Table 4

Comparison of sequence and structure design. The quantitative comparison of performance in CDR design, refer to Figure 7. In the ‘Training Dataset’ column the number of sequences (seq.) and structures (struct.) used for training is shown.

NameClassModelTraining
Dataset
DescriptionRef
AbDiffuserAntibodyDDPMpOAS
(105k seq.)
HER2 [104]
(9k struct.)
Strengths: Can handle variable length sequences.
 Limitations: Does not consider the antigen or the epitope.
 Applications: Full Ab 3D structure and sequence design of variable length.
[101]
DiffAbAntibodyDDPMSAbDabStrengths: Side-chains orientations design.
 Limitations: Relies on an Ab framework bound to the target Ag.
 Applications: Sequence-structure co-design, sequence design of CDRs for given backbone structures, and Ab optimization.
[100]
EAGLEAntibodyDDPMOAS
(100M seq.)
SAbDab
(8k struct.)
Strengths: Use of sequence embedding and CLIP models with Ag structure.
 Limitations: CLIP has limited impact on model performance; requires knowledge of both the antigen and the epitope.
 Applications: Ab sequence designed conditioned on the Ag structure.
[102]
FvHallucinatorAntibodyDeepAbAbDb
abYbank
(11k struct.)
Strengths: Designs substitutions highly enriched in human repertoire; integrated folding model.
 Limitations: Does not consider the Ag and optimization.
 Applications: Generate libraries of Ab sequences with fixed structure.
[96]


[105]
RefineGNNAntibodyGNNSAbDabStrengths: Modifies a generated subgraph to accommodate new residues.
 Limitations: Does not consider the epitope.
 Applications: Sequence and structure co-design of CDRs with enhanced binding specificity or neutralization capabilities.
[93]
NameClassModelTraining
Dataset
DescriptionRef
AbDiffuserAntibodyDDPMpOAS
(105k seq.)
HER2 [104]
(9k struct.)
Strengths: Can handle variable length sequences.
 Limitations: Does not consider the antigen or the epitope.
 Applications: Full Ab 3D structure and sequence design of variable length.
[101]
DiffAbAntibodyDDPMSAbDabStrengths: Side-chains orientations design.
 Limitations: Relies on an Ab framework bound to the target Ag.
 Applications: Sequence-structure co-design, sequence design of CDRs for given backbone structures, and Ab optimization.
[100]
EAGLEAntibodyDDPMOAS
(100M seq.)
SAbDab
(8k struct.)
Strengths: Use of sequence embedding and CLIP models with Ag structure.
 Limitations: CLIP has limited impact on model performance; requires knowledge of both the antigen and the epitope.
 Applications: Ab sequence designed conditioned on the Ag structure.
[102]
FvHallucinatorAntibodyDeepAbAbDb
abYbank
(11k struct.)
Strengths: Designs substitutions highly enriched in human repertoire; integrated folding model.
 Limitations: Does not consider the Ag and optimization.
 Applications: Generate libraries of Ab sequences with fixed structure.
[96]


[105]
RefineGNNAntibodyGNNSAbDabStrengths: Modifies a generated subgraph to accommodate new residues.
 Limitations: Does not consider the epitope.
 Applications: Sequence and structure co-design of CDRs with enhanced binding specificity or neutralization capabilities.
[93]
Table 4

Comparison of sequence and structure design. The quantitative comparison of performance in CDR design, refer to Figure 7. In the ‘Training Dataset’ column the number of sequences (seq.) and structures (struct.) used for training is shown.

NameClassModelTraining
Dataset
DescriptionRef
AbDiffuserAntibodyDDPMpOAS
(105k seq.)
HER2 [104]
(9k struct.)
Strengths: Can handle variable length sequences.
 Limitations: Does not consider the antigen or the epitope.
 Applications: Full Ab 3D structure and sequence design of variable length.
[101]
DiffAbAntibodyDDPMSAbDabStrengths: Side-chains orientations design.
 Limitations: Relies on an Ab framework bound to the target Ag.
 Applications: Sequence-structure co-design, sequence design of CDRs for given backbone structures, and Ab optimization.
[100]
EAGLEAntibodyDDPMOAS
(100M seq.)
SAbDab
(8k struct.)
Strengths: Use of sequence embedding and CLIP models with Ag structure.
 Limitations: CLIP has limited impact on model performance; requires knowledge of both the antigen and the epitope.
 Applications: Ab sequence designed conditioned on the Ag structure.
[102]
FvHallucinatorAntibodyDeepAbAbDb
abYbank
(11k struct.)
Strengths: Designs substitutions highly enriched in human repertoire; integrated folding model.
 Limitations: Does not consider the Ag and optimization.
 Applications: Generate libraries of Ab sequences with fixed structure.
[96]


[105]
RefineGNNAntibodyGNNSAbDabStrengths: Modifies a generated subgraph to accommodate new residues.
 Limitations: Does not consider the epitope.
 Applications: Sequence and structure co-design of CDRs with enhanced binding specificity or neutralization capabilities.
[93]
NameClassModelTraining
Dataset
DescriptionRef
AbDiffuserAntibodyDDPMpOAS
(105k seq.)
HER2 [104]
(9k struct.)
Strengths: Can handle variable length sequences.
 Limitations: Does not consider the antigen or the epitope.
 Applications: Full Ab 3D structure and sequence design of variable length.
[101]
DiffAbAntibodyDDPMSAbDabStrengths: Side-chains orientations design.
 Limitations: Relies on an Ab framework bound to the target Ag.
 Applications: Sequence-structure co-design, sequence design of CDRs for given backbone structures, and Ab optimization.
[100]
EAGLEAntibodyDDPMOAS
(100M seq.)
SAbDab
(8k struct.)
Strengths: Use of sequence embedding and CLIP models with Ag structure.
 Limitations: CLIP has limited impact on model performance; requires knowledge of both the antigen and the epitope.
 Applications: Ab sequence designed conditioned on the Ag structure.
[102]
FvHallucinatorAntibodyDeepAbAbDb
abYbank
(11k struct.)
Strengths: Designs substitutions highly enriched in human repertoire; integrated folding model.
 Limitations: Does not consider the Ag and optimization.
 Applications: Generate libraries of Ab sequences with fixed structure.
[96]


[105]
RefineGNNAntibodyGNNSAbDabStrengths: Modifies a generated subgraph to accommodate new residues.
 Limitations: Does not consider the epitope.
 Applications: Sequence and structure co-design of CDRs with enhanced binding specificity or neutralization capabilities.
[93]
Close
This Feature Is Available To Subscribers Only

Sign In or Create an Account

Close

This PDF is available to Subscribers Only

View Article Abstract & Purchase Options

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

Close