NMR Software

NMRtist

A Cloud Based All-in-One Solution for Protein Assignment and Structure

Highlights

Rapid Analysis
Reducing the lengthy process of analyzing and assigning protein NMR spectra from weeks to just a few hours
Improved Accuracy
Using machine learning algorithms to improve the accuracy of spectral assignment and analysis
Intuitive User-Experience
Designed for effortless usage, providing a user-friendly platform to access and analyze protein NMR spectra with ease

New Opportunities for NMR Protein Functional Analysis
Biomolecular nuclear magnetic resonance (NMR), in combination with other structural biology tools like crystallography, Cryo-EM and AlphaFold, is crucial in elucidating and understanding protein dynamics, function, structural rearrangements and ultimately disease pathways.

Automation of NMR Protein Functional Analysis Utilizing Artificial Intelligence (AI)
Advances in NMR technology and computational methods have significantly improved the speed and accuracy of spectral assignments, cementing biomolecular NMR’s role as an indispensable tool in modern functional structural biology. NMRtist1 is a cloud-based, AI-supported software platform developed by ETH Zurich in collaboration with Bruker to enable secure and user-friendly access to protein NMR spectra analysis. Utilizing the advanced ARTINA2,3 algorithm, NMRtist automates the lengthy and manual process of protein spectra analysis, including peak picking and assignment, reducing the timeline from weeks to just a few hours. This game-changing software provides unprecedented efficiency and reliable results with improved accuracy, making biomolecular NMR a more competitive tool for researchers in the field of structural biology.

Features & Benefits

NMRtist at a Glance:

  • Cloud based, AI supported software platform
  • High performance, secure cloud architecture
  • Upcoming integration of automated assignments into other workflows
  • Fully automated peak picking of multi-dimensional NMR spectra
  • Performs resonance assignment in no time
  • Fully automated structure determination

 

Cloud Computation and AI Taking the Tedious Tasks off Your Hands

The ARTINA algorithm consists of the three individual steps which can be executed separately or in consecutive order.

Peak Picking

The initial step in analyzing NMR spectra is typically peak picking. In biomolecular NMR, these spectra often have three or more dimensions, making manual peak picking or even manual inspection of peak lists an extremely time-consuming process. To address this challenge, NMRtist employs a neural network trained on a meticulously curated set of datasets, primarily comprising three- and four-dimensional protein spectra. This approach significantly contributes to the overall time savings achievable with the NMRtist platform.

Slice out of a peak picked 3D 13C-edited NOESY-HSQC

Resonance Assignment

The second step is resonance assignment. What usually is a days- or weeks-long undertaking can be automated and executed in a matter of minutes or hours. The ARTINA shift assignment application first uses a deep convolutional neural network to detect positions of signals in the selected NMR spectra (see ARTINA peak picking application above). Afterwards, the detected signals undergo FLYA4 automated chemical shift assignment. The method returns protein chemical, together with assigned peak lists for the individual NMR spectra.


Backbone assignment overview including confidence levels for a 19.5 kDa protein

3D Structure Calculation

The third step is 3D structure calculation. When proteins or other biomolecules cannot be crystallized, or when AI-based predictions need verification or refinement, NMR-based structure determination becomes essential. However, this process is traditionally time-consuming and often requires an expert familiar with specialized software packages. The ARTINA structure determination application simplifies this by implementing end-to-end protein structure solving. Given a set of spectra and a protein sequence as input, the application generates the aforementioned peak lists and assignments, followed by a fully automated structure calculation using CYANA5.

Example of a 19.5 kDa cancer related protein structure, determined using NMRtist. Total measurement time was 2 weeks followed by a 24 hour NMRtist run

Other Applications

Protein Dynamics

Easy access to assignments via the NMRtist platform benefits a wide range of applications, with the determination of backbone dynamics being just one example. A minimal set of experiments was utilized to obtain backbone amide assignments. By combining rapid pulsing and sparse sampling techniques, the total measurement time for a small protein was reduced to just under 2 hours.

 

 

NMRtist successfully achieved a complete backbone assignment, which was then used alongside relaxation measurements to assess the rigidity of the protein backbone using DynamicsCenter™. The integration of bioTop for setup and acquisition, NMRtist for obtaining assignments, and DynamicsCenter for advanced data analysis creates an almost fully automated, straightforward method.

Results of automated backbone assignment using a minimal set of experiments



Lipari-Szabo type order parameters for ubiquitine shown as histogram (left) of heatmap (right)

Next Steps

Working together with the developers at the ETH in Zurich, Bruker will provide the users with a high performance, secure platform to run ARTINA jobs. Security will be provided using state of the art cloud infrastructure as well as regular security audits by 3rd parties.

In addition, Bruker will provide and interface from within TopSpin and bioTop to integrate the NMRTist platform seamlessly into other workflows, for example using automated assignment results to determine protein dynamics.

More Information

References

  1. Klukowski, P., Riek, R. & Güntert, P. NMRtist: an online platform for automated biomolecular NMR spectra analysis, Bioinformatics, 2023.
  2. Klukowski, P., Riek, R. & Güntert, P. Rapid protein assignments and structures from raw NMR spectra with the deep learning technique ARTINA (2022). Nature Communications 13, 6151.
  3. Klukowski, P., Riek, R. & Güntert, P. Time-optimized protein NMR assignment with an integrative deep learning approach using AlphaFold and chemical shift prediction, Science Advances, 2023.
  4. Schmidt, E., & Güntert, P. (2012). A new algorithm for reliable and general NMR resonance assignment. Journal of the American Chemical Society, 134, 12817-12829.
  5. Güntert, P. & Buchner, L. (2015). Combined automated NOE assignment and structure calculation with CYANA. J. Biomol. NMR 62, 453-471.