henhaobio
Sign in

bioRxivBioinformaticsPreprint

HANSEN: An Integrated Structural and Functional Proteome Resource for Structure-Guided Drug Discovery in Mycobacterium leprae

Sundeep Chaitanya VedithiiD, Rebecca Rees, Sony MalhotraiD, Asma MuniriD, Modestas Matusevicius, Ali F. AlsulamiiD, Christopher A. BeaudoiniD, Kiran Sai Sunkara, Madhusmita DasiD, Tom L. BlundelliD, Rodrigo Andres Floto

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

Abstract

Leprosy remains a leading infectious cause of preventable disability, yet its causative agent, Mycobacterium leprae (M. leprae), is structurally under-characterised. Only ten Protein Data Bank (PDB) entries represent seven of its 1,603 protein-coding genes. We present HANSEN, a proteome-wide structural and functional resource for M. leprae. Monomeric and oligomeric models were generated with AlphaFold 3, Boltz-1, Boltz-2 and Chai-1, and annotated with per-residue confidence, predicted aligned error and, for assemblies, interface confidence. Ligand-binding pockets were predicted with AF2BIND, P2Rank and fpocket, template-derived ligands were modelled within oligomeric complexes, residue-level B-cell epitope propensity was estimated with DiscoTope-3.0, and gene essentiality was transferred from Mycobacterium tuberculosis transposon-sequencing labels. These features are integrated in a relational database with interactive visualisation and combined into a calibrated Target Priority Score that ranks all 1,603 proteins into four tiers and recovers established antimycobacterial targets. HANSEN (https://hansen-leprosy.medschl.cam.ac.uk/home) provides a practical basis for target prioritisation and structure-guided drug discovery in leprosy.

In-depth summary

Claude Sonnet 4.6 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (29 Aug 2026) · CC BY 4.0

Web resource offering predicted monomeric and oligomeric structures for 1,603 M. leprae proteins, with pockets, epitope scores, essentiality and a target priority ranking, plus a nanopore resistance-calling module.

M. leprae cannot be grown in axenic culture, and only seven unique proteins have experimental structures. The authors describe HANSEN as the first M. leprae-specific platform to combine proteome-wide models, including oligomers that the monomer-only AlphaFold database models lack, with functional features.

Key results

  1. ShownHANSEN covers 1,603 records with predicted monomers from four methods and 726 oligomeric records for 720 proteins, plus pocket, epitope and prioritisation layers.

    Oligomeric records: 528 homo and 198 hetero (704 from Boltz-2, 22 from AlphaFold 3). Consensus pockets: 47,194 (aggregated over models, not distinct sites). DiscoTope-3.0 flags 216,748 of 1,078,217 residues (20.1%). Only seven proteins have experimental structures. · 2.1 A proteome-wide structural atlas for M…; 2.3 Oligomeric Assembly Representation; 2.4 Binding Site Predictions; 2.6 B-cell Epitope Propensity

  2. SuggestedAgainst the seven M. leprae proteins with experimental structures, AlphaFold 3 and Chai-1 matched best. The RuvA case lowered the Boltz-1 and Boltz-2 means, and pLDDT tracked local accuracy.

    Mean TM-score 0.90 (AlphaFold 3), 0.89 (Chai-1), 0.85 (Boltz-2), 0.82 (Boltz-1); median Cα-RMSD 1.9 Å. RuvA (ML0482): Boltz-1 0.21, Boltz-2 0.34. Excluding RuvA the Boltz-1 and Boltz-2 means rise to 0.92 and 0.93. pLDDT vs Cα-lDDT Pearson r 0.64 over 4,624 residues. · 2.9 Benchmarking Model Accuracy and Confidence…

  3. ShownThe methods largely agree with each other and with AFDB models. Residual disagreement sits mostly in flexible regions and inter-domain arrangement, not in domain folds.

    Boltz-1 vs Boltz-2 mean TM-score 0.83, pLDDT r = 0.96. Median best-pair TM-score 0.94; 93% of proteins reach 0.5 (115 do not). Versus AFDB (1,598 proteins, 6,379 pairs), 86 to 90% reach TM-score 0.5, median 0.934 for AlphaFold 3. Restricting to confident residues, over 98% reach 0.5. · 2.10 Cross-method concordance; 2.10.1 Concordance with AFDB

  4. SuggestedThe Target Priority Score places known drug targets near the top of the ranking of 1,603 proteins, supporting its use as a triage tool.

    Of 24 validated reference targets, 6 are in the top 50 (chance 3%), 11 in the top 200 (chance 12%) and 23 in the top 400 (chance 25%). Tiers: 50 high-priority, 123 strong, 498 moderate, 932 exploratory. gyrB ranks 5, qcrB 23, rpoB 24. · 2.7 Essentiality and Target Prioritisation; 2.11 Recovery of Validated Drug Targets by the…

  5. SuggestedProteomeLM-Ess, trained on M. tuberculosis essentiality labels, transfers moderately to M. leprae. It assigns 90 proteins as essential and 79 as likely essential.

    Homology-grouped five-fold CV AUROC 0.84 in M. tuberculosis; transfer AUROC 0.78 on 241 M. leprae proteins with M. tuberculosis orthologues, only 39 labelled essential. Classes: 90 essential, 79 likely-essential, 101 uncertain, 1,333 non-essential. · 2.7.1 Cross-species essentiality prediction with…; 5.9 ProteomeLM-derived Contextual Essentiality

Methods, data and reproducibility

Proteome-scale structure prediction with AlphaFold 3, Boltz-1, Boltz-2 and Chai-1, with template-guided oligomer modelling, was combined with pocket, epitope, essentiality and weighted priority scoring in a Flask web app. A Nanopore amplicon module (Lep-AMR) reports resistance variants in rpoB, folP1 and gyrA.

  • M. leprae proteome records (UniProt-derived) modelled by four methods · Mycobacterium leprae · n = 1,603 records; 1,599 unique gene names
  • Experimental M. leprae structures used as accuracy references · Mycobacterium leprae · n = 7 proteins · 7OA5 (RuvA); 5IE8 (ML1382)
  • AFDB M. leprae TN proteome release (model version v6) used for concordance · Mycobacterium leprae · n = 1,598 proteins, 6,379 model pairs · UP000000806_272631_MYCLE
  • M. tuberculosis H37Rv proteome with saturating-transposon essentiality labels, used to train the essentiality head · Mycobacterium tuberculosis
  • Reference set of validated drug targets for benchmarking the priority score · Mycobacterium leprae · n = 24 targets
  • Single-barcode folP1 Nanopore amplicon dataset used to demonstrate Lep-AMR · Mycobacterium leprae · n = 2,306 reads
Methods & tools
AlphaFold 3, Boltz-1, Boltz-2, Chai-1, ABCFold, fpocket, P2Rank, AF2Bind, DiscoTope-3.0, ProteomeLM-Ess, EPI2ME wf-amplicon, Medaka
Software & environment
Lep-AMR uses EPI2ME wf-amplicon 1.2.2, minimap2 2.28 and Medaka 2.2.0 (model dna_r10.4.1_e8.2_400bps_hac@v4.2.0). No environment file is stated for the rest.
Reproducibility gaps
GitHub is cited only for the Supplementary Note SN1 AFDB-comparison code; release of modelling, scoring and web-app code is not stated. Seq3D is a manuscript in preparation; the folP1 demonstration data are unpublished.
Limitations (5)
  • The resource is hypothesis-generating. Predicted pockets, pLDDT and PAE reflect model self-assessment, not biological validity, and priority scores need experimental validation. · stated by the authors
  • Essentiality is transferred cross-species from M. tuberculosis. Labels for M. leprae come from orthology, only 39 of 241 aligned proteins are essential, and cross-validation is the more conservative performance estimate. · stated by the authors
  • The AFDB comparison measures reproducibility, not accuracy. Pocket totals are aggregated over models and are not distinct binding sites. AlphaFold 3 oligomers lack ligand coordinates. · stated by the authors
  • Experimental accuracy benchmarking rests on only seven proteins, and the target-recovery test uses a 24-target reference set assembled by the authors. · noted in this summary
  • Lep-AMR is shown on a single unpublished folP1 sample with no sensitivity or specificity evaluation across genes or samples. · noted in this summary

Read this if you work on leprosy or mycobacterial target discovery and want open predicted structures, pockets and ranked candidates, or need to judge how far multi-method agreement can stand in for missing experimental structures.

Citations
0

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
THISHe '24Bannerman '19Andrade '23Todor '26Bannerman '23Hossain '26
59% semantic matchBioinformatics
Zhijie He, Cong Wang, Xudong Guo+7 more·Bioinformatics Advances·Jan 1, 2024· 1 citations