henhaobio
Sign in

Protein ScienceBiophysicsPeer-reviewed journal article

MAGPIE: an interactive tool for visualizing and analyzing protein-ligand interactions

Daniel RodrígueziD, Kyle C. WeberiD, Belen Sundberg, Anum Azam GlasgowiD

PDFDOIPreprintCodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Quantitative tools to compile and analyze biomolecular interactions among chemically diverse binding partners would improve therapeutics design and aid in the study of molecular evolution. Here we present MAGPIE (Mapping Areas of Genetic Parsimony In Epitopes), a publicly available software package for simultaneously visualizing and analyzing thousands of interactions between a single protein or small molecule ligand (the "target") and all of its protein binding partners ("binders"). MAGPIE generates an interactive 3D visualization from a set of protein complex structures that share the target ligand, as well as sequence logo-style amino acid frequency graphs that show all the amino acids from the set of protein binders that interact with user-defined target ligand positions or chemical groups. MAGPIE highlights all the salt bridge and hydrogen bond interactions made by the target in the visualization and as separate amino acid frequency graphs. Finally, MAGPIE collates the most common target-binder interactions as a list of "hotspots," which can be used to analyze trends or guide the de novo design of protein binders. As an example of the utility of the program, we used MAGPIE to probe how two ligands bind orthologs of a well-conserved glycolytic enzyme for a detailed understanding of evolutionarily conserved interactions involved in its activation and inhibition. MAGPIE is implemented in Python 3 and freely available at https://github.com/glasgowlab/MAGPIE, along with sample datasets, usage examples, and helper scripts to prepare input structures.

In-depth summary

Claude Sonnet 4.6 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (4 Mar 2024) · CC BY 4.0

MAGPIE aligns many protein complexes on a shared ligand and shows binder residues in interactive 3D, with amino acid frequency graphs and DBSCAN-derived interaction hotspots.

Sequence logos built from alignments need alignable binders and cannot cover small-molecule ligands, while structural search and small-molecule alignment tools do not organize conserved binding interactions across diverse proteins. MAGPIE pools many complexes around one target ligand and reports residue frequencies, hydrogen bonds, salt bridges and hotspots in 3D.

Key results

  1. ShownIn a set of SARS-CoV-2 RBD binders, MAGPIE separated two epitope loops: loop 1 contacts are almost all serines and glycines, while loop 2 contacts are chemically more varied.

    27 of 63 binders contact at least one of the two loops via CDRs. Serines are 60% of hydrogen bond partners in loop 1 (residues 127-131); loop 2 is residues 146-149. Default hotspot settings (2 Å, minimum 15 AAs) gave five hotspots. · 2.2. Case study 1: antibody-antigen complexes in…

  2. ShownSplitting coenzyme A complexes into conformer pools showed that the same ligand atoms have different local environments and hotspot counts in different ligand conformations.

    199 complexes chosen at random from >600 PDB hits were sorted into 31 conformer pools (2.5 Å RMSD). Pool 1 (22 structures) gave six hotspots (2 Å, minimum 20 AAs); pool 2 (21 models) gave two hotspots dominated by hydrophilic and charged AAs. · 2.3. Case study 2: protein complexes with a…; Data collection and curation; Figure legends

  3. SuggestedIn predicted bacterial PFK-1 ortholog models, PEP and ADP show distinct conformer and hotspot patterns, and a hydrophobic ADP hotspot overlaps the hotspot unique to PEP Conformer 1.

    72 PEP-bound models (7 phyla, 39 orders) gave four conformers (0.4 Å bins), two hotspots each (1.5 Å, minimum 15 AAs). 65 ADP-bound models (4 phyla, 15 orders) gave four conformers (0.7 Å), 58 in the first, with 6 hotspots (minimum 20 AAs). · 2.4. Case study 3: trends in inhibitor binding…; Figure legends

  4. SuggestedThe PFK-1 input models were built with AlphaFold2, Rosetta relaxation and ligand placement by structural alignment, then filtered. The final set spans considerable sequence diversity.

    Alignment to the PEP-bound G. stearothermophilus structure (PDB 4I4I) gave pocket all-atom RMSD 0.371 Å and global RMSD 0.493 Å. Mean sequence identity was 50.2% within a phylum vs 29.54% between phyla. · 2.4. Case study 3: trends in inhibitor binding…

  5. SuggestedComparing all-contact and hydrogen-bond frequency graphs separated polar from nonpolar contacts at the same RBD position.

    About 15% of interactions with F127 involve proline, tyrosine and alanine, yet none of these AAs contribute to F127 hydrogen bonds in the dataset. F127 pi stacks with them in individual structures. · 2.2. Case study 1: antibody-antigen complexes in…

Methods, data and reproducibility

Structures sharing a target ligand are cleaned, protonated and aligned with PyMOL-based helper scripts. A K-DTree finds neighboring binder residues, Plotly renders 3D views, LogoMaker draws frequency graphs, and DBSCAN clusters hotspots.

  • Nonredundant antibody Fab and nanobody complexes with the SARS-CoV-2 spike RBD (from a previously compiled RBD-antibody dataset) · SARS-CoV-2 (antigen); binder species not stated · n = 63 nonredundant binders (from a 68-member subset)
  • Coenzyme A-bound bacterial enzyme structures from the PDB (1.5-3 Å resolution), randomly chosen from >600 hits · bacteria (various) · n = 199 structures; 31 conformer pools · PubChem identifier 87642
  • AlphaFold2 models of PEP-bound bacterial PFK-1 orthologs, ligand placed by alignment to a reference complex · bacteria (7 phyla, 39 orders) · n = 72 models · PDB 4I4I (reference)
  • AlphaFold2 models of ADP-bound bacterial PFK-1 orthologs, aligned to the E. coli reference · bacteria (4 phyla, 15 orders) · n = 65 models · PDB 1PFK (reference)
Methods & tools
MAGPIE, PyMOL, K-DTree, Plotly, LogoMaker, DBSCAN (scikit-learn), AlphaFold2, Rosetta relax, US-align, UniProt, PDB advanced search, Google Colab
Software & environment
Python 3 in Jupyter notebook format; Google Colab notebook available; local installation requirements are listed in supplementary section 9 (not in the main text).
Assay conditions & compute
Uses multiple CPU processors; Colab suitable for fewer than about 1000 structures; larger or private datasets should use a local or multithreaded run. No runtimes reported.
Reproducibility gaps
Supplementary sections 2, 6, 8 and 9 hold key methods and are not in the text; package versions are not pinned; availability of the PFK-1 model coordinates is not stated.
Limitations (4)
  • Colab runs are suited only to datasets under about 1000 structures; private or larger datasets need a local or multithreaded installation. · stated by the authors
  • Few bacterial PFK-1 structures have been experimentally solved, so case study 3 relies on AlphaFold2 models with PEP and ADP placed by structural alignment to reference complexes. · stated by the authors
  • PFK-1 findings rest on AlphaFold2 and Rosetta models with ligands placed by alignment to reference crystal structures (a mutant for PEP); interpretations such as conformer enrichment are untested hypotheses. · noted in this summary
  • Hotspot thresholds varied between case studies (2 Å or 1.5 Å; minimum 15 or 20 AAs), and no sensitivity analysis, background comparison or statistical test of enrichment is reported. · noted in this summary

Read this if you need to compare residue-level interactions across many structures or models sharing a ligand, for binder design or evolutionary analysis.

Citations
4
FWCI
0.54
Field percentile
top 42%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISPanda '26Young '21Kynast '22Ribeiro '25Sharma '26Heyne '20
66% semantic matchBioinformatics
Pritam Kumar Panda·bioRxiv·Aug 9, 2026