Clinical antibodies have become the standard of care for multiple conditions, extending and improving the quality of life for millions of patients, from cancer and autoimmune diseases to emerging disease-modifying therapies in neurological conditions. Protein aggregation is not only a concern for their developability, influencing the yield and cost of the final product, but aggregated therapeutics display reduced biological activity and can induce unwanted immune responses, ultimately affecting patient safety. Past case studies have shown the A3D and its latest update, A4D, as valid tools to detect and control aggregation in antibodies. These tools use the protein 3D structures to identify solvent-exposed, spatially proximal residues that create the aggregation-driving patches. The A4D-Ab database reports the aggregation propensity for over 2,400 structures from more than 300 clinical antibodies, allowing systematic explorations of the aggregation propensity of therapeutic antibodies.
The A4D-Ab database utilizes the Aggrescan4D (A4D) aggregation scoring metrics. Using the antibodies' available 3D structures to identify residues that are solvent-exposed and accounting for proximity on the protein surface, capturing the spatial context that sequence-only methods inevitably miss.
A4D estimates the influence of the 3D context and pH on the charge state and hydrophobicity of ionizable acids, improving the understanding of pH-dependent aggregation propensity and allowing a more precise detection of problematic residues for next-generation therapeutics. pH awareness also helps define the pH or pH-rationale ranges for optimal experimental and formulation strategies.
A4D-Ab collects experimentally solved structures for clinical antibodies with >95% sequence identity and reports the aggregation propensity not just of a single structure, but of the structurally derived protein ensembles - allowing the identification of lower-population structural aggregation patches. Each entry is enriched with metadata (therapeutic area, target, isotype, clinical status) and, following FAIR principles, linked to established databases and to the companies sponsoring their clinical development.
The database is structured as three progressive levels: (1) multiple antibodies, (2) a summary of all predictions for a specific antibody, and (3) predictions for a specific structure.
An overview of the clinical antibody structurome, allowing comparison of clinical-grade antibodies by target, therapeutic area, clinical status, and aggregated A4D metrics. Use filters, search, and sortable columns to identify candidates of interest.
A dedicated view for each antibody capturing its structural heterogeneity. Since a single antibody may have multiple experimental structures, this section enables direct comparison of how different conformations affect solubility. It includes metadata to understand the clinical context - therapeutic area, target, clinical status, company - as well as direct links to gold-standard antibody and drug resources. This level also reports average A4D metrics for all structures, as well as the antibody formulation pH (when disclosed).
The A4D metrics have been reworked for in-depth analysis of the spatial, sequential and pH-dependent distribution of aggregation propensity. This level incorporates a novel plot of aggregation-prone amino acids, the changes in aggregation propensity with pH, and a reworked structural view of the results. Since CDRs contribute significantly to antibody aggregation propensity, A4D-Ab reports scores for the full antibody and its CDR regions.
Navigating the Database
Understanding the Structure Page
A global measure of aggregation propensity for the whole structure. Lower (more negative) scores indicate decreased propensity for aggregation. Positive scores indicate regions at elevated risk of aggregation.
The predicted pH values at which the antibody shows the lowest and the highest aggregation propensity - relevant for formulation buffer design.
Measures how closely a structural model matches the target antibody sequence. A4D-Ab reports 3 recalculated levels accounting for 95–98%, 99%, or 100%, depending on the source PDB structure used.
Get a list of all available antibodies.
[{'antibody_id': 'A4DAB057', 'antibody_name': 'Canakinumab'},
{'antibody_id': 'A4DAB049', 'antibody_name': 'Bintrafusp'}, ...]
Get details for a specific antibody. Use the antibody_id from the list endpoint. Returns antibody metadata and a list of structures (with jid for each).
SingleAntibodyFullData
├── antibody: AntibodyHighLevel
│ ├── antibody_id: str
│ ├── name: str
│ ├── target: str | None
│ ├── domain: str | None
│ ├── clinical_status: str | None
│ ├── subclass: str | None
│ ├── isotype: str | None
│ ├── company: str | None
│ ├── indication: str | None
│ ├── synonyms: str | None
│ ├── structures_count: int
│ ├── max_seq_identity: str
│ ├── aggregate_a4d_score: float
│ └── external_ids: ExternalIds
│ ├── imgt: str | None
│ ├── abcd: str | None
│ ├── pubchem: list[str] | None
│ └── chembl: list[str] | None
│
└── structures: list[StructureHighLevel]
└── StructureHighLevel
├── structure_id: str
├── jid: str
├── pdb_id: str | None
├── seq_identity: float | None
├── avg_a4d_score: float
├── ph_max_solubility: float
├── score_at_ph_max: float
├── ph_min_solubility: float
├── score_at_ph_min: float
├── a4d_score_for_cdr: float | None
├── a4d_score_for_non_cdr: float | None
├── ph_formulation: float | None
└── score_at_ph_formulation: float | None
Get full details for a specific structure (corresponds to Level 3 in the UI). Use the jid from the antibody endpoint.
LowLevelData ├── structure_id: str ├── jid: str ├── antibody_id: str ├── antibody_name: str ├── pdb_id: str ├── chains: list[str] ├── seq_identity: str ├── started: str ├── sequence: str │ ├── table: AggrescanTable │ ├── min: float │ ├── max: float │ ├── avg: float │ ├── sum: float │ └── tab: list[tuple[str, str, str, str]] │ ├── ph_scores: PhTable │ └── ph_scores: dict[str, list[float]] │ ├── cdr_data: CdrData │ ├── hchain: str │ ├── lchain: str │ ├── HCDR1: list[int] | None │ ├── HCDR2: list[int] | None │ ├── HCDR3: list[int] | None │ ├── LCDR1: list[int] | None │ ├── LCDR2: list[int] | None │ └── LCDR3: list[int] | None │ ├── a4d_score_for_cdr: float | None ├── a4d_score_for_non_cdr: float | None ├── h_chain_score_cdr: float | None ├── h_chain_score_non_cdr: float | None ├── l_chain_score_cdr: float | None └── l_chain_score_non_cdr: float | None
Download all result files for a structure as a .tar.gz archive. Example Python script:
import requests
JID = "f248514cc791da"
URL = f"https://lcbio.pl/a4d/A4D-Ab/api/antibody/structure/data/get/{JID}"
OUTPUT_PATH = "structure_data.tar.gz"
with requests.get(URL, stream=True) as response:
response.raise_for_status()
with open(OUTPUT_PATH, "wb") as f:
for chunk in response.iter_content(chunk_size=1024 * 1024):
if chunk:
f.write(chunk)
print(f"Saved to: {OUTPUT_PATH}")
A4D-Ab is a collaborative effort between the Protein Folding and Conformational Diseases Laboratory at the Universitat Autònoma de Barcelona, the Bioinformatics and Multiomics Analysis Laboratory at the Medical University of Białystok and the Laboratory of Computational Biology at the University of Warsaw.
Both A4D-Ab and the Aggrescan4D method are described in dedicated publications, please cite them when using data from this resource: A4D publication (Nucleic Acids Research, 2024)