We created a webtool, AbDiver to facilitate exploration of BCR diversity from AIRR/NGS datasets. The main aim is to provide an overview of ‘natural diversity’ with respect to a single query amino acid sequence of an antibody that one might need to engineer (e.g. at lead optimization stage).
The server allows for:
Positional profiling of a query sequence using AA frequencies triangulated from multiple independent studies. Your query sequence is annotated with positional AA frequency statistics aggregated from multiple AIRR/NGS studies (e.g. same 2nd and 3rd frequent amino acids). This is regardless of different individuals, sequencing techniques across independent studies, to emphasize repeatable patterns rather than study-specific biases.
Find sequence-similar V-region and clonotypes (provide CDR3+germline). This is to facilitate discovery naturally-occurring sequences similar to your lead candidate (like here).
Underlying data for this public service are the freely available sequences from OAS.
We were doing benchmarking of the system on its usability at optimization stage, how much information one can find when faced with engineering-stage Ab - when the manuscript goes through all the hoops will post here… In the meantime thanks a lot for checking it out.
Preempting an already-posed question - is it a humanization tool using NGS? Not really, it is a tool to explore the natural diversity of antibodies with implications to immunogenicity, stability, specificity and beyond. On this, there was some really nice work on employing OAS specifically for humanization using ML here and here. But we were asked to add ‘align to organism of choice’ buttons for completeness.