Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat
Sign Up
Cited by 16Advanced NMR Techniques and ApplicationsProtein Structure and DynamicsMolecular spectroscopy and chirality

The DOI matches its Crossref record; 57 of the 57 references it cites were found in a registry or an index.

Robust automated backbone triple resonance NMR assignments of proteins using Bayesian-based simulated annealing

Open access
Bishop, Anthony C., Torres-Montalvo, Glorisé, Kotaru, Sravya, Mimun, Kyle, Wand, A. Joshua·2023·Nature Communications
DOI 10.1038/s41467-023-37219-zPMC PMC10030768

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

Assignment of resonances of nuclear magnetic resonance (NMR) spectra to specific atoms within a protein remains a labor-intensive and challenging task. Automation of the assignment process often remains a bottleneck in the exploitation of solution NMR spectroscopy for the study of protein structure-dynamics-function relationships. We present an approach to the assignment of backbone triple resonance spectra of proteins. A Bayesian statistical analysis of predicted and observed chemical shifts is used in conjunction with inter-spin connectivities provided by triple resonance spectroscopy to calculate a pseudo-energy potential that drives a simulated annealing search for the most optimal set of resonance assignments. Termed Bayesian Assisted Assignments by Simulated Annealing (BARASA), a C++ program implementation is tested against systems ranging in size to over 450 amino acids including examples of intrinsically disordered proteins. BARASA is fast, robust, accommodates incomplete and incorrect information, and outperforms current algorithms – especially in cases of sparse data and is sufficiently fast to allow for real-time evaluation during data acquisition.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41467-023-37219-z

Cites

57 of 57 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Outline of the BARASA triple resonance assignment algorithm. a The search engine rests on a Bayesian-based simulated annealing protocol that uses a specific-heat mechanism to guide cooling. Crosspeaks lists drawn from triple resonance spectra are assembled into putative spin systems, which are then randomly assigned to positions within the primary sequence of the protein. Sequential adjacency in the primary sequence is provided by apparent connectivities derived from triple resonance NMR spectra. Predicted chemical shifts, based on a high-resolution structural model or gleaned from empirical amino acid-specific distributions, are incorporated into the system energy using Bayesian statistics. Throughout annealing, crosspeaks may move among spin systems with overlapping resonances, changing the energies of the affected spin systems. Annealing involves Monte Carlo swapping of both crosspeak assignments to spin systems and spin system assignments to locations in the sequence. The concept of dynamic swapping of individual crosspeaks or entire spin systems is outlined in Fig. 2. Annealing continues until energy equilibration is achieved. The temperature is then lowered and the system re-equilibrated. Annealing is stopped when the termination criteria are met and a local minimization routine is performed. b The final resonance assignments are developed from results of multiple independent simulated annealing runs. c Shown is a ribbon representation of maltose binding protein (PDB code: 1DMB [10.2210/pdb1DMB/pdb]) color-coded according to assignment status following analysis by BARASA: correctly assigned residues (blue); unassigned residues (white), prolines (red). See main text for further details.
Fig. 1 Outline of the BARASA triple resonance assignment algorithm. a The search engine rests on a Bayesian-based simulated annealing protocol that uses a specific-heat mechanism to guide cooling. Crosspeaks lists drawn from triple resonance spectra are assembled into putative spin systems, which are then randomly assigned to positions within the primary sequence of the protein. Sequential adjacency in the primary sequence is provided by apparent connectivities derived from triple resonance NMR spectra. Predicted chemical shifts, based on a high-resolution structural model or gleaned from empirical amino acid-specific distributions, are incorporated into the system energy using Bayesian statistics. Throughout annealing, crosspeaks may move among spin systems with overlapping resonances, changing the energies of the affected spin systems. Annealing involves Monte Carlo swapping of both crosspeak assignments to spin systems and spin system assignments to locations in the sequence. The concept of dynamic swapping of individual crosspeaks or entire spin systems is outlined in Fig. 2. Annealing continues until energy equilibration is achieved. The temperature is then lowered and the system re-equilibrated. Annealing is stopped when the termination criteria are met and a local minimization routine is performed. b The final resonance assignments are developed from results of multiple independent simulated annealing runs. c Shown is a ribbon representation of maltose binding protein (PDB code: 1DMB [10.2210/pdb1DMB/pdb]) color-coded according to assignment status following analysis by BARASA: correctly assigned residues (blue); unassigned residues (white), prolines (red). See main text for further details.CC BY · PubMed Central
Fig. 2 Dynamic assembly of spin systems during simulated annealing. a Spin systems (orange puzzle pieces) begin in the cache (black box) and are initialized by random assignment to the sequence (purple pieces). Spin systems can then be swapped with others or moved to different locations of the sequence or to the cache. Spin systems or cross peaks in their respective caches have no priority and are randomly selected. Swaps are accepted or rejected with a probability based on the change in energy of the proposed swap. b The energy of each spin system depends on how it fits with the adjacent spin system (adjacency energy) and with the predicted shifts for that residue location (chemical shift energy). Exchange of crosspeaks between spin systems can be thought of as changing the puzzle piece shape. See main text and supplementary material for details.
Fig. 2 Dynamic assembly of spin systems during simulated annealing. a Spin systems (orange puzzle pieces) begin in the cache (black box) and are initialized by random assignment to the sequence (purple pieces). Spin systems can then be swapped with others or moved to different locations of the sequence or to the cache. Spin systems or cross peaks in their respective caches have no priority and are randomly selected. Swaps are accepted or rejected with a probability based on the change in energy of the proposed swap. b The energy of each spin system depends on how it fits with the adjacent spin system (adjacency energy) and with the predicted shifts for that residue location (chemical shift energy). Exchange of crosspeaks between spin systems can be thought of as changing the puzzle piece shape. See main text and supplementary material for details.CC BY · PubMed Central
Fig. 3 Performance of BARASA using a comprehensive set of triple resonance experiments. Comparison of automated assignment algorithms. Results of automated resonance assignments by BARASA utilizing raw crosspeak lists drawn from a relatively comprehensive set of triple resonance experiments. Compared to manually curated resonance assignments obtained for eight test proteins: interleukin-1β (IL-1β), interleukin-1 receptor antagonist (C66A, C122A) (IL-1Ra); indole-3-glycerol phosphate synthase (R43S) (IGPS), maltose binding protein (MBP), non-ribosomal peptide synthetase (Cy1), thymidylate synthase (ecTS), V5 domain of protein kinase C (V5dm), and intrinsically disordered region of human ANP32A (hIDD). Shown are the fractions of residues that are accurately matched (green), mismatched (magenta), or missing (i.e., unassigned) (blue) to the reference assignments. *In the case of hIDD, a number of de novo assignments were indicated by BARASA and are included as mismatching with the reference assignments. See main text and Table 1. Source data are provided in the Source Data file.
Fig. 3 Performance of BARASA using a comprehensive set of triple resonance experiments. Comparison of automated assignment algorithms. Results of automated resonance assignments by BARASA utilizing raw crosspeak lists drawn from a relatively comprehensive set of triple resonance experiments. Compared to manually curated resonance assignments obtained for eight test proteins: interleukin-1β (IL-1β), interleukin-1 receptor antagonist (C66A, C122A) (IL-1Ra); indole-3-glycerol phosphate synthase (R43S) (IGPS), maltose binding protein (MBP), non-ribosomal peptide synthetase (Cy1), thymidylate synthase (ecTS), V5 domain of protein kinase C (V5dm), and intrinsically disordered region of human ANP32A (hIDD). Shown are the fractions of residues that are accurately matched (green), mismatched (magenta), or missing (i.e., unassigned) (blue) to the reference assignments. *In the case of hIDD, a number of de novo assignments were indicated by BARASA and are included as mismatching with the reference assignments. See main text and Table 1. Source data are provided in the Source Data file.CC BY · PubMed Central
Fig. 4 Performance of BARASA under data sparse conditions. Shown are the fractions of residues that are accurately matched (green), mismatched (magenta) or missing (i.e., unassigned) (blue) to the reference assignments. Panels a–d correspond to results from input data sets where entire spin systems were discarded from the crosspeak lists. The ordinate axis is the fraction of retained spin systems and the dashed lines indicate the maximum fraction of possible matching assignments. The effects of random spin system depletion on the analysis of MBP both randomly (a) and as stretches of five consecutive residues being discarded (b). A similar analysis of ecTS with either individual (c) or groups of five consecutive spin systems being discarded (d). For the conditions 0.8 and 0.6 fractions retained, ten random data sets retaining the indicated fraction of spin systems were generated. The performance of BARASA on each data set is shown as a single orange solid circle, with the bar height representing the arithmetic mean. The full data set (“1.0” condition) results were taken from Fig. 3. Only one result with the full data set was measured to avoid the comparison of run-to-run variation with variation due to differences in the input data set. The effects of restricting connectivity information by utilizing only a single pair of triple resonance experiments with either residue-type statistics (BMRB) (e) or structure-based (SHIFTX + ) (f) chemical shift predictions for MBP. Similarly, for ecTS using only residue-type statistics (BMRB) (g) or structure-based (SHIFTX + ) (h) chemical shift predictions. The effect of random depletion of crosspeaks from the comprehensive set of triple resonance experiments where the indicated percentages each type of crosspeak that are retained is illustrated for the MBP (i) and ecTS (j) data sets and used with residue-type statistics (BMRB) or structure-based (SHIFTX+) predicted chemical shifts. Results of ten individual runs (n = 10) are plotted as solid orange circles and bar heights represent the arithmetic mean. Source data are provided as a Source Data file.
Fig. 4 Performance of BARASA under data sparse conditions. Shown are the fractions of residues that are accurately matched (green), mismatched (magenta) or missing (i.e., unassigned) (blue) to the reference assignments. Panels a–d correspond to results from input data sets where entire spin systems were discarded from the crosspeak lists. The ordinate axis is the fraction of retained spin systems and the dashed lines indicate the maximum fraction of possible matching assignments. The effects of random spin system depletion on the analysis of MBP both randomly (a) and as stretches of five consecutive residues being discarded (b). A similar analysis of ecTS with either individual (c) or groups of five consecutive spin systems being discarded (d). For the conditions 0.8 and 0.6 fractions retained, ten random data sets retaining the indicated fraction of spin systems were generated. The performance of BARASA on each data set is shown as a single orange solid circle, with the bar height representing the arithmetic mean. The full data set (“1.0” condition) results were taken from Fig. 3. Only one result with the full data set was measured to avoid the comparison of run-to-run variation with variation due to differences in the input data set. The effects of restricting connectivity information by utilizing only a single pair of triple resonance experiments with either residue-type statistics (BMRB) (e) or structure-based (SHIFTX + ) (f) chemical shift predictions for MBP. Similarly, for ecTS using only residue-type statistics (BMRB) (g) or structure-based (SHIFTX + ) (h) chemical shift predictions. The effect of random depletion of crosspeaks from the comprehensive set of triple resonance experiments where the indicated percentages each type of crosspeak that are retained is illustrated for the MBP (i) and ecTS (j) data sets and used with residue-type statistics (BMRB) or structure-based (SHIFTX+) predicted chemical shifts. Results of ten individual runs (n = 10) are plotted as solid orange circles and bar heights represent the arithmetic mean. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 5 Comparison of BARASA against other assignment algorithms. Performance of BARASA, FLYA, AutoAssign (AA), and I-PINE against reference triple resonance assignments of six protein systems: a IL-1β; b IL-1Ra; c IGPS; d MBP; e CY1; f ecTS; g V5dm; h hIDD. Shown are the fractions of residues that are accurately matched to the reference assignments (green), incorrectly matched (magenta) or missing (i.e., unassigned) (blue). *BARASA and, to a lesser extent, FLYA extended the reference assignments for hIDD considerably (Supplementary Table 10). The extended assignments are therefore denoted here as mismatching. Source data are provided as a Source Data file.
Fig. 5 Comparison of BARASA against other assignment algorithms. Performance of BARASA, FLYA, AutoAssign (AA), and I-PINE against reference triple resonance assignments of six protein systems: a IL-1β; b IL-1Ra; c IGPS; d MBP; e CY1; f ecTS; g V5dm; h hIDD. Shown are the fractions of residues that are accurately matched to the reference assignments (green), incorrectly matched (magenta) or missing (i.e., unassigned) (blue). *BARASA and, to a lesser extent, FLYA extended the reference assignments for hIDD considerably (Supplementary Table 10). The extended assignments are therefore denoted here as mismatching. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 6 Performance of BARASA and FLYA with depleted crosspeak lists. The effects of random crosspeak depletion on the analysis of MBP (a) and ecTS (b) comprehensive triple resonance data sets with partial retention of the indicated crosspeak types (see text and Fig. 4). Shown are the fractions of residues that are accurately matched (green), mismatched (magenta) or missing (i.e., unassigned) (blue) to the reference assignments. Ten independent data sets (n = 10) were randomly generated for each depletion condition. The results of analysis by BARASA for each data set are shown as solid orange circles and the bar heights correspond to the mean. Source data are provided as a Source Data file.
Fig. 6 Performance of BARASA and FLYA with depleted crosspeak lists. The effects of random crosspeak depletion on the analysis of MBP (a) and ecTS (b) comprehensive triple resonance data sets with partial retention of the indicated crosspeak types (see text and Fig. 4). Shown are the fractions of residues that are accurately matched (green), mismatched (magenta) or missing (i.e., unassigned) (blue) to the reference assignments. Ten independent data sets (n = 10) were randomly generated for each depletion condition. The results of analysis by BARASA for each data set are shown as solid orange circles and the bar heights correspond to the mean. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 7 Sensitivity of automated resonance assignment of Cy1 to breadth of triple resonance experimental foundation and chemical shift prediction precision. The resonance assignment by BARASA using the indicated cross crosspeak types from the triple resonance spectra and, residue-type (BMRB) chemical shift statistics (a) or chemical shifts predicted by SHIFTX + based on a structural model provided by AlphaFold2 (b). Triple resonance data sets include the peaks from the following spectra: HNCA/HN(CO)CA (Cα), HN(CA)CB/HN(COCA)CB (Cβ) and HNCO/HN(CA)CO (CO). Bar heights indicate the fractions of residues that are accurately matched (green), mismatched (magenta) or missing (i.e., unassigned) (blue) to the reference assignments. Equivalent runs with FLYA (c) using the data set of (b) reinforce the conclusion that BARASA is more robust to non-ideal data. Source data are provided as a Source Data file.
Fig. 7 Sensitivity of automated resonance assignment of Cy1 to breadth of triple resonance experimental foundation and chemical shift prediction precision. The resonance assignment by BARASA using the indicated cross crosspeak types from the triple resonance spectra and, residue-type (BMRB) chemical shift statistics (a) or chemical shifts predicted by SHIFTX + based on a structural model provided by AlphaFold2 (b). Triple resonance data sets include the peaks from the following spectra: HNCA/HN(CO)CA (Cα), HN(CA)CB/HN(COCA)CB (Cβ) and HNCO/HN(CA)CO (CO). Bar heights indicate the fractions of residues that are accurately matched (green), mismatched (magenta) or missing (i.e., unassigned) (blue) to the reference assignments. Equivalent runs with FLYA (c) using the data set of (b) reinforce the conclusion that BARASA is more robust to non-ideal data. Source data are provided as a Source Data file.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Research works printed but not linked: 2

  1. Montelione, G. T. & Wagner, G. Conformation-independent sequential NMR connections in polypeptides by H1-C13-N15 triple-resonance experiments. J. Magn. Reson. 87, 183–188 (1990).
  2. Grzesiek, S. & Bax, A. Improved 3D triple-resonance NMR techniques applied to a 31 kDa protein. J. Magn. Reson. 96, 432–440 (1992).

Metadata

Authors
Bishop, A. C., Torres-Montalvo, G., Kotaru, S., Mimun, K., & Wand, A. J.
Affiliations
  • Texas A&M University (US)
  • University of Pennsylvania (US)
Corresponding author
Wand, A. Joshua
ORCID
Venue
Nature Communications
Publisher
Springer Science and Business Media LLC
Published
2023-03-21
Volume
14
Issue
1
Pages
1556-1556
ISSN
2041-1723 (ISSN-L)
DOI
10.1038/s41467-023-37219-z
PMC
PMC10030768
OpenAlex
W4328126951
Cited by
16
References
57
Citations by year
2026: 4 · 2025: 4 · 2024: 7 · 2023: 1
FWCI
1.48
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • Research Square (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Cancer Prevention and Research Institute of Texas (RP210043)
  • G. Harold and Leila Y. Mathers Foundation (MF-1809-00155)
  • Texas A and M University (n/a)
  • U.S. Department of Health & Human Services | National Institutes of Health (R01 GM129076)
Topics
  • Physical Sciences › Chemistry › Spectroscopy › Advanced NMR Techniques and Applications
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
  • Physical Sciences › Chemistry › Spectroscopy › Molecular spectroscopy and chirality
Keywords
chemical shift prediction, intrinsically disordered proteins, protein structure determination, data sparsity
MeSH terms
Algorithms, Amino Acids, Bayes Theorem, Magnetic Resonance Spectroscopy, Proteins, Nuclear Magnetic Resonance, Biomolecular

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.