CoChat
Sign Up
Cited by 8Metabolomics and Mass Spectrometry StudiesAdvanced NMR Techniques and ApplicationsSpectroscopy and Chemometric Analyses

The DOI matches its Crossref record; 30 of the 30 references it cites were found in a registry or an index.

RTExtract: time-series NMR spectra quantification based on 3D surface ridge tracking

Open access
Wu, Yue, Judge, Michael T, Arnold, Jonathan, Bhandarkar, Suchendra M, Edison, Arthur S·2020·Bioinformatics
DOI 10.1093/bioinformatics/btaa631PMC PMC7755419
Open this paper in CoChat — free accountOpen PDF

Abstract

Motivation: Time-series nuclear magnetic resonance (NMR) has advanced our knowledge about metabolic dynamics. Before analyzing compounds through modeling or statistical methods, chemical features need to be tracked and quantified. However, because of peak overlap and peak shifting, the available protocols are time consuming at best or even impossible for some regions in NMR spectra. Results: We introduce Ridge Tracking-based Extract (RTExtract), a computer vision-based algorithm, to quantify time-series NMR spectra. The NMR spectra of multiple time points were formulated as a 3D surface. Candidate points were first filtered using local curvature and optima, then connected into ridges by a greedy algorithm. Interactive steps were implemented to refine results. Among 173 simulated ridges, 115 can be tracked (RMSD < 0.001). For reproducing previous results, RTExtract took less than 2 h instead of ∼48 h, and two instead of seven parameters need tuning. Multiple regions with overlapping and changing chemical shifts are accurately tracked. Availability and implementation: Source code is freely available within Metabolomics toolbox GitHub repository (https://github.com/artedison/Edison_Lab_Shared_Metabolomics_UGA/tree/master/metabolomics_toolbox/code/ridge_tracking) and is implemented in MATLAB and R. Supplementary information: Supplementary data are available at Bioinformatics online.

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 · 10.1093/bioinformatics/btaa631

Cites

30 of 30 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1. Illustration of the concept of HandK curvatures. N is the normal vector, and the curve is one of the intersecting curves. The two principal curvatures, κ1 and κ2, correspond to the two vectors. Gaussian curvature (K) and mean curvature (H) are calculated by κ1 and κ2. For the ridge surface shown here, κ2<0 and κ1≈0 which results in H<0 and K≈0. Computation of H and K curvature is in Section 2.1. Other surface types are illustrated in Supplementary Figure S1
Fig. 1. Illustration of the concept of HandK curvatures. N is the normal vector, and the curve is one of the intersecting curves. The two principal curvatures, κ1 and κ2, correspond to the two vectors. Gaussian curvature (K) and mean curvature (H) are calculated by κ1 and κ2. For the ridge surface shown here, κ2<0 and κ1≈0 which results in H<0 and K≈0. Computation of H and K curvature is in Section 2.1. Other surface types are illustrated in Supplementary Figure S1CC BY-NC · PubMed Central
Fig. 2. Illustration of the RTExtract algorithm. The algorithm is presented based on an example time-series NMR dataset, which was measured under aerobic conditions (A) (Judge et al., 2019). The stack spectral plot (A) shows changes of whole NMR spectra (X-axis chemical shift δ) through time (Y-axis), and each time point is distinguished by a different line color as in all stack spectral plot in this article. Two regions (ppm [5.5, 8.5] and [1.3, 2.2]) are expanded to show spectral complexity. (B) A column slice (ppm [1.30, 1.35], indicated as orange box and arrow in (A) is chosen as show case for RTExtract (B–F). H and K curvatures were computed based on the 7 by 7 window (B) around each point, which was used to define ridge points (H < 0 and |K|<1). (C) To form candidate points (red points), these ridge points are intersected with local maxima and combined with the first Nmax global maximal points for each spectrum. (D) These points were connected in two steps by a greedy method to form ridges (red line). The first step was based on chemical shift distance (Lgap) between points, and the second step was based on distance and angles. (E) Ridges were refined and manually selected for feature quantification. For the red ridge in (E), intensity and chemical shift (δ) were plotted against time (F). The intensity is measured in arbitrary units (AUs), time is measured in hours, and chemical shift (δ) is measured in ppm. Details on the RTExtract algorithm and tuning parameters can be found in Sections 2.1 and 2.2
Fig. 2. Illustration of the RTExtract algorithm. The algorithm is presented based on an example time-series NMR dataset, which was measured under aerobic conditions (A) (Judge et al., 2019). The stack spectral plot (A) shows changes of whole NMR spectra (X-axis chemical shift δ) through time (Y-axis), and each time point is distinguished by a different line color as in all stack spectral plot in this article. Two regions (ppm [5.5, 8.5] and [1.3, 2.2]) are expanded to show spectral complexity. (B) A column slice (ppm [1.30, 1.35], indicated as orange box and arrow in (A) is chosen as show case for RTExtract (B–F). H and K curvatures were computed based on the 7 by 7 window (B) around each point, which was used to define ridge points (H < 0 and |K|<1). (C) To form candidate points (red points), these ridge points are intersected with local maxima and combined with the first Nmax global maximal points for each spectrum. (D) These points were connected in two steps by a greedy method to form ridges (red line). The first step was based on chemical shift distance (Lgap) between points, and the second step was based on distance and angles. (E) Ridges were refined and manually selected for feature quantification. For the red ridge in (E), intensity and chemical shift (δ) were plotted against time (F). The intensity is measured in arbitrary units (AUs), time is measured in hours, and chemical shift (δ) is measured in ppm. Details on the RTExtract algorithm and tuning parameters can be found in Sections 2.1 and 2.2CC BY-NC · PubMed Central
Fig. 3. Evaluation of the RTExtract algorithm on simulated datasets. (A) Compound quantification for all ridge points was plotted. For each ridge peak, extracted compound concentrations normalized to DSS (Y-axis, [C]ex) were plotted against simulated compound concentration (X-axis, [C]sim). The black diagonal represents perfect quantification and its slope is 1. Four overlapped peaks mapped to four compounds [1: alanine (blue); 2: glycerol (pink); 3: serine (green); 4: ethanol (purple)] were selected as examples (A–C). All other ridge overlapped peaks are in gray. (B) In the stack spectra, one spectrum was highlight in black, which was then decomposed into compound peaks involved in the overlap (C). In (C), the black line is the full simulated spectrum with all the compound peaks. (D) The acetate peak was simulated with chemical shift variation under different pH conditions and overlapped with another peak. (E) The extracted concentrations ([C]ex) were plotted against simulated concentrations ([C]sim). (F) Extracted chemical shifts (δex) were plotted against simulated chemical shift (δsim). Performance of concentration and chemical shift estimation with more compounds can be found in Supplementary Figures S5 and S6. Compound concentration was simulated with arbitrary unit (AU) and chemical shift δ was evaluated in ppm
Fig. 3. Evaluation of the RTExtract algorithm on simulated datasets. (A) Compound quantification for all ridge points was plotted. For each ridge peak, extracted compound concentrations normalized to DSS (Y-axis, [C]ex) were plotted against simulated compound concentration (X-axis, [C]sim). The black diagonal represents perfect quantification and its slope is 1. Four overlapped peaks mapped to four compounds [1: alanine (blue); 2: glycerol (pink); 3: serine (green); 4: ethanol (purple)] were selected as examples (A–C). All other ridge overlapped peaks are in gray. (B) In the stack spectra, one spectrum was highlight in black, which was then decomposed into compound peaks involved in the overlap (C). In (C), the black line is the full simulated spectrum with all the compound peaks. (D) The acetate peak was simulated with chemical shift variation under different pH conditions and overlapped with another peak. (E) The extracted concentrations ([C]ex) were plotted against simulated concentrations ([C]sim). (F) Extracted chemical shifts (δex) were plotted against simulated chemical shift (δsim). Performance of concentration and chemical shift estimation with more compounds can be found in Supplementary Figures S5 and S6. Compound concentration was simulated with arbitrary unit (AU) and chemical shift δ was evaluated in ppmCC BY-NC · PubMed Central
Fig. 4. Reproduction of compound quantification results from our previous method. In each plot, the X-axes indicate time, and the Y-axes indicates Mean Scaled Ridge Intensity. Red curves are from aerobic datasets, and blue curves are from anaerobic datasets. Details in computing Mean Scaled Ridge Intensity can be found in the previous publication (Judge et al., 2019). Each row indicates quantification for one compound, including the previous, the RTExtract methods and residuals (differences between the quantifications by the two methods). Comparison for more compounds can be found in Supplementary Figure S7
Fig. 4. Reproduction of compound quantification results from our previous method. In each plot, the X-axes indicate time, and the Y-axes indicates Mean Scaled Ridge Intensity. Red curves are from aerobic datasets, and blue curves are from anaerobic datasets. Details in computing Mean Scaled Ridge Intensity can be found in the previous publication (Judge et al., 2019). Each row indicates quantification for one compound, including the previous, the RTExtract methods and residuals (differences between the quantifications by the two methods). Comparison for more compounds can be found in Supplementary Figure S7CC BY-NC · PubMed Central
Fig. 5. Evaluation of RTExtract on complex overlapping regions on the experimental datasets. Two ROIs (A–B) were selected as examples. Peaks in these ROIs can be precisely tracked, and the parts that are problematic in the previous method are indicated with stars. The middle six peaks in B are annotated to glutamate. Different point colors indicate different tracked peaks. Performance of the algorithm for less complex regions is in Figure 4 and Supplementary Figure S7. More example results from RTExtract and comparison with the previous method are in Supplementary Figure S8. Tracking for B is given as an example in Supplementary File S1
Fig. 5. Evaluation of RTExtract on complex overlapping regions on the experimental datasets. Two ROIs (A–B) were selected as examples. Peaks in these ROIs can be precisely tracked, and the parts that are problematic in the previous method are indicated with stars. The middle six peaks in B are annotated to glutamate. Different point colors indicate different tracked peaks. Performance of the algorithm for less complex regions is in Figure 4 and Supplementary Figure S7. More example results from RTExtract and comparison with the previous method are in Supplementary Figure S8. Tracking for B is given as an example in Supplementary File S1CC BY-NC · 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

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

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

Cited by

Sorted by citations

See all 8

Related

See all

Metadata

Authors
Wu, Y., Judge, M. T., Arnold, J., Bhandarkar, S. M., & Edison, A. S.
Affiliations
  • Institute of Bioinformatics
  • Department of Genetics
  • Department of Computer Science
  • University of Georgia (US)
Corresponding author
Edison, Arthur S
ORCID
Venue
Bioinformatics
Publisher
Oxford University Press (OUP)
Published
2020-07-12
Volume
36
Issue
20
Pages
5068-5075
ISSN
  • 1367-4803 (ISSN-L)
  • 1367-4811
DOI
10.1093/bioinformatics/btaa631
PMC
PMC7755419
OpenAlex
W3031610872
Cited by
8
References
30
Citations by year
2026: 1 · 2024: 1 · 2023: 3 · 2022: 2 · 2020: 1
FWCI
0.49
Type
journal article
Language
en
Open access
Licence
Locations
  • Bioinformatics (publishedVersion, cc-by-nc)
  • PubMed Central (publishedVersion, cc-by-nc)
  • PubMed (publishedVersion)
Funders
  • National Science Foundation [NSF (1713746)
  • Georgia Research Alliance
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Metabolomics and Mass Spectrometry Studies
  • Physical Sciences › Chemistry › Spectroscopy › Advanced NMR Techniques and Applications
  • Physical Sciences › Chemistry › Analytical Chemistry › Spectroscopy and Chemometric Analyses
Keywords
NMR quantification, 3D surface analysis, peak overlap, peak shift, metabolic dynamics, computer vision, greedy algorithm, metabolomics
MeSH terms
Algorithms, Magnetic Resonance Imaging, Magnetic Resonance Spectroscopy, Software, Metabolomics

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