Using the abYsis database for drug discovery

Posted 30th April 2021 by Nick Noakes

Professor Andrew Martin, UCL, speaking at the 4th Global Pharma R&D AI, Data Science and Informatics Summit, described using the abYsis database and workbench for drug discovery. He showed how it is possible to explore an annotate antibody sequence and structure, including comparisons with observed residue distributions. The database can also aid with humanization by making sequences more human and library design.

Antibodies are becoming hugely important as drugs, half of the drugs named by the WHO-INN are biologics, and approximately half of those are antibodies. In November 2020, in the U.S. or Europe, there were 92 approved antibody-based drugs, plus 17 in review. And another seven have been withdrawn. There are 197 in development for COVID-19 alone. And 10 of those had names assigned by the WHO-INN in the last couple of months.Everyone reading this will be familiar with antibody structures. However, there are many antibody formats, which mostly are bio-specifics and, in some cases, multi specifics. (You can read a 2015 paper, (Spiess et al) that shows some of the formats in development at that time.) The focus of Professor Martin’s presentation was on the combining site and FV fragments, the variable domains that come together to form that binding site and how the development abYsis, which is freely available and accessible at www.abYsis.org.

The utility of abYsis lies in its ability to take random sequence data and create a visual representation of sequence information, including various annotations, easily. It is based on a complex processing pipeline, using public DNA & protein sequence data and 3d structural data. This data is numbered in a canonical loop assignment, looking for post-translational modifications and liabilities and feed into a relational database that can calculate various statistics about frequencies of residues etc. Tools also include germline analysis, humanness, and unusual residue analysis. Individual user sequence data can pass through the same process to generate the same view.

“abYsis allows us to explore an annotate antibody sequence and structure, including comparisons with observed residue distributions. It can aid with humanization by which I mean, making sequences more human and also with library design.”

Andrew Martin

The presentation explored:
• interactive sorting,
• calculation of residue frequencies
• how to increase the humanness of a sequence.

Watch the presentation here.

Looking to the future, Andrew described the library design tool, developed in collaboration with Twist, which allows users to generate libraries, which have variation at specific positions offering potential explanations for and solutions to immunogenicity. Developments include a high throughput system, which allows, for example, screening of large libraries of sequences from next-generation sequencing and enabling screening based on various developability constraints.

Andrew Martin, Professor, Bioinformatics and Computational Biology, UCL

Tags: antibody, biologics, drug discovery