promise in advancing the field by providing faster results than arduous experimental approaches that are the current standard in antibody discovery.

    “Computational approaches hold promise in advancing the field by providing faster results than arduous experimental approaches that are the current standard in antibody discovery [5].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [5] Krawczyk, 2017, Computational tools for aiding rational antibody design, Methods Mol Biol source

      Not checked in the free run.

  • Premisecitation

    The increasing number of structural, sequence and experimental data on antibodies being deposited in the public domain provides the necessary foundation for improving data-driven methods.

    “The increasing number of structural [16], sequence [17] and experimental data [18–21] on antibodies being deposited in the public domain provides the necessary foundation for improving such data-driven methods.”
    Quote in the paperSubstantial claim0 of 6 sources checked
    • cites [16] Dunbar, 2013, SAbDab: the structural antibody database, Nucleic Acids Res source

      Not checked in the free run.

    • cites [17] Kovaltsuk, 2018, Observed antibody space: a resource for data mining next-generation sequencing of antibody repertoires, J Immunol source

      Not checked in the free run.

    • cites [18] Vita, 2010, The immune epitope database 2.0, Nucleic Acids Res source

      Not checked in the free run.

    • cites [19] Sirin, 2016, AB-Bind: antibody binding mutational database for computational affinity predictions, Protein Sci source

      Not checked in the free run.

    • cites [20] Koenig, 2017, Mutational landscape of antibody variable domains reveals a switch modulating the interdomain conformational dynamics and antigen binding, Proc Natl Acad Sci U S A source

      Not checked in the free run.

    • cites [21] Jain, 2017, Biophysical properties of the clinical-stage antibody landscape, Proc Natl Acad Sci source

      Not checked in the free run.

  • Premisecitation

    CDRH3 is very important for antigen recognition.

    “CDRH3 possesses the highest sequence and structural diversity of the six CDRs [33] and is very important for antigen recognition [34, 35].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [34] Tsuchiya, 2016, The diversity of H3 loops determines the antigen-binding tendencies of antibody CDR loops, Protein Sci source

      Not checked in the free run.

    • cites [35] Xu, 2000, Diversity in the CDR3 region of V H is sufficient for most antibody specificities, Immunity source

      Not checked in the free run.

  • Premisecitation

    There are pharmaceutically focused computational approaches that aid in assessing the immunogenicity and biophysical properties of antibodies.

    “There are also pharmaceutically focused computational approaches that aid in assessing the immunogenicity [14] and biophysical properties [15] of antibodies.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [14] Abhinandan, 2007, Analyzing the ‘Degree of Humanness’ of antibody sequences, J Mol Biol source

      Not checked in the free run.

    • cites [15] Obrezanova, 2015, Aggregation risk prediction for antibodies and its application to biotherapeutic development, MAbs source

      Not checked in the free run.

  • Classified as probably not a substantial claim. 62
  • May combine two claims. 35

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    Computational advances in protein modelling and design can have a tangible impact on antibody-based therapeutic development.

    “Computational advances in protein modelling and design can have a tangible impact on antibody-based therapeutic development.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Antibodies are currently the largest class of biotherapeutics.

    “Due to their binding versatility, antibodies are currently the largest class of biotherapeutics, with five monoclonal antibodies ranked in the top 10 blockbuster drugs.”
    • Also stated in: § Introduction · plain
      “Because of such favorable binding properties, antibodies have been harnessed for therapeutic purposes and are currently the largest class of biotherapeutics.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [4] Kaplon, 2018, Antibodies to watch in, MAbs

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Five of the current top-selling blockbusters are monoclonal antibodies: adalimumab, infliximab, rituximab, bevacizumab and trastuzumab.

    “Five of the current top-selling blockbusters are monoclonal antibodies: adalimumab and infliximab (anti-TNFα), rituximab (anti-CD20), bevacizumab (anti-VEGF), trastuzumab (anti-HER2/neu) and their market presence is still expanding [4].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [4] Kaplon, 2018, Antibodies to watch in, MAbs

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Each of the estimated 5 × 109 B-cells in an organism produces a distinct B-cell receptor or antibody through somatic recombination of variable, diversity, joining and constant gene segments.

    “Each of the estimated 5 × 109 B-cells in an organism [23] produces a distinct B-cell receptor (membrane-bound) or antibody (in soluble form) through somatic recombination of variable (V), diversity (D), joining (J) and constant (C) gene segments [28, 29].”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [23] Briney, 2019, Commonality despite exceptional diversity in the baseline human antibody repertoire, Nature source

      Not checked in the free run.

    • cites [28] Tonegawa, 1983, Somatic generation of antibody diversity, Nature source

      Not checked in the free run.

    • cites [29] Hesslein, 2001, Factors and forces controlling V(D) J recombination, Adv Immunol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Resources exist that curate the therapeutic antibody information, such as TABS and SAbDab-Therapeutic-Antibodies.

    “Resources exist that curate the therapeutic antibody information, such as TABS (https://tabs.craic.com/) and SAbDab-Therapeutic-Antibodies [13].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [13] Raybould, 2019, Five computational developability guidelines for therapeutic antibody profiling, Proc Natl Acad Sci U S A source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Tools such as DIGIT and Abysis source their data from the European Nucleotide Archive (ENA) and the National Center for Biotechnology Information (NCBI); their sequence volumes are of the order of 105 and contain multiple artificially engineered sequences.

    “Tools such as DIGIT and Abysis source their data from the European Nucleotide Archive (ENA) [60] and the National Center for Biotechnology Information (NCBI) [61]; their sequence volumes are of the order of 105 and contain multiple artificially engineered sequences.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [60] Leinonen, 2011, The European nucleotide archive, Nucleic Acids Res source

      Not checked in the free run.

    • cites [61] Resource, 2017, Coordinators. Database Resources of the National Center for Biotechnology Information, Nucleic Acids Res source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Despite the intrinsic sequence diversity in the CDRs, all hypervariable loops with the exception of CDRH3 adopt a constrained set of conformations termed canonical classes.

    “Despite the intrinsic sequence diversity in the CDRs, all hypervariable loops with the exception of CDRH3 adopt a constrained set of conformations termed canonical classes [32].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [32] Chothia, 1987, Canonical structures for the hypervariable regions of immunoglobulins, J Mol Biol source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Natural preferences in antibody repertoires can be used as a reference to assess the biophysical properties of therapeutic antibodies or to develop naturally focused surface display libraries.

    “Among others, such natural preferences can be used as a reference to assess the biophysical properties of therapeutic antibodies [13] or to develop naturally focused surface display libraries [26].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [13] Raybould, 2019, Five computational developability guidelines for therapeutic antibody profiling, Proc Natl Acad Sci U S A source

      Not checked in the free run.

    • cites [26] Zhai, 2011, Synthetic antibodies designed on natural sequence landscapes, J Mol Biol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Most repositories collect both antibody and nanobody data; however, there are also some databases specializing only in the latter, e.g. sdAb-DB.

    “Most of these repositories collect both antibody and nanobody data; however, there are also some databases specializing only in the latter, e.g. sdAb-DB [58].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [58] Wilton, 2018, sdAb-DB: the single domain antibody database, ACS Synth Biol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Altogether, computational antibody analysis methods have matured enough to allow for wider applications in therapeutic development.

    “Altogether, computational antibody analysis methods have matured enough to allow for wider applications in therapeutic development.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    CDRH3 possesses the highest sequence and structural diversity of the six CDRs.

    “CDRH3 possesses the highest sequence and structural diversity of the six CDRs [33] and is very important for antigen recognition [34, 35].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [33] Regep, 2017, The H3 loop of antibodies shows unique structural characteristics, Proteins Struct Funct Bioinforma source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    These molecules are of increasing therapeutic interest with the first nanobody therapeutic (caplacizumab) approved in 2018.

    “Therefore, these molecules are of increasing therapeutic interest with the first nanobody therapeutic (caplacizumab) approved in 2018 [42].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [42] Duggan, 2018, Caplacizumab: first global approval, Drugs source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The leading database of germline antibody sequences is the International Immunogenetics Information System (IMGT) and it is widely used to derive gene assignments for recombined antibodies.

    “The leading database of germline antibody sequences is the International Immunogenetics Information System (IMGT) [46] and it is widely used to derive gene assignments for recombined antibodies.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [46] Lefranc, 2015, IMGT R, the international ImMunoGeneTics information system R 25 years on, Nucleic Acids Res source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The Protein Data Bank (PDB) is the main global repository of 3D structure information for proteins.

    “The Protein Data Bank (PDB) is the main global repository of 3D structure information for proteins [65].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [65] Berman, 2000, The protein data bank, Nucleic Acids Res source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Upon antigen exposure, the antibody-producing B-cells undergo a natural process of affinity maturation, based on somatic hypermutation.

    “Upon antigen exposure, the antibody-producing B-cells undergo a natural process of affinity maturation, based on somatic hypermutation [31].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [31] Peters, 1996, Somatic hypermutation of immunoglobulin genes is linked to transcription initiation, Immunity source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    According to SAbDab, of the approximately 150,000 structures deposited in the PDB to date as many as 3500 are identified as containing at least one antibody (or nanobody) chain.

    “According to SAbDab, of the approximately 150 000 structures deposited in the PDB to date as many as 3500 are identified as containing at least one antibody (or nanobody) chain.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [16] Dunbar, 2013, SAbDab: the structural antibody database, Nucleic Acids Res source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    SAbDab specifically allows for a bulk download of its weekly updatable database, providing an up-to-date resource for applications such as antibody modelling or docking.

    “SAbDab specifically allows for a bulk download of its weekly updatable database, providing an up-to-date resource for applications such as antibody modelling or docking.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [16] Dunbar, 2013, SAbDab: the structural antibody database, Nucleic Acids Res source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Due to the high-throughput nature of NGS, sequence volumes are of the order of 108 and are associated with non-trivial error rates.

    “Due to the high-throughput nature of NGS, sequence volumes are of the order of 108 and are associated with non-trivial error rates [62].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [62] Kovaltsuk, 2018, Filtering next-generation sequencing of the Ig gene repertoire data using antibody structural information, J Immunol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The accumulated methodology of computational antibody protocols could potentially be applied to novel antibody formats with intrinsically better biophysical properties, such as nanobodies.

    “The accumulated methodology of computational antibody protocols could potentially be applied to novel antibody formats with intrinsically better biophysical properties, such as nanobodies [27].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [27] Bannas, 2017, Nanobodies and nanobody-based human heavy chain antibodies as antitumor therapeutics, Front Immunol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Certain resources, such as Observed Antibody Space, address this by offering an annotation of the predicted sequence errors.

    “Certain resources, such as Observed Antibody Space, address this by offering an annotation of the predicted sequence errors [62].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [62] Kovaltsuk, 2018, Filtering next-generation sequencing of the Ig gene repertoire data using antibody structural information, J Immunol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The recent decade saw the advent of next generation sequencing (NGS) of B-cell receptor (antibody) repertoires.

    “Specifically, the recent decade saw the advent of next generation sequencing (NGS) of B-cell receptor (antibody) repertoires [22].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [22] Miho, 2018, Computational strategies for dissecting the high-dimensional complexity of adaptive immune repertoires, Front Immunol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    Computational approaches to analyze and design antibodies rely on the availability of suitable datasets.

    “Computational approaches to analyze and design antibodies rely on the availability of suitable datasets.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    NGS provides a snapshot of millions of antibody sequences sampled from the theoretically possible 1012-1015 antibody sequences in a human repertoire.

    “NGS provides a snapshot of millions of antibody sequences sampled from the theoretically possible 1012–1015 antibody sequences in a human repertoire [23, 24].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [23] Briney, 2019, Commonality despite exceptional diversity in the baseline human antibody repertoire, Nature source

      Not checked in the free run.

    • cites [24] Glanville, 2009, Precise determination of the diversity of a combinatorial antibody library gives insight into the human immunoglobulin repertoire, Proc. Natl Acad Sci U S A source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    In humans, the H and L chains can naturally assemble into five isotypes: IgG, IgD, IgE (all three monomers), IgA (dimer) and IgM (pentamer).

    “In humans, the H and L chains can naturally assemble into five isotypes: IgG, IgD, IgE (all three monomers), IgA (dimer) and IgM (pentamer) [30].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30] Storb, 2016, Somatic hypermutation and class switch recombination, Encycl Immunobiol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Together with the diversity introduced by V(D)J recombination, somatic hypermutation can produce an estimated 1012-1015 possible antibody sequences.

    “Together with the diversity introduced by V(D) J recombination, somatic hypermutation can produce an estimated 1012–1015 possible antibody sequences [23, 24].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [23] Briney, 2019, Commonality despite exceptional diversity in the baseline human antibody repertoire, Nature source

      Not checked in the free run.

    • cites [24] Glanville, 2009, Precise determination of the diversity of a combinatorial antibody library gives insight into the human immunoglobulin repertoire, Proc. Natl Acad Sci U S A source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    SabDab and Abysis allow retrieval of particular structures given a query antibody sequence (SAbDab) or by using more advanced features such as canonical classes of the CDRs (Abysis).

    “SabDab and Abysis allow retrieval of particular structures given a query antibody sequence (SAbDab) or by using more advanced features such as canonical classes of the CDRs (Abysis).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [16] Dunbar, 2013, SAbDab: the structural antibody database, Nucleic Acids Res source

      Not checked in the free run.

    • cites [47] Swindells, 2017, abYsis: integrated antibody sequence and structure—management, analysis, and prediction, J Mol Biol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Similarities between antibody amino acid sequences further allow for the creation of a standardized reference framework, or numbering scheme, giving each variable region amino acid an identifier.

    “Similarities between antibody amino acid sequences further allow for the creation of a standardized reference framework, or numbering scheme, giving each variable region amino acid an identifier [151].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [151] Dondelinger, 2018, Understanding the significance and implications of antibody numbering and antigen-binding surface/residue definition, Front Immunol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Binding affinity data is contained in resources such as SAbDab and PDBBind.

    “Such data is contained in resources such as SAbDab and PDBBind [54].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [54] Wang, 2005, The PDBbind database: methodologies and updates, J Med Chem source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Kabat and IMGT definitions are based on sequence alignments identifying conserved positions in the variable region whereas Chothia takes into account the 3D structure of the CDR loops.

    “Kabat and IMGT definitions are based on sequence alignments identifying conserved positions in the variable region [152, 153] whereas Chothia takes into account the 3D structure of the CDR loops.”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [152] Wu, 1970, An analysis of the sequences of the variable regions of Bence Jones proteins and myeloma light chains and their implications for antibody complementarity, J Exp Med source

      Not checked in the free run.

    • cites [153] Lefranc, 2011, IMGT unique numbering for the variable (V), constant (C), and groove (G) domains of IG, TR, MH, IgSF, and MhSF, Cold Spring Harb Protoc

      Not checked in the free run.

    • cites [32] Chothia, 1987, Canonical structures for the hypervariable regions of immunoglobulins, J Mol Biol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Other, more specialized, resources exist such as Ab-Bind, which hold data on 1101 mutations across 32 antibody complexes, and SKEMPI, which curates binding energy data for available structures but is not limited to antibody information only.

    “Other, more specialized, resources exist such as Ab-Bind [19], which hold data on 1101 mutations across 32 antibody complexes, and SKEMPI [55], which curates binding energy data for available structures but is not limited to antibody information only.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [19] Sirin, 2016, AB-Bind: antibody binding mutational database for computational affinity predictions, Protein Sci source

      Not checked in the free run.

    • cites [55] Moal, 2012, SKEMPI: a structural kinetic and energetic database of mutant protein interactions and its use in empirical models, Bioinformatics source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    Antibody-based therapeutics are developed via well-established processes that can be broadly categorized into Lead Identification and Lead Optimization.

    “Antibody-based therapeutics are developed via well-established processes that can be broadly categorized into Lead Identification and Lead Optimization.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Data on epitopes targeted by antibodies can be readily downloaded from the IEDB that now links such information to epitope-specific antibody sequences.

    “Data on epitopes targeted by antibodies can be readily downloaded from the IEDB that now links such information to epitope-specific antibody sequences [67].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [67] Mahajan, 2018, Epitope specific antibodies and T cell receptors in the immune epitope database, Front Immunol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Since the seminal work to define a standard numbering scheme for antibodies was carried out by Kabat in 1970, the Chothia and IMGT schemes have been adopted as the main alternatives.

    “Since the seminal work to define a standard numbering scheme for antibodies was carried out by Kabat in 1970 [152], the Chothia [32] and IMGT [153] schemes have been adopted as the main alternatives.”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [152] Wu, 1970, An analysis of the sequences of the variable regions of Bence Jones proteins and myeloma light chains and their implications for antibody complementarity, J Exp Med source

      Not checked in the free run.

    • cites [32] Chothia, 1987, Canonical structures for the hypervariable regions of immunoglobulins, J Mol Biol source

      Not checked in the free run.

    • cites [153] Lefranc, 2011, IMGT unique numbering for the variable (V), constant (C), and groove (G) domains of IG, TR, MH, IgSF, and MhSF, Cold Spring Harb Protoc

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    To the best of our knowledge, there are three freely available software packages to perform numbering of antibodies, ANARCI, Abnum and AbRSA, to act as the first step in computational antibody analysis such as homology modelling.

    “To the best of our knowledge, there are three freely available software packages to perform numbering of antibodies, ANARCI [83], Abnum [82] and AbRSA[81], to act as the first step in computational antibody analysis such as homology modelling.”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [83] Dunbar, 2015, ANARCI: Antigen receptor numbering and receptor classification, Bioinformatics source

      Not checked in the free run.

    • cites [82] Abhinandan, 2008, Analysis and improvements to Kabat and structurally correct numbering of antibody variable domains, Mol Immunol source

      Not checked in the free run.

    • cites [81] Li, 2019, AbRSA: a robust tool for antibody numbering, Protein Sci source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    CDRH3 is the most challenging region and is modelled to >5 Å RMSD in some targets.

    “AMA II reported that the overall accuracy of modelling the entire antibody FV is 1.1 Å Root Mean Square Deviation (RMSD) on average, with the most challenging region being the CDRH3, which is modelled to >5 Å RMSD in some targets.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [7] Almagro, 2014, Second antibody modeling assessment (AMA-II), Proteins Struct Funct Bioinforma source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Other resources such as the Immune Epitope Database (IEDB) link structural information to experimentally derived epitope data.

    “Other resources such as the Immune Epitope Database (IEDB) [18] link structural information to experimentally derived epitope data.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [18] Vita, 2010, The immune epitope database 2.0, Nucleic Acids Res source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Modelled antibody structures can be used at the Lead Identification stage to select surface exposed paratope residues for mutations or to characterize binding with respect to the cognate epitope.

    “Modelled structures can be used at the Lead Identification stage to select surface exposed paratope residues for mutations [110] or to characterize the binding with respect to the cognate epitope [119].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [110] Krawczyk, 2013, Antibody i-Patch prediction of the antibody binding site improves rigid local antibody–antigen docking, Protein Eng Des Sel source

      Not checked in the free run.

    • cites [119] Krawczyk, 2014, Improving B-cell epitope prediction and its application to global antibody–antigen docking, Bioinformatics source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    During Lead Identification and Optimization, molecules are assessed for unfavorable characteristics such as immunogenicity or poor biophysical properties.

    “During Lead Identification and Optimization, molecules are assessed for unfavorable characteristics such as immunogenicity or poor biophysical properties.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Arbitrary molecular surfaces appear to be indistinguishable from epitope regions, suggesting that predictions should be performed with reference to a particular antibody.

    “However, arbitrary molecular surfaces appear to be indistinguishable from epitope regions [167–169] suggesting that predictions should be performed with reference to a particular antibody [116, 118–120].”
    Quote in the paperSubstantial claim0 of 7 sources checked
    • cites [167] Kunik, 2013, The indistinguishability of epitopes from protein surface is explained by the distinct binding preferences of each of the six antigen-binding loops, Protein Eng Des Sel source

      Not checked in the free run.

    • cites [168] Kringelum, 2013, Structural analysis of B-cell epitopes in antibody: protein complexes, Mol Immunol source

      Not checked in the free run.

    • cites [169] Greenbaum, 2007, Towards a consensus on datasets and evaluation metrics for developing B-cell epitope prediction tools, J Mol Recognit source

      Not checked in the free run.

    • cites [116] Rapberger, 2007, Identification of discontinuous antigenic determinants on proteins based on shape complementarities, J Mol Recognit source

      Not checked in the free run.

    • cites [118] Sela-Culang, 2014, Using a combined computational-experimental approach to predict antibody-specific B cell epitopes, Structure source

      Not checked in the free run.

    • cites [119] Krawczyk, 2014, Improving B-cell epitope prediction and its application to global antibody–antigen docking, Bioinformatics source

      Not checked in the free run.

    • cites [120] Jespersen, 2019, Antibody specific B-cell epitope predictions: leveraging information from antibody–antigen protein complexes, Front Immunol source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Re-epitoping, pioneered by Ofran and collaborators, tests existing antibodies for complementarity to a target epitope and uses the best candidates to computationally construct focused surface display libraries.

    “A further approach to data mining existing structures to improve antibody affinity is ‘re-epitoping’, pioneered by Ofran and collaborators [184]. Here, existing antibodies are tested for complementarity to a target epitope and the best candidates are used to computationally construct focused surface display libraries.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [184] Nimrod, 2018, Computational design of epitope-specific functional antibodies, Cell Rep

      Not checked in the free run.

  • May combine two claims.

    Interpretationnone given

    Existing epitope prediction systems still do not provide reliable predictions, limiting their applicability in molecular docking.

    “The most challenging aspect is identification of the epitope since, despite the great efforts of the community in developing accurate epitope prediction methods, existing systems still do not provide reliable predictions, limiting their applicability in molecular docking.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The methods outlined offer the potential for discovering specific and selective binders computationally, reducing the experimental effort during the Lead Identification stage.

    “The methods outlined above offer the potential for discovering specific and selective binders computationally, reducing the experimental effort during the Lead Identification stage.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Characterization of the antibody-antigen interaction is important when filing therapeutic antibody patents.

    “From a legal perspective, characterization of the antibody–antigen interaction is of importance when filing therapeutic antibody patents [163].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [163] Deng, 2017, Enhancing antibody patent protection using epitope mapping information, MAbs source

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    Selected molecules from Lippow and co-workers had better affinity for the target, demonstrating that in some scenarios computational approaches alone can be used for affinity maturation.

    “Selected molecules had better affinity for the target, demonstrating that in some scenarios computational approaches alone can be used for affinity maturation.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [181] Lippow, 2007, Computational design of antibody-affinity improvement beyond in vivo maturation, Nat Biotechnol source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The majority of epitopes are conformational in nature, therefore predictors that use structural antigen information offer more accurate results than linear methods.

    “The majority of epitopes are conformational in nature therefore predictors that use structural antigen information offer more accurate results than linear methods [165, 166].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [165] Haste Andersen, 2006, Prediction of residues in discontinuous B-cell epitopes using protein 3D structures, Protein Sci source

      Not checked in the free run.

    • cites [166] Gao, 2014, Computational prediction of B cell epitopes from antigen sequences, Methods Mol Biol source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    Triangulation of results from paratope prediction, epitope prediction and antibody-antigen docking methods could provide a relatively fast and cost-effective route to obtaining reliable information on which to base rational antibody design decisions.

    “Therefore, as methods for antibody–antigen interaction prediction improve, it is to be expected that triangulation of results from paratope prediction, epitope prediction and antibody–antigen docking methods could provide a relatively fast and cost-effective route to obtaining reliable information on which to base rational antibody design decisions.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationreasoning

    Ab initio antibody design methods are still emerging and need to be validated across multiple projects in an industrial setting to assess their utility.

    “Ab inito methods such as these are still emerging and although some of them demonstrated the validity of their constructs experimentally there exists for them to be validated across multiple projects in industrial setting to assess their utility.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Raybould et al. stipulate five computational guidelines to define favorable biophysical properties in antibody therapeutics.

    “In this study Raybould et al. stipulate five computational guidelines to define favorable biophysical properties in antibody therapeutics.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [13] Raybould, 2019, Five computational developability guidelines for therapeutic antibody profiling, Proc Natl Acad Sci U S A source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    Identification of hydrophobic regions is an important step in aggregation prediction that ideally requires a crystal structure of the antibody or a reliable homology model.

    “Identification of hydrophobic regions is an important step in aggregation prediction that ideally requires a crystal structure of the antibody or a reliable homology model.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    Score values diverging significantly from the natural distribution are highlighted and associated sequences flagged for developability risk.

    “Score values diverging significantly from the natural distribution are highlighted and the associated sequences flagged for developability risk.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [13] Raybould, 2019, Five computational developability guidelines for therapeutic antibody profiling, Proc Natl Acad Sci U S A source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The field currently benefits from a steady stream of new data from NGS of B-cell Receptors and increasing resources for downloading such data.

    “In this respect, the field currently benefits from a steady stream of new data from NGS of B-cell Receptors (BCRs, to be used as proxy for antibodies) [212, 213] and there is an increasing number of resources available for downloading such data [17], which are being used to analyze therapeutic antibodies [13].”
    Quote in the paperNot a substantial claim0 of 4 sources checked
    • cites [212] Georgiou, 2014, The promise and challenge of high-throughput sequencing of the antibody repertoire, Nat Biotechnol source

      Not checked in the free run.

    • cites [213] Rubelt, 2017, Adaptive Immune Receptor Repertoire Community recommendations for sharing immune-repertoire sequencing data, Nat Immunol source

      Not checked in the free run.

    • cites [17] Kovaltsuk, 2018, Observed antibody space: a resource for data mining next-generation sequencing of antibody repertoires, J Immunol source

      Not checked in the free run.

    • cites [13] Raybould, 2019, Five computational developability guidelines for therapeutic antibody profiling, Proc Natl Acad Sci U S A source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The Adaptive Boosting model for aggregation prediction was trained and validated on a dataset of 500 antibodies with calculated biophysical properties.

    “The model was trained and validated on a dataset of 500 antibodies with calculated biophysical properties.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [15] Obrezanova, 2015, Aggregation risk prediction for antibodies and its application to biotherapeutic development, MAbs source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The connection between computational predictions of immune epitopes and ADAs generated against biotherapeutics remains largely untested.

    “We must emphasize that connection between computational predictions of immune epitopes and ADAs generated against biotherapeutics remains largely untested.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Simple models based on abundance or sequence similarity can identify sequences that are not antigen-specific or miss less abundant antigen-specific sequences.

    “Such simple models however can still identify sequences that are not antigen-specific or miss less abundant antigen-specific sequences [215].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [215] Galson, 2015, Analysis of B cell repertoire dynamics following hepatitis B vaccination in humans, and enrichment of vaccine-specific antibody sequences, EBioMedicine source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Current bioinformatic analyses of NGS repertoires focus on large-scale decoding of immune responses, with several potential applications for therapeutic design.

    “Current bioinformatic analyses of NGS repertoires focus on large-scale decoding of the immune responses, with several potential applications for therapeutic design [22, 25, 214].”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [22] Miho, 2018, Computational strategies for dissecting the high-dimensional complexity of adaptive immune repertoires, Front Immunol source

      Not checked in the free run.

    • cites [25] Brown, 2019, Augmenting adaptive immunity: progress and challenges in the quantitative engineering and analysis of adaptive immune receptor repertoires source

      Not checked in the free run.

    • cites [214] Miho, 2019, Large-scale network analysis reveals the sequence space architecture of antibody repertoires, Nat Commun source

      Not checked in the free run.

  • May combine two claims.

    Premisereasoning

    Germline sequences only offer a limited view of overall mutational antibody diversity, which can be addressed by computational humanization comparing the animal-derived therapeutic to the distribution of amino acids in human antibody sequences.

    “Germline sequences however only offer a limited view of overall mutational antibody diversity, which can be addressed by computational humanization, comparing the animal-derived therapeutic to the distribution of amino acids in human antibody sequences (Figure 2D, Table 2E).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationnone given

    Measures such as the DI and the aggregation propensity risk score rely on hydrophobicity scales and charge annotations, and these parameters alone contain useful information.

    “Measures such as the DI and the aggregation propensity risk score rely on the hydrophobicity scales and charge annotations, demonstrating that there is useful information in these parameters alone.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Generation of ADAs is a multi-factorial issue and depends upon patients' genetic background and disease history as well as quality attributes of the protein therapeutics, particularly presence of aggregates and other degradants even in very minute quantities.

    “Generation of ADAs is a multi-factorial issue and depends upon patients genetic background and disease history as well as quality attributes of the protein therapeutics, particularly, presence of aggregates and other degradants even in very minute quantities [190, 191].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [190] Jiskoot, 2016, Mouse models for assessing protein immunogenicity: lessons and challenges, J Pharm Sci source

      Not checked in the free run.

    • cites [191] Singh, 2011, Impact of product-related factors on immunogenicity of biotherapeutics, J Pharm Sci source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    An alternative approach to developability pre-selection is to use natural antibody sequences under the assumption that they have favorable biophysical properties.

    “An alternative approach is to use natural antibody sequences under the assumption that they have favorable biophysical properties [13].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [13] Raybould, 2019, Five computational developability guidelines for therapeutic antibody profiling, Proc Natl Acad Sci U S A source

      Not checked in the free run.

    • The authors state this as an assumption: Natural antibody sequences have favorable biophysical properties.

  • Classified as probably not a substantial claim.

    Premisecitation

    Among Raybould et al.'s guidelines, a structure-based hydrophobicity score is calculated and compared to the distribution of the same score in naturally sourced NGS sequences.

    “Among these a structure-based hydrophobicity score is calculated and the value is then compared to the distribution of the same score in naturally sourced NGS sequences.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [13] Raybould, 2019, Five computational developability guidelines for therapeutic antibody profiling, Proc Natl Acad Sci U S A source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    Development of computational methods to aid antibody engineering relies on successful exploration and exploitation of new data sources.

    “Development of computational methods to aid antibody engineering relies on successful exploration and exploitation of new data sources.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Humanized and even fully human antibodies can elicit immune responses among patients receiving such medicines and generate ADAs against them.

    “Indeed, humanized and even fully human antibodies can elicit immune responses among the patients receiving such medicines and generate ADAs against them.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Lauer and co-workers derived a developability index (DI) from a 2-year measurement of biophysical properties for 12 antibodies and demonstrated that it correlated well with favorable biophysical properties of their antibodies.

    “Lauer and co/workers carried out a 2-year long measurement of biophysical properties for 12 antibodies [148] from which they derived a score, the developability index (DI), and demonstrated that it correlated well with the favorable biophysical properties of their antibodies.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [148] Lauer, 2012, Developability index: a rapid in silico tool for the screening of antibody aggregation propensity, J Pharm Sci source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The absence of the L chain means that nanobody CDRs are distinct from those of antibodies in both sequence and structure, and as a result nanobodies are able to bind antibody-inaccessible epitopes in enzyme active sites, viral capsids and G protein coupled receptors.

    “The absence of the L chain means that the nanobody’s CDRs are distinct from those of antibodies in both sequence and structure, and as a result nanobodies are able to bind antibody-inaccessible epitopes in enzyme active sites, viral capsids and G protein coupled receptors [233, 234].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [233] Staus, 2016, Allosteric nanobodies reveal the dynamic range and diverse mechanisms of G-protein-coupled receptor activation, Nature source

      Not checked in the free run.

    • cites [234] Steyaert, 2011, Nanobody stabilization of G protein-coupled receptor conformational states, Curr Opin Struct Biol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    Increasing amount of data from next generation sequencing holds the potential to improve the computational methods and thus its applicability to therapeutic design.

    “Increasing amount of data from next generation sequencing holds the potential to improve the computational methods and thus its applicability to therapeutic design.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Nanobody paratopes contain on average nearly three additional residues, and the paratope is drawn from a wider set of aligned sequence positions than classical antibody VH domain paratopes, comparable to the set used by the entire VH-VL paratope.

    “Perhaps more significant in terms of challenges to computational modelling tools is that nanobody paratopes contain on average nearly three additional residues, and moreover the paratope is drawn from a wider set of aligned sequence positions than classical antibody VH domain paratopes, comparable to the set used by the entire VH-VL paratope [230, 232].”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [230] Muyldermans, 2013, Nanobodies: natural single-domain antibodies, Annu Rev Biochem. source

      Not checked in the free run.

    • cites [232] Mitchell, 2018, Comparative analysis of nanobody sequence and structure data, Proteins Struct Funct Bioinforma source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    Existing computational antibody design knowledge and tools may benefit emerging biotherapeutic modalities akin to antibodies, such as nanobodies.

    “Existing computational antibody design knowledge and tools may benefit emerging biotherapeutic modalities akin to antibodies, such as nanobodies.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    Predicting interface information can provide an initial guide for experimental epitope mapping efforts or offer a starting point for a therapeutic campaign by providing the basis for focused surface display libraries to design a novel antibody binder for a given epitope.

    “Predicting interface information can provide an initial guide for experimental epitope mapping efforts or offer a starting point for a therapeutic campaign by providing the basis for focused surface display libraries to design a novel antibody binder for a given epitope.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Nanobodies are naturally occurring H chain only antibodies in camelids and sharks.

    “One of the most promising is the H chain only antibody, or nanobody that are naturally occurring in camelids (llamas, alpacas and camels) [27] and sharks [228, 229] (reviewed by Muyldermans [230] and Bannas et al. [27]).”
    Quote in the paperNot a substantial claim0 of 4 sources checked
    • cites [27] Bannas, 2017, Nanobodies and nanobody-based human heavy chain antibodies as antitumor therapeutics, Front Immunol source

      Not checked in the free run.

    • cites [228] Feige, 2014, The structural analysis of shark IgNAR antibodies reveals evolutionary principles of immunoglobulins, Proc Natl Acad Sci source

      Not checked in the free run.

    • cites [229] Griffiths, 2013, Shark variable new antigen receptor (VNAR) single domain antibody fragments: stability and diagnostic applications, Antibodies source

      Not checked in the free run.

    • cites [230] Muyldermans, 2013, Nanobodies: natural single-domain antibodies, Annu Rev Biochem. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Established bioinformatics methods such as protein structural homology modelling and molecular docking can be readily applied to the specific case of antibodies in a therapeutic setting.

    “Established bioinformatics methods such as protein structural homology modelling and molecular docking can be readily applied to the specific case of antibodies in a therapeutic setting.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Despite the similarity between antibodies and nanobodies, systematic benchmarking will still be needed to determine whether development of nanobodies can benefit from computational antibody protocols in their current form or whether they need to be adjusted accordingly.

    “However, despite the similarity between antibodies and nanobodies, systematic benchmarking will still be needed to determine whether development of nanobodies can benefit from computational antibody protocols in their current form or whether they need to be adjusted accordingly.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    Antibody-specific benchmarking challenges will emphasize the shortcomings and advantages of each method and enable improvements to be developed in a focused manner, specifically with regard to their utility in therapeutic development process.

    “Antibody-specific benchmarking challenges will emphasize the shortcomings and advantages of each method and enable improvements to be developed in a focused manner, specifically with regard to their utility in therapeutic development process.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationcitation

    Benchmarking of bioinformatic antibody methods on a par with existing protein-generic initiatives such as CASP, CAPRI or CAMEO will benefit the entire computational antibody field.

    “Holistically, benchmarking of bioinformatic antibody methods on a par with existing protein-generic initiatives such as CASP, CAPRI or CAMEO [237] will benefit the entire computational antibody field.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [237] Haas, 2018, Continuous Automated Model EvaluatiOn (CAMEO) complementing the critical assessment of structure prediction in CASP12, Proteins Struct Funct Bioinforma source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationnone given

    The largest added value of computational methods will continue to be in providing time and cost-efficient ways of guiding experimental methods.

    “Indeed, their largest added value will continue to be in providing time and cost-efficient ways of guiding experimental methods.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Current approaches to bring antibody therapeutics to market have remained experimentally focused, with animal immunization and surface display technologies accounting for the majority of molecules developed to date.

  • May combine two claims.

    Premisecitation

    Analysis of antibodies derived from 600 donors followed by construction of a library based on natural positional preferences has demonstrated that such constraints can be adapted to library construction and used to identify binding antibodies.

    “This has already been demonstrated to some extent through analysis of antibodies derived from 600 donors [24] followed by construction of a library based on natural positional preferences [26].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [24] Glanville, 2009, Precise determination of the diversity of a combinatorial antibody library gives insight into the human immunoglobulin repertoire, Proc. Natl Acad Sci U S A source

      Not checked in the free run.

    • cites [26] Zhai, 2011, Synthetic antibodies designed on natural sequence landscapes, J Mol Biol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    The increasing amount of antibody-specific data in the public domain facilitates the maturation of computational antibody design methods, resulting in a growing uptake as part of standard pharmaceutical discovery processes.

    “The increasing amount of antibody-specific data in the public domain facilitates the maturation of computational antibody design methods, resulting in a growing uptake as part of standard pharmaceutical discovery processes.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    CDRH3 loop residues that are highly variable in both sequence and structure dominate antigen-contacting residues within nanobodies, suggesting that the nanobody-antigen interface will be difficult to model using tools that have been developed in the context of classical antibody VH domains.

    “CDRH3 loop residues that are highly variable in both sequence and structure dominate antigen-contacting residues within nanobodies, suggesting that the nanobody-antigen interface will be difficult to model using tools that have been developed in the context of classical antibody VH domains.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    Further progress in the development of antibody-specific computational tools will be associated with access to more and diverse data in the public domain.

    “Further progress in the development of antibody-specific computational tools will be associated with access to more and diverse data in the public domain.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationnone given

    Computational antibody methods might be transferable to other immunoglobulin formats such as nanobodies, which have more potential in certain therapeutic areas.

    “Computational antibody methods might be transferable to other immunoglobulin formats such as nanobodies, which have more potential in certain therapeutic areas.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Exploiting the vast amount of data generated by NGS will facilitate the derivation of more reliable 'humanness' and 'developability' profiles with which to guide antibody therapeutic discovery.

    “Exploiting the vast amount of data generated by NGS will facilitate the derivation of more reliable ‘humanness’ and ‘developability’ profiles with which to guide antibody therapeutic discovery.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    Systematic benchmarking of the tools can emphasize the advantages of computational methods and where these can be used to support therapeutic pipelines.

    “Systematic benchmarking of the tools can emphasize the advantages of computational methods and where these can be used to support therapeutic pipelines.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Structural modelling can offer insight on exposed residues to be used for mutagenesis to either optimize binding, reduce immunogenicity or provide information on hydrophobicity patches related to detrimental biophysical properties.

    “Structural modelling can offer insight on exposed residues to be used for mutagenesis to either optimize binding, reduce immunogenicity or provide information on hydrophobicity patches related to detrimental biophysical properties.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    Further development of bioinformatic methods for NGS analysis will depend as much on improving the algorithms as on the quality of the data itself.

    “At this stage, further development of bioinformatic methods for NGS analysis will depend as much on improving the algorithms as on the quality of the data itself.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Nanobodies exhibit less sequence and structure variation across their framework regions, and similar levels of sequence variation as classical antibodies within the CDRH1 and CDRH2 loops.

    “Specifically, nanobodies were found to exhibit less sequence and structure variation across their framework regions, and similar levels of sequence variation as classical antibodies within the CDRH1 and CDRH2 loops [232].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [232] Mitchell, 2018, Comparative analysis of nanobody sequence and structure data, Proteins Struct Funct Bioinforma source

      Not checked in the free run.

84 of 84

  • The paper relies on this without support.

    Experimentally determining the 3D structure of the antibody–antigen complex using X-ray crystallography is the gold standard for obtaining atomic-level epitope–paratope interaction information.

    “The ‘gold standard’ for obtaining this information is by experimentally determining the 3D structure of the antibody–antigen complex using X-ray crystallography.”

  • The paper relies on this without support.

    Experimental structural methods are very time and resource consuming and their success is not guaranteed.

    “Experimental methods can be very time and resource consuming with success not being guaranteed.”

  • The paper relies on this without support.

    Understanding epitope–paratope interactions at the atomic level is key to rational development of effective therapeutics.

    “Understanding the epitope–paratope interactions at the atomic level is key to rational development of effective therapeutics.”

  • The authors state this as an assumption.

    Natural antibody sequences have favorable biophysical properties.

    “An alternative approach is to use natural antibody sequences under the assumption that they have favorable biophysical properties [13].”

  • The paper relies on this without support.

    Most NGS datasets produced to date do not offer paired H and L chain sequences.

    “In addition, most of the NGS datasets produced to date do not offer paired H and L chain sequences.”

  • The paper relies on this without support.

    The VL domain is much less structurally variable.

    “Given that the VL domain is much less structurally variable, this suggests that the application of computational structural modelling tools to nanobody–antigen interactions will require innovation.”

  • The paper relies on this without support.

    Current approaches to bring antibody therapeutics to market have remained experimentally focused, with animal immunization and surface display technologies accounting for the majority of molecules developed to date.

    “Current approaches to bring these molecules to the market have remained experimentally focused, with animal immunization and surface display technologies accounting for the majority of molecules developed to date.”
7 of 7

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Table 2E is cited but not present

    Minor

    The text directs readers to Table 2E for humanization methods, but Table 2 has no E section; its humanization entries appear under the D heading.

1 of 1

Cites

Flagged first, then by citations

1 of the 242 references couldn’t be matched to a known paper.

Research works printed but not linked: 3

  1. Kindt, 2007, Kuby Immunology
  2. Inc. CCG, 2016, Molecular Operating Environment (MOE), 2016.08
  3. MacKerel, 1998, CHARMM: the energy function and its parameterization with an overview of the program, Encycl Comput Chem, 1, 271

Cited by

Sorted by citations

Related

Authors
Norman, R. A., Ambrosetti, F., Bonvin, A. M. J. J., Colwell, L. J., Kelm, S., Kumar, S., & Krawczyk, K.
Affiliations
  • Pistoia Alliance Inc., USA
  • Utrecht University (NL)
  • University of Cambridge (GB)
  • UCB Pharma (United Kingdom) (GB)
  • Boehringer Ingelheim (United States) (US)
  • NaturalAntibody, Germany
Corresponding author
Krawczyk, Konrad
ORCID
Venue
Briefings in Bioinformatics
Publisher
Oxford University Press (OUP)
Published
2019-10-18
Volume
21
Issue
5
Pages
1549-1567
ISSN
  • 1467-5463 (ISSN-L)
  • 1477-4054
DOI
10.1093/bib/bbz095
PMC
PMC7947987
OpenAlex
W2980681402
Semantic Scholar
aeee70fa6bd5b6d3659d9486f33655afbadbf02c
Cited by
219
References
242
Citations by year
2026: 24 · 2025: 43 · 2024: 31 · 2023: 35 · 2022: 37 · 2021: 29 · 2020: 18 · 2019: 1 · 2012: 1
FWCI
3.18
Type
journal article
Language
en
Open access
Licence
Locations
  • Briefings in Bioinformatics (publishedVersion, public-domain)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Utrecht University Repository (Utrecht University) (submittedVersion, other-oa)
  • IRIS Research product catalog (Sapienza University of Rome) (submittedVersion)
Funders
  • European Union Horizon 2020 BioExcel (675728, 823830)
  • EOSC-hub (777536)
  • Simons Foundation
Topics
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Monoclonal and Polyclonal Antibodies Research
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › vaccines and immunoinformatics approaches
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
Keywords
therapeutic antibody design, antibody engineering, monoclonal antibodies, nanobodies, antibody database, protein modeling, antibody-antigen binding, biologic therapy
MeSH terms
Antibodies, Monoclonal, Protein Conformation, Computational Biology, Databases, Protein, Molecular Docking Simulation

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review