Welcome to the 2026 Global NMR Discussion Meetings Online Conference Hub. Use the drop down menus above to search presentations by topic or to see work from previous years.
-
Frequency-dependent NMR relaxation : Insights into the Structure and Dynamics of ‘WiS’ Electrolytes
Angel Mary Chiramel Tony (University of Rostock, Germany)
LinkedIn: @Angel Mary Chiramel Tony; X: @AngelMaryCT1; Bluesky: @angeltony2000.bsky.social
Abstract: “The “Water-in-Salt Electrolytes” are promising candidates for safer high-voltage aqueous lithium-ion chemistry. Our system of concern for the NMR relaxation investigation is Li-bis(trifluoromethanesulfonyl)imide with H2O/D2O mixtures in 1:3, 1:4 and 1:5 ratios for shedding light on the structure and dynamics of this battery electrolyte as a model system. The frequency-dependant R1(spin-lattice) and R2(spin-spin) relaxation rates for several orders of magnitude is calculated with the computational framework employing MD simulations and applying a correction factor to account for the system size dependency and accessible time scales. Our approach is based on combining the analytical theory of Hwang and Freed (HF) for the long-range intermolecular contribution of the magnetic dipole-dipole correlation function with MD simulations . We show that the correlation functions due to the HF-theory do asymptotically converge with our MD simulation results at long times. We are successful in dissecting the intermolecular and intramolecular contribution of relaxation rates describing the translational and rotational dynamics with the NMR active nuclei 1H and 19F on water and anion molecules respectively. The results show that both longitudinal and transverse relaxation rates increase with increasing temperature and water content indicating enhanced dynamics. Morever the complimentery calculations shed light into the Li ion transport mechanism in the system. The key finding is that Li ion dynamics is compensated with anion repalcing water in highly concentrated systems making it an ideal candidate for battery electrolytes.”
4 responses to “Frequency-dependent NMR relaxation : Insights into the Structure and Dynamics of ‘WiS’ Electrolytes”
-
Nice talk! How do you determine the time range over which the MD correlation function can be safely matched to the Hwang–Freed tail?
-
Hello Yujie,
Many thanks for the interesting question.
For long times, Gn_inter(t) shows a t −3/2 scaling behaviour, and the curve determined from MD simulation asymptotically approaches that of the Hwang and Freed model. The time frame where this happens is given by a time interval T_tr, which can be computed via eqn 27 in the paper below as a proof of concept(in ChemRxiv).
https://chemrxiv.org/doi/abs/10.26434/chemrxiv.15005933/v1
For sufficiently large box sizes, this time-frame is essentially determined by the spin-spin inter-diffusion coefficient D′ and the dimension of the cubic simulation box with box-length L. We have shown that for times t ≥ T_tr both curves are practically indistinguishableIn (Refs.31,32 in Arxiv paper)We determine the time at which the convergence happens, and the time interval at which the application of the correction factor(Gn_inter(t),MD – Gn_inter(t),HF) can be trusted in this manner.
(Detailed in the published papers if you’re interested in the theory and methodology)
https://pubs.aip.org/jcp/article/162/5/054501/3333273/When-theory-meets-experiment-What-does-it-take-to
https://pubs.aip.org/jcp/article/160/7/074102/3265437/Computing-the-frequency-dependent-NMR-relaxationI hope that clarifies your query.
Please feel free to reach out to me if you have more questions or curiosities.Best
Angel -
Hello Angel,
nice results, thank you!
Do you have a (possibly chemical) idea why the 19F _intra_molecular contribution to the relaxation depends so drastically on the number of water molecules?
-
Hello Sergei,
Thanks a lot for your question.
Here the intermolecular contribution for 19F is compromised with the increasing water molecules in the system( W5 being the system with five times no.of water molecules compared to TFSI anion molecules) and intramolecular 19F contribution becomes dominant. Even though it sounds intuitive, it is interesting that the dynamics of Li is not as compromised as TFSI anion in highly concentrated W3 system( close to critical highest concentration,~19 m) which is shown in the non- Gaussian alpha factor calculation for Diffusion of Li,TFSI and water molecules in three different composition. This indicates both water and anion molecules are involved in compensation of Li ion dynamics in such high concentration.
However we don’t have more insights into the chemical phenomenon ( if you meant exchange mechanism etc..) as this study focuses on deciphering the structure and Dynamics of WiS system to shed light into the Li ion transport mechanism.Let me know if I need to clarify anything further.
Looking forward to answering more questions you have.Best
Angel
-
-
Improved 2D-HMQC Spectroscopy Through Perfect Echo Refocusing and ASAP Polarization Transfer
Nidhi Tiwari (Centre of Biomedical Research, India)
LinkedIn: Nidhi Tiwari, X: @TiwariNidhi05
Abstract: The speed of multidimensional NMR spectroscopy can be increased by an order of magnitude by shortening the recycle delay between scans. The consequent loss of longitudinal magnetization due to incomplete relaxation can be retrieved if undisturbed polarization is transferred from nearby proton spins not directly attached to 13C. In ASAP (Acceleration by Sharing Adjacent Polarization) HMQC, an ASAP block based on homonuclear Hartmann-Hahn mixing is incorporated, which consistently provides higher signal enhancement under identical total preparation time by transferring polarization from 12C attached (donor) protons to 13C attached (acceptor) protons, leading to repeated revival of detectable magnetization during short recovery delays.[1] Later on, this ASAP mechanism was also demonstrated to be useful for HSQC and NOAH (NMR by Ordered Acquisition using 1H detection); however, in HSQC, further improvement was achieved using the ZIP element, which overcomes the JHH (homonuclear ¹H-¹H J-evolution) modulation of the remote proton transverse magnetization by storing it before the start of the t1 evolution.
Recently, we have been addressing this JHH modulation in the HMQC class of experiments by combining the concept of ASAP and Perfect Echo-based refocusing of JHH in HMQC. This Perfect Echo-based ASAP-HMQC offers performance similar to that of ASAP-HSQC and better than that of only ASAP-HMQC.
Comparative analysis of ASAP-HMQC, ASAP-HSQC, and Perfect Echo ASAP-HMQC will be presented in the work.Leave a Reply
2 responses to “Improved 2D-HMQC Spectroscopy Through Perfect Echo Refocusing and ASAP Polarization Transfer”
-
Getting the best sensitivity out of HSQC and HMQC experiments is hugely important. Which of the sensitivity-enhancing schemes would work with large proteins that are nonetheless too small to warrant TROSY?
-
Thank you for your question. I think for proteins that are moderately large but not in the molecular weight regime where TROSY becomes advantageous, ASAP-based polarization transfer is likely to provide the most direct sensitivity benefit by utilizing otherwise unused proton magnetization. The perfect echo modifications mainly improve spectral quality by reducing homonuclear J-coupling and line broadening, which can indirectly enhance sensitivity through improved resolution. Therefore, ASAP-based approaches would generally be expected to offer the larger sensitivity gains, while perfect echo refocusing can further improve peak shapes and spectral clarity.
-
-
Isolation of Megastigmanes from Ficus sycomorus and In Silico Design of Novel Cyclohexenone Derivatives as Tubulin Inhibitors for Breast Cancer Therapy
Dauda Garba (University of Abuja Nigeria, Nigeria)
LinkedIn: Dauda Garba
Abstract: Breast cancer demands novel therapies with improved efficacy and reduced toxicity. Tubulin, particularly the colchicine-binding site, is an ideal target for disrupting microtubule dynamics. While medicinal plants offer chiral bioactive compounds, determining their absolute configuration is challenging. This study integrates phytochemistry and computational design to develop cyclohexenone-based tubulin inhibitors. Two megastigmane derivatives, vomifoliol (A1) and its 13-hydroxy analog (A2), were isolated from Ficus sycomorus and characterized via NMR, LCMS, and ECD. Guided by their scaffold and SAR, fourteen derivatives were designed in silico. SwissADME and ProTox-III confirmed drug-likeness and favorable ADMET profiles. Molecular docking against tubulin (PDB: 1SA0) identified six compounds with superior binding (−8.0 to −9.0 kcal/mol) over colchicine (−7.9 kcal/mol). Lig9 showed the strongest affinity (−9.0 kcal/mol), with key interactions at CYS241, LEU242, and ILE378. These results position Lig9 as a promising lead for breast cancer therapy, warranting experimental validation.
Leave a Reply
-
Motional Heterogeneity of Native Bone Collagen under Dehydration Revealed by 13C Solid–State NMR Relaxation Measurements
Bijaylaxmi Patra (Centre of BioMedical Research (CBMR), India)
LinkedIn: @Bijaylaxmi Patra; X: @BijaylaxmiNMR; Bluesky: @bijaylaxmi.bsky.social
Abstract: Collagen, the most abundant protein in animals, is a major structural constituent of muscle and connective tissues and plays a crucial role in determining their mechanical properties. Its structure and stability are strongly influenced by interactions with the surrounding environment, particularly water; however, the molecular basis of these effects remains insufficiently understood. In this study, we probe dehydration–induced changes in collagen dynamics within the native bone extracellular matrix (ECM) using 13C solid–state Nuclear Magnetic Resonance (NMR) relaxation measurements (T1 and T2) along with rotational correlation time analysis to examine water–collagen interactions at the molecular level. Our residue–specific investigation uncovers distinct motional behavior among the aliphatic carbons of the Gly–Pro–Hyp triplet and alanine residues, which together account for nearly 70% of type I collagen. The extracted 13C correlation times reveal substantial motional heterogeneity, where dehydration primarily suppresses the mobility of hydroxyproline Cβ, while H/D exchange significantly affects hydroxyproline Cα, Cβ, and Cγ, as well as glycine Cα. These results provide new insights into hydration–dependent collagen dynamics in native bone and demonstrate the utility of 13C relaxation measurements for investigating water–mediated stabilization and motion in collagen, with broader implications for understanding pathological processes and designing biomimetic materials.
Leave a Reply
4 responses to “Motional Heterogeneity of Native Bone Collagen under Dehydration Revealed by 13C Solid–State NMR Relaxation Measurements”
-
Great presentation! I have two questions:
1. Your results suggest that dehydration and H/D exchange affect different collagen residues in distinct ways. Do you think these residue-specific dynamics are primarily governed by local hydrogen-bonding networks, or could larger-scale structural rearrangements of the collagen fibril also contribute?
2. How the presence (or lack of) minerals like Ca2+ would affect the collagen structure and dynamics?Many thanks!
-
Hi Karen,
1)That’s a very interesting question. Based on our data, we believe the primary effect arises from changes in the local hydrogen-bonding network, particularly the tightly bound water associated with the collagen triple helix. This is supported by the residue-specific responses, especially the pronounced sensitivity of hydroxyproline, which is known to stabilize collagen through water-mediated hydrogen bonds. However, we cannot completely rule out larger-scale structural rearrangements. Dehydration may also reduce intermolecular spacing between collagen molecules and alter fibril packing, which could contribute to the observed changes in dynamics. Therefore, we think the residue-specific behavior likely results from a combination of local hydrogen-bonding changes and subtle structural reorganization of the collagen fibril, although our current NMR data are more directly sensitive to the local molecular dynamics.2) Bone is a composite material where collagen and hydroxyapatite are intimately connected, so the mineral phase plays an important role in collagen dynamics. Calcium ions help stabilize the mineral phase, which mechanically constrains the surrounding collagen network. If the mineral content were reduced or removed, we would expect the collagen matrix to become less constrained and more mobile, leading to faster molecular motions and likely shorter rotational correlation times. In our study, however, the mineral composition was kept unchanged. We only altered the hydration state, so the changes we observed can primarily be attributed to differences in water-mediated interactions rather than changes in mineral content. Studying the combined effects of mineral loss and hydration would indeed be an interesting direction for future work.
-
Excellent presentation!
When you dehydrate the samples, to what degree are water molecules removed, i.e. what is the percent mass attributed to water that is successfully removed? I don’t have a sense for distances between collagen supramolecular fibers, but as a crystalline structure with voids present between fibers, what happens to the distances between them upon dehydration? Are the samples denser and are there increased surface interactions between fibers? -
Hi Sean,
Thank you for the question.
In our study, the bone samples were lyophilized for 24 hours to induce dehydration. This protocol is commonly used to remove free and loosely bound water, while a fraction of the tightly bound water associated with the collagen matrix and mineral interface remains. We did not perform a gravimetric analysis to determine the exact percentage of water removed, so I cannot quote a precise mass loss. Our assessment of dehydration is based on the changes observed in the ¹H NMR spectra, which clearly indicate a substantial reduction in the mobile water population.Regarding the structural changes, dehydration is expected to reduce the water layer between collagen and the mineral phase. Previous solid-state NMR work by Ratan Rai(https://doi.org/10.1021/jp2025768), which also used 24-hour lyophilization, showed that dehydration decreases the distance between collagen and the hydroxyapatite surface, indicating that the collagen network becomes more closely packed. They also reported reduced bound-water intensity after one day of dehydration and concluded that collagen comes closer to the mineral surface as the hydrogen-bonding network weakens.
Therefore, yes, dehydration is expected to make the matrix denser, with reduced intermolecular and collagen–mineral spacing and consequently stronger intermolecular interactions. This interpretation is also consistent with our relaxation data, where most residues show longer T₁ values and increased rotational correlation times, indicating restricted molecular motion and a more rigid collagen framework. While we did not directly measure fibrillar spacing using X-ray scattering or electron microscopy, our NMR data are fully consistent with tighter collagen packing upon dehydration.
-
-
NMR-based serum metabolomics of dementia patients
Rimjhim Trivedi (Centre of Biomedical Research, India)
LinkedIn: Rimjhim Trivedi, X: @rimjhimtrivedi1
Abstract: Cognitive decline is a major age-related neurodegenerative condition lacking minimally invasive biomarkers. Systemic metabolic changes often mirror brain bioenergetic dysfunction, offering potential indicators of disease progression. This study used quantitative 1H NMR-based serum metabolomics to identify metabolic signatures linked to cognitive decline (CD, n = 40) compared with cognitively normal controls (NCD, n = 54). High-field 800 MHz 1H NMR spectroscopy was applied, followed by multivariate (PCA, OPLS-DA, Random Forest) and univariate analyses. Receiver operating characteristic (ROC) analysis assessed diagnostic performance, while correlations with Mini-Mental State Examination (MMSE) scores evaluated clinical relevance. Thirty-three metabolites and key ratios were quantified. CD subjects showed elevated glucose, pyruvate, acetate, acetoacetate, creatine, mannose, and urea, with reduced lactate, alanine, histidine, glutamine, glutamate, and branched-chain amino acids. Ratios including glutamine-to-glucose (QGR), alanine-to-glucose (AGR), and histidine-to-tyrosine (HTR) were significantly decreased. AGR and QGR declined with disease severity, and AGR correlated positively with MMSE scores. ROC analysis demonstrated strong diagnostic performance for QGR (AUC = 0.876), AGR (AUC = 0.840), and HTR (AUC = 0.822). Pathway analysis indicated impaired glycolysis-mitochondrial coupling, reduced pyruvate dehydrogenase activity, altered amino acid metabolism, and compensatory ketone/fatty-acid utilization. These findings highlight serum metabolomics as a promising non-invasive approach for assessing cognitive decline and monitoring progression.
Leave a Reply
-
NMR-guided structural and dynamic characterization of β-PIX SH3 domain for the design of peptide inhibitors targeting oncogenic protein-protein interactions
Srijon Sen (IIT Kharagpur, India)
Abstract: Protein–protein interactions (PPIs) mediated by Src homology 3 (SH3) domains are central to oncogenic signaling and represent challenging therapeutic targets. In breast cancer, the β-PIX SH3 domain interacts with the proline-rich region of Cbl, sequestering it and impairing EGFR downregulation, thereby promoting aberrant signaling. Here, we employ solution-state NMR spectroscopy to structurally and dynamically characterize the β-PIX SH3 domain and to elucidate its interaction with designed peptide inhibitors.
Uniformly ¹⁵N-labeled β-PIX SH3 domain was expressed and purified, and ¹⁵N–¹H HSQC spectra exhibited well-dispersed resonances indicative of a folded protein. Backbone resonance assignments were achieved with high completeness. Secondary structure propensity analysis confirmed a canonical SH3 fold comprising five β-strands and a short 3₁₀ helix. ¹⁵N relaxation measurements (R₁, R₂, and heteronuclear NOE) revealed a structurally stable core with localized flexibility in the RT loop, a key determinant of ligand binding.
Peptide inhibitors were designed using computational approaches and molecular dynamics simulations. NMR titration experiments showed significant chemical shift perturbations upon peptide binding, enabling mapping of the interaction interface on the SH3 domain. Cyclic peptides exhibited enhanced binding compared to linear counterparts, as evidenced by larger perturbations and improved binding affinities. These observations were further supported by isothermal titration calorimetry and fluorescence anisotropy measurements.
Overall, this study demonstrates the power of NMR in resolving structural, dynamic, and interaction features of the β-PIX SH3 domain and provides a framework for the rational design of peptide-based inhibitors targeting SH3-mediated PPIs in cancer.
Leave a Reply
2 responses to “NMR-guided structural and dynamic characterization of β-PIX SH3 domain for the design of peptide inhibitors targeting oncogenic protein-protein interactions”
-
Thank you very much for the nice presentation.
Did you identify any common features among the 9 selected peptides? Are they similar from a structural point of view?
Did you measure any difference in the relaxation properties of your protein upon addition of the peptides?
Would it be possible to follow the interaction also from the peptide point of view rather than from the protein point of view? -
Thanks, Marco, for the encouraging remarks on the presentation and for these insightful questions. I will try answering them.
In response to your question: Did you identify any common features among the 9 selected peptides? Are they similar from a structural point of view? Yes, all 9 peptides share a common 14-residue scaffold (G-x2-P-S-I-D-x7-S-T-K-x11-S-x13-C) built on the invariant Pro3–Thr9–Lys10 core of the class-II PxxP/basic-residue motif inherited from the parent Cbl PRR template. Diversification is limited to just four positions (2, 7, 11, 13), the sites the DDmut-PPI scan flagged as most stabilizing (ΔΔG < 0). The series is therefore a focused affinity-maturation set on one conserved motif, not nine independent ligands. Consistent with this, AlphaFold3 models show all nine docking into the same β-PIX SH3 groove in a shared polyproline-II-like pose, with substitutions only probing adjacent subpockets, and SH3-side CSP mapping (T18, S26, E37, H46) supports this same convergent binding surface.
In response to your question: Did you measure any difference in the relaxation properties of your protein upon addition of the peptides? I have not yet repeated this in the peptide-bound state. The 15N relaxation data shown were collected on the apo β-PIX SH3 domain only, establishing intrinsic ps–ns flexibility in the loops and hinting at μs–ms exchange even unbound. We plan to pursue this next by comparing bound vs apo-state relaxation, which would test whether binding rigidifies these loops or alters the exchange contribution to R₂, and CPMG relaxation dispersion would let us distinguish conformational selection from induced fit.
And finally, in response to your question: Would it be possible to follow the interaction also from the peptide point of view rather than from the protein point of view? Yes definitely. Since the peptides are produced recombinantly, isotope labeling is directly feasible with our existing expression pipeline. Two peptide-observed approaches follow naturally:
1. Reciprocal titration: titrate unlabeled SH3 into ¹⁵N- (or ¹⁵N/¹³C-) labeled peptide and follow the peptide's own amide shifts, giving a residue-level epitope on the peptide side to compare against the SH3-side CSP map.
2. trNOE/STD-NMR: with sub-stoichiometric SH3 and unlabeled peptide, transferred-NOE would test whether the free (likely disordered) peptide adopts the predicted polyproline-II pose only upon binding.
These would directly complement the protein-observed titration and CSP data already shown.
-
-
Nuclear spin-induced envelope modulation in single NV diamond magnetometry
Abhishek Joshi (Indian Institute of Science Education and Research Bhopal, India)
LinkedIn: Abhishek Joshi
Abstract: Nuclear spins are an ubiquitous probe of the atomic and molecular environment. However, the Nuclear Magnetic Resonance signal does not provide the information related to the hyperfine coupling strength and the nature of the coupled electron. On the other hand, Electron Spin Resonance (ESR), weak magnetic fields from nuclear spins coupled to an electron spin are detected through envelope modulation arising from hyperfine coupling. In this context, the nitrogen-vacancy (NV) center in diamond is a versatile platform for room-temperature ESR studies: its features are optically induced spin polarization at room temperature, coherent manipulation with microwaves, and optical readout via spin-dependent fluorescence.
Prior work has shown that a magnetic field transverse to the NV axis induces electron-nuclear state mixing, thus enabling forbidden microwave transitions that produce envelope modulations in spin-echo experiments. Recently, analogous ^15N-induced modulations were observed in the NV coherence measured in Ramsey experiments performed at low fields (∼ 10 mT). Here, we extend these studies to higher magnetic fields of 45 mT approaching the NV center’s excited-state level anti-crossing (ESLAC), a regime where the current theoretical description based on the perturbation theory fails.
By analyzing the experimental data, we extracted hyperfine parameters, enhancement factors of the nuclear gyromagnetic ratio, and the conditional nuclear Larmor frequencies. Our findings are significant for developing nuclear spin-based quantum sensors and nanoscale quantum sensing in biological systems where NV-nanodiamond probes function in environments with poorly controlled or misaligned magnetic fields.Leave a Reply
3 responses to “Nuclear spin-induced envelope modulation in single NV diamond magnetometry”
-
Hello!
Thanks for the presentation. I have two questions:
1. Can this modulation effect be exploited to study different close-by spins, e.g. 13C? Would a higher gamma be in favour of the effect?
2. Is the experimental linewidth of 15N nuclear spectrum always determined by T2? Is it feasible to expect an underlying distribution of hyperfine coupling parameters?
Thank you!
-
Hi,
Very interesting work. Can you elaborate more on what was written in the abstract regarding the breakdown of perturbation theory at higher fields?
Thanks! -
Hi Abhishek, wonderful talk!!!
-
-
NV-Centre-Based Deuterium NMR: Chemical Information at Low Field with Interfacial Sensitivity
Utsab Banerjee (Technical University of Munich, Germany)
LinkedIn: Utsab Banerjee, X: @UtsabBan1729, BlueSky: @utsab1729.bsky.social
Abstract: Nitrogen-vacancy (NV) centres in diamond are powerful quantum sensors capable of performing NMR spectroscopy on spin systems far beyond the reach of conventional inductive detection. Positioned a few nanometers beneath the diamond surface, these optically addressable defects sense statistical nuclear spin fluctuations from nanoscale volumes, circumventing the Boltzmann polarisation bottleneck that demands large samples and high magnetic fields. Deuterium (²H) is an ideal target: its quadrupolar Pake patterns encode molecular geometry and dynamics, with intrinsic breadths naturally matching NV-accessible spectral resolution. Resolving such lineshapes at the nanoscale has remained an outstanding challenge.
Here, we demonstrate nanoscale ²H NMR using shallow NV centre ensembles in an isotopically enriched diamond chip, at magnetic fields orders of magnitude lower than conventional spectrometers. Correlation spectroscopy protocols built on dynamical decoupling sequences yield powder-like ²H quadrupolar Pake patterns from deuterated samples on the diamond surface, directly comparable to bulk solid-state ²H NMR.
NV-NMR delivers a spin sensitivity many orders of magnitude beyond conventional inductive detection while retaining the rich spectral information of the quadrupolar interaction, fittable with standard solid-state NMR software. Variable-temperature measurements reveal distinct interfacial physics: the polymer shows suppressed dynamics consistent with an elevated local glass-transition temperature at the diamond surface, while the molecular solid undergoes a progressive lineshape collapse tracking its solid-liquid phase transition. This work establishes NV-based quadrupolar ²H NMR as a new sensing modality for probing molecular dynamics at surfaces and interfaces, with ultimate prospects for single-molecule detection.Leave a Reply
4 responses to “NV-Centre-Based Deuterium NMR: Chemical Information at Low Field with Interfacial Sensitivity”
-
This is very interesting! I noticed that the NV spectra clearly resolves the main Pake horns, but some of the weaker outer shoulders visible in the bulk spectra and simulations seem to be suppressed. Is this limited by the XY8 bandwidth or the detection sensitivity? Have you considered stepping the XY8 filter centre across the quadrupolar spectrum and stitching together several frequency windows?
-
Thank you for your comment. Actually, it’s a bit of both. The XY8-N filter has a finite passband , so spectral weight at the outer shoulders — which is already lower-intensity in the powder average — sits closer to the edge of that window and gets less filter gain to begin with. Combined with our current sensitivity, that puts the weakest crystallite orientations right near our detection floor.
Stepping the filter centre and stitching windows is exactly the approach we’re currently pursuing. Thanks again for the comment.
-
Impressive!
What would be the lowest field (approximately) at which deuterium could be detected with this method?
-
The practical limit is set by our own coherence time: XY8-N is a resonant technique, so the pulse spacing τ has to match half the deuterium Larmor period. As field drops, the Larmor frequency drops too, so τ has to get longer — but τ is capped by how long the NV can maintain phase coherence under decoupling (our T2 in this experiment). So there’s a lowest Larmor frequency (700 kHz), and hence lowest field ( 100 mT), below which we simply can’t space the pulses far enough to stay on resonance within our coherence window.
-
-
-
Physics-Informed Neural Network for Conformational State Prediction from NMR Spectra
David Joseph (Germany)
LinkedIn: David Joseph; X: @DaJo_1729; Bluesky: @dajo-1729.bsky.social
Abstract: Determining conformational ensembles—and thereby the underlying conformational landscape—of peptides and proteins remains a central challenge in structural biology. Conventional NMR approaches rely on iterative, expert-driven interpretation of chemical shifts, NOEs, and scalar couplings, which can be time-consuming and difficult to scale. We introduce a physics-informed neural network for conformational state prediction directly from NMR spectra. The proposed model is a transformer-based architecture that maps raw 2D NMR experiments—including NOESY, HSQC, and TOCSY spectra—to multi-state conformational ensembles. For each state, the network predicts backbone torsion angles, side-chain dihedrals, inter-residue distances, and corresponding Boltzmann populations. Training is performed on a fully synthetic dataset generated by combining physics-based conformational sampling with a differentiable NMR pulse-sequence simulator, enabling the creation of paired (spectrum, ensemble) data without reliance on experimentally determined structures. Preliminary results on held-out validation data indicate that the model accurately recovers backbone geometries and population distributions from spectral inputs alone. Ongoing work focuses on extending the approach to longer sequences, variable ensemble sizes, and application to experimental NMR datasets.
Leave a Reply
4 responses to “Physics-Informed Neural Network for Conformational State Prediction from NMR Spectra”
-
Nice talk! Where does the physics enter the neural network training?
-
Hi Yujie,
The physics goes into the loss function mainly.
Dav
-
-
Hi,
Interesting work, thanks. Is your neural network limited by the size of the peptide? Can it do small proteins, in the case where the NMR data is good.
Also, how good does the NMR data need to be in terms of SNR and resolution?-
Hi Daphana,
At the moment, the method is limited to peptides of up to about 25 amino acids, as this was a proof-of-concept to test feasibility. I am currently looking into scaling to larger systems, but this requires substantial computational resources, which is not easy to come by these days! In tests with synthetic noise, an SNR of approximately 2–10 was sufficient, although real spectra would provide a more meaningful benchmark.
Dav
-
-
-
Precursor-Dependent Lignin Biosynthesis in Grasses Revealed by DNP-Enhanced Solid-State NMR
Priya Sahu (Michigan State University, United States)
LinkedIn: Priya Sahu
Abstract: Lignin is a major structural component of plant secondary cell walls and a key determinant of biomass utilization. In grasses, both phenylalanine and tyrosine contribute to lignin biosynthesis, but their respective roles in shaping the native lignin polymer have remained unclear. Here, we combine precursor-specific 13C isotope labeling with dynamic nuclear polarization (DNP)-enhanced solid-state NMR to directly track aromatic amino acid incorporation into lignin in intact Brachypodium distachyon cell walls. Conventional solid-state NMR established tissue-specific labeling patterns, while DNP provided up to ~40-fold sensitivity enhancement, enabling multidimensional 13C-13C correlation experiments on selectively labeled samples. We found that phenylalanine is the dominant precursor for canonical guaiacyl and syringyl lignin, whereas tyrosine preferentially contributes to hydroxyphenyl lignin and ferulate moieties characteristic of grass cell walls. Analysis of a C3H knockdown mutant further revealed precursor-dependent metabolic plasticity: phenylalanine-derived lignification was strongly impaired, while tyrosine-derived lignification remained comparatively resilient through alternative metabolic routing. These results demonstrate how DNP-enhanced solid-state NMR can directly connect precursor metabolism with polymer architecture in intact plant cell walls, providing new insights into the metabolic regulation of lignin biosynthesis in grasses.
Leave a Reply
-
Prediction of 13C-NMR Chemical Shifts of Small Organic and Drug Molecules Using Machine Learning
Surajit Das (Tata Institute of Fundamental Research Hyderabad, India)
LinkedIn: Surajit Das, X: surajit_das_x, BlueSky: surajitdas.bsky.social
Abstract: Accurate prediction of NMR chemical shifts is essential for mapping the structure-property relationships of complex molecules. In this work, we use the kernel ridge regression (KRR) machine learning architecture with an in-house-built aBoB-RBF(n) atomic representation to predict 13C-NMR chemical shifts. To train the model, we use the QM9NMR dataset, which contains 130,831 molecules with up to 9 non-hydrogen (CONF) atoms, encompassing nearly 832k carbon atomic environments. This vast pool of chemical diversity allows us to train on 100k data points with a 50k test set, yielding a mean absolute error of 1.69 ppm, an improvement over previously reported models. We also demonstrate model performance on diverse validation sets, including the drug-like Drug12 and Drug40 datasets; the GDBm dataset, which contains an increasing number of CONF atoms; and the Pyrimidinone dataset, which features biologically relevant substituted uracil and pyrimidinone molecules. Additionally, we developed a Python-based mlqm9nmr module, available on GitHub, that allows users to quickly predict chemical shifts using our KRR model. Finally, we developed a web-based interface that uses a given molecule’s SMILES string to predict 13C NMR chemical shifts with the trained KRR model.
Leave a Reply
4 responses to “Prediction of 13C-NMR Chemical Shifts of Small Organic and Drug Molecules Using Machine Learning”
-
Hi Surajit, nice work. Could you elaborate on the choice of the aBob-RBF representation of the molecules. Why does it improve the accuracy of prediction significantly. And how does it compare to other representations of the atomic structure ?
-
The aBoB-RBF descriptor is based upon two-body Coulombic interactions combined with element-specific bags that encode the local chemical environment around a target atom. The continuous representation using the radial basis functions (RBF) creates a smooth, well-behaved descriptor space. Furthermore, incorporation of the nearest-neighbor information, specifically in aBoB-RBF(4), which includes contributions from the four nearest atoms (valency of a carbon atom), substantially enhances the local chemical information.
The constitution of aBoB-RBF(4) relies on two-body interactions; our benchmarking against aSLATM shows that the neighbor representations enable it to effectively capture the many-body interaction effects at a significantly lower computational cost. On the QM9NMR test set of 50,000 carbon shieldings, aBoB-RBF(4) achieves an out-of-sample mean absolute error of 1.69 ppm (https://doi.org/10.1063/5.0306349), outperforming previously reported state-of-the-art models using descriptor such as SOAP and FCHL (1.88 ppm; https://doi.org/10.1088/2632-2153/abe347) and MACE neural-network embeddings (1.87 ppm; https://doi.org/10.1039/d4dd00098f).
-
Very nice work, Surajit!
Since the errors increase for the larger drug molecules, and your descriptors use PM7 geometries while the DFT reference values use B3LYP geometries, have you disentangled how much of this loss comes from geometry mismatch versus model extrapolation? -
We disentangled the geometric differences between B3LYP and PM7, as well as other cheap, force-field-based geometries. While one would definitely expect a mismatch between DFT and semi-empirical or force-field-based methods, it is not very significant for the smaller QM9 geometries. For big drug molecules, using the same DFT-level geometries would be ideal, but the process is computationally more expensive than the NMR prediction itself. We found that the higher error in the larger drug molecules arises mainly from their conformations, which are due to factors such as flexible, rotatable bonds and different intramolecular interactions. In short, the increase in MAE for the larger drug molecules is due to the combined effect of geometric mismatch and conformation, which we are currently addressing in our work. Thank you
-
-
Pulsar Studio: A Graphical Interface for Quantum Optimal-Control Pulse Design
David Joseph (Germany)
LinkedIn: David Joseph; X: @DaJo_1729; Bluesky: @dajo-1729.bsky.social
Abstract: Quantum optimal-control methods routinely outperform analytic shaped pulses for broadband and band-selective spin manipulation and coherence transfer, and can be made robust to instrumental limitations, B_0/B_1 inhomogeneity, and relaxation. Yet their adoption remains limited by a steep tooling barrier, i.e. designing pulse shapes requires installing a scientific programming environment, learning a package API, and writing scripts by hand. We present Pulsar Studio, a cross-platform desktop application that removes this barrier by wrapping the Pulsar.jl optimal-control library behind a guided graphical interface. Crucially, end users need no programming environment and write no code. The application bundles its own Julia runtime and optimization backend, so spectroscopists install a single package and begin designing pulses immediately. During optimization, the interface streams live diagnostics — fidelity convergence, gradient norm, and an updating waveform snapshot — allowing users to judge and cancel runs interactively. Designed pulses can be exported directly to the instrument-ready formats supported by Pulsar.jl. The complete experiment — system definition, constraints, and run history — are saved as portable project files, supporting reproducible and shareable workflows. By making quantum optimal-control pulse design approachable without any coding, Pulsar Studio lowers the barrier for adapting robust pulse engineering into everyday magnetic-resonance practice.
Leave a Reply
6 responses to “Pulsar Studio: A Graphical Interface for Quantum Optimal-Control Pulse Design”
-
Interesting! Have you compared the simulated pulse performance with any experimental results yet?
-
Hi Yujie,
This has not yet been completed specifically using Pulsar results. However, you can refer to my earlier publications (1. https://www.science.org/doi/10.1126/sciadv.adj1133; 2. https://doi.org/10.1016/j.jmr.2026.108081; 3. https://doi.org/10.1016/j.jmr.2026.108030), where similar comparisons were carried out. The insights gained from that work have been incorporated into the package, hopefully allowing users to more easily adapt pulses to a given instrument.
Dav
-
-
It is a nice-looking product! Web-based app?
In EPR, the major source of pulse non-ideality is the non-flat transfer function of microwave components (~ instrumental frequency-dependent power distortion). One can estimate it theoretically (yielding an analytical expression) or measure it experimentally (-> numeric dataset). Does your software allow for taking/uploading either of these formats for pulse compensation?
I hope my question is clear. Thanks!
-
Hi Sergei,
This is not a web-based tool; it needs to be installed locally on a computer. If you would like to register for beta testing, you can do so here: https://forms.gle/hcQmsBsgRMkYX3bo9.
If I understood you correctly, this functionality is not currently included in Pulsar software. If you are interested in such applications and can point me to relevant publications, I would be happy to explore whether it is feasible to incorporate instrument response functions into the optimization process. As an alternative, one can compensate for known distortions in the input pulse shape so that the resulting output matches the desired signal more closely.
Dav
-
-
Very useful tool. Have you compared the simulated B₀/B₁ robustness profiles with experimental profiles after exporting the pulse to a spectrometer, to assess how hardware-dependent waveform distortions affect the achieved fidelity?
-
Hi Shubha,
I am glad you found it useful, If you would like to register for beta testing, you can do so here: https://forms.gle/hcQmsBsgRMkYX3bo9.
Such comparisons of simulated B₀/B₁ robustness profiles with experimental profiles were carried out in one of my earlier work (https://www.science.org/doi/10.1126/sciadv.adj1133) if you feel brave enough please check out the ~90 page long ESI, which contains extensive benchmarks of this sort from experiment and simulations. Hence they were not repeated for the GUI.Dav
-
-
Leave a Reply