Dataset Information

Structure-free antibody paratope similarity prediction for in silico epitope binning via protein language models.

ABSTRACT: Antibodies are an important group of biological molecules that are used as therapeutics and diagnostic tools. Although millions of antibody sequences are available, identifying their structural and functional similarity and their antigen binding sites remains a challenge at large scale. Here, we present a fast, sequence-based computational method for antibody paratope prediction based on protein language models. The paratope information is then used to measure similarity among antibodies via protein language models. Our computational method enables binning of antibody discovery hits into groups as the function of epitope engagement. We further demonstrate the utility of the method by identifying antibodies targeting highly similar epitopes of the same antigens from a large pool of antibody sequences, using two case studies: SARS CoV2 Receptor Binding Domain (RBD) and Epidermal Growth Factor Receptor (EGFR). Our approach highlights the potential in accelerating antibody discovery by enhancing hit prioritization and diversity selection.

SUBMITTER: Ghanbarpour A

PROVIDER: S-EPMC9941125 | biostudies-literature | 2023 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Structure-free antibody paratope similarity prediction for <i>in silico</i> epitope binning via protein language models.

Ghanbarpour Ahmadreza A Jiang Min M Foster Denisa D Chai Qing Q

iScience 20230125 2

Antibodies are an important group of biological molecules that are used as therapeutics and diagnostic tools. Although millions of antibody sequences are available, identifying their structural and functional similarity and their antigen binding sites remains a challenge at large scale. Here, we present a fast, sequence-based computational method for antibody paratope prediction based on protein language models. The paratope information is then used to measure similarity among antibodies via pro ...[more]

PMID: 36824280

Similar Datasets

Project description:Binding of platelet receptor glycoprotein Ibα (GPIbα) to the A1 domain of von Willebrand factor (vWF) is a critical step in both physiologic hemostasis and pathologic thrombosis, for initiating platelet adhesion to subendothelium of blood vessels at sites of vascular injury. Gain-of-function mutations in GPIbα contribute to an abnormally high-affinity binding of platelets to vWF and can lead to thrombosis, an accurate complication causing heart attack and stroke. Of various antithrombotic monoclonal antibodies (mAbs) targeting human GPIbα, 6B4 is a potent one to inhibit the interaction between GPIbα and vWF-A1 under static and flow conditions. Mapping paratope to epitope with mutagenesis experiments, a traditional route in researches of these antithrombotic mAbs, is usually expensive and time-consuming. Here, we suggested a novel computational procedure, which combines with homology modeling, rigid body docking, free and steered molecular dynamics (MD) simulations, to identify key paratope residues on 6B4 and their partners on GPIbα, with hypothesis that the stable hydrogen bonds and salt bridges are the important linkers between paratope and epitope residues. Based on a best constructed model of 6B4 bound with GPIbα, the survival ratios and rupture times of all detected hydrogen bonds and salt bridges in binding site were examined via free and steered MD simulations and regarded as indices of thermal and mechanical stabilizations of the bonds, respectively. Five principal paratope residues with their partners were predicted with their high survival ratios and/or long rupture times of involved hydrogen bonds, or with their hydrogen bond stabilization indices ranked in top 5. Exciting, the present results were in good agreement with previous mutagenesis experiment data, meaning a wide application prospect of our novel computational procedure on researches of molecular of basis of ligand-receptor interactions, various antithrombotic mAbs and other antibodies as well as theoretically design of biomolecular drugs.

Dataset Information

Structure-free antibody paratope similarity prediction for in silico epitope binning via protein language models.

Publications

Structure-free antibody paratope similarity prediction for <i>in silico</i> epitope binning via protein language models.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets