Machine learning in NMR spectroscopy.

NMR spectroscopy is a versatile technique for studies of molecular structures, dynamic processes, and intermolecular interactions across a broad range of systems, including small molecules, macromolecules, biomolecular assemblies, and materials in both solution and solid-state environments. As the complexity of NMR studies continues to pose challenges for practitioners, the integration of machine learning is recognized as a promising research direction for improving data acquisition, processing, and analysis. Here, we summarize recent findings in this area, highlighting common applications such as signal detection, chemical shift assignment, structure determination, chemical shift prediction, non-uniform sampling reconstruction, and denoising.

Super-resolution triple-resonance NMR spectroscopy for the sequential assignment of proteins.

To study the structure and dynamics of proteins by nuclear magnetic resonance (NMR), sequence-specific assignment is needed, which can be obtained by acquiring and analyzing multiple triple-resonance experiments with the three-dimensional TROSY-HNCA, the most sensitive stand-alone experiment with which sequential assignment is, in principle, possible. However, gaining an unambiguous assignment solely from this spectrum is generally not possible because amino acid-type information cannot be gleaned only from the C shifts and the low resolution in the C dimension, which is limited by the homonuclear coupling of the C and C nuclei. Here, super-resolution NMR is applied to the TROSY-HNCA and HNcoCA experiments, yielding pseudo-decoupling, which results in a four- to fivefold resolution enhancement in the C dimension, essential for the assignment, which allows for straightforward assignment of proteins as large as 500 residues based on simulations.

Time-optimized protein NMR assignment with an integrative deep learning approach using AlphaFold and chemical shift prediction.

Chemical shift assignment is vital for nuclear magnetic resonance (NMR)-based studies of protein structures, dynamics, and interactions, providing crucial atomic-level insight. However, obtaining chemical shift assignments is labor intensive and requires extensive measurement time. To address this limitation, we previously proposed ARTINA, a deep learning method for automatic assignment of two-dimensional (2D)-4D NMR spectra.

Chemical shift transfer: an effective strategy for protein NMR assignment with ARTINA.

Chemical shift transfer (CST) is a well-established technique in NMR spectroscopy that utilizes the chemical shift assignment of one protein (source) to identify chemical shifts of another (target). Given similarity between source and target systems (e.g.

NMRtist: an online platform for automated biomolecular NMR spectra analysis.

Summary: We present NMRtist, an online platform that combines deep learning, large-scale optimization and cloud computing to automate protein NMR spectra analysis. Our website provides virtual storage for NMR spectra deposition together with a set of applications designed for automated peak picking, chemical shift assignment and protein structure determination. The system can be used by non-experts and allows protein assignments and structures to be determined within hours after the measurements, strictly without any human intervention.

Rapid protein assignments and structures from raw NMR spectra with the deep learning technique ARTINA.

Nuclear Magnetic Resonance (NMR) spectroscopy is a major technique in structural biology with over 11,800 protein structures deposited in the Protein Data Bank. NMR can elucidate structures and dynamics of small and medium size proteins in solution, living cells, and solids, but has been limited by the tedious data analysis process. It typically requires weeks or months of manual work of a trained expert to turn NMR measurements into a protein structure.

PDBcor: An automated correlation extraction calculator for multi-state protein structures.

Allostery and correlated motion are key elements linking protein dynamics with the mechanisms of action of proteins. Here, we present PDBCor, an automated and unbiased method for the detection and analysis of correlated motions from experimental multi-state protein structures. It uses torsion angle and distance statistics and does not require any structure superposition.

A Mandible with the Temporomandibular Joint-A New FEM Model Dedicated to Strength and Fatigue Calculations of Bonding Elements Used in Fracture and Defect Surgery.

The aim of the study was to develop a new FEM (finite element method) model of a mandible with the temporal joint, which can be used in the numerical verification of the work of bonding elements used in surgical operations of patients with mandibular fractures or defects. Most of such types of numerical models are dedicated to a specific case. The authors engaged themselves in building a model that can be relatively easily adapted to various types of tasks, allowing to assess stiffness, strength and durability of the bonded fragments, taking into account operational loads and fatigue limit that vary in time.

Chemical shift-based identification of monosaccharide spin-systems with NMR spectroscopy to complement untargeted glycomics.

Motivation: A better understanding of oligosaccharides and their wide-ranging functions in almost every aspect of biology and medicine promises to uncover hidden layers of biology and will support the development of better therapies. Elucidating the chemical structure of an unknown oligosaccharide remains a challenge. Efficient tools are required for non-targeted glycomics.

Application of Dirichlet process mixture model to the identification of spin systems in protein NMR spectra.

Analysis of structure, function and interactions of proteins by NMR spectroscopy usually requires the assignment of resonances to the corresponding nuclei in protein. This task, although automated by methods such as FLYA or PINE, is still frequently performed manually. To facilitate the manual sequence-specific chemical shift assignment of complex proteins, we propose a method based on Dirichlet process mixture model (DPMM) that performs automated matching of groups of signals observed in NMR spectra to corresponding nuclei in protein sequence.

NMRNet: a deep learning approach to automated peak picking of protein NMR spectra.

Motivation: Automated selection of signals in protein NMR spectra, known as peak picking, has been studied for over 20 years, nevertheless existing peak picking methods are still largely deficient. Accurate and precise automated peak picking would accelerate the structure calculation, and analysis of dynamics and interactions of macromolecules. Recent advancement in handling big data, together with an outburst of machine learning techniques, offer an opportunity to tackle the peak picking problem substantially faster than manual picking and on par with human accuracy.

A versatile pipeline for the multi-scale digital reconstruction and quantitative analysis of 3D tissue architecture.

A prerequisite for the systems biology analysis of tissues is an accurate digital three-dimensional reconstruction of tissue structure based on images of markers covering multiple scales. Here, we designed a flexible pipeline for the multi-scale reconstruction and quantitative morphological analysis of tissue architecture from microscopy images. Our pipeline includes newly developed algorithms that address specific challenges of thick dense tissue reconstruction.

Computer vision-based automated peak picking applied to protein NMR spectra.

Motivation: A detailed analysis of multidimensional NMR spectra of macromolecules requires the identification of individual resonances (peaks). This task can be tedious and time-consuming and often requires support by experienced users. Automated peak picking algorithms were introduced more than 25 years ago, but there are still major deficiencies/flaws that often prevent complete and error free peak picking of biological macromolecule spectra.