ABodyBuilder-ML

Machine Learning based antibody Fv modelling

This web application is designed for small-scale structural modelling. If you would like to perform large-scale analyses, please contact: opig@stats.ox.ac.uk

• ABodyBuilder-ML can be used to produce models of the antibody Fv regions.
• Submit heavy and light chain sequences to the forms below and ABodyBuilder-ML will:
• Number the sequences with ANARCI
• Choose the best templates for the VH and VL domains separately using SAbDab
• Predict the VH-VL orientation using ABangle and place the template domains in the predicted pose
• Model the CDR loops with ABlooper(*)
• Generate side chains with PEARS
• Renumber the final models with ANARCI
• Annotate the estimated loop model quality, recognise potential sequence liabilities and visualise different regions of the structure.
• An example of the output produced by ABodyBuilder-ML can be seen here.

(*)NB: ABodyBuilder-ML uses the deep-learning approach to CDR loop modelling of ABlooper, replacing template-based modelling previously done with FREAD and Sphinx by the former ABodyBuilder pipeline. For a full description of the original ABodyBuilder pipeline, see this publication

To cite ABlooper, please refer to Abanades B, et al. ABlooper: fast accurate antibody CDR loop structure prediction with accuracy estimation, Bioinformatics (2022)

IMPORTANT: ABodyBuilder-ML will become deprecated in the near future, as ABodyBuilder2 is now our main tool for antibody structure modelling, which will affect other tools such as TAP and Hu-mAb. To receive notifications for any of these updates, follow our Twitter account.

Heavy chain sequence:

load example

Light chain sequence:

load example

Job name:

my_fv_model

Specify a PDB blacklist?

Choose a numbering scheme to annotate the final models with:

IMGT Kabat Chothia Martin

Model