Molecular replacement (MR) is the primary method used to solve the phase problem in macromolecular crystallography. When there are no suitable homologues available for conventional MR, one alternative option is to predict the structure via bioinformatic means. This is referred to as ab initio or de novo modelling and has dramatically increased in accuracy in recent years with the availability of better residue-contact predictions derived through covariance analysis and deep learning. Covariance-assisted ab initio models are now available on a large scale in new generation databases, and therefore provide an easily obtainable source of potential search models as a supplement to the PDB. We have previously shown that using such structure predictions obtained from the GREMLIN and PconsFam databases could be processed with AMPLE to successfully solve structures through MR [1]. Here we explore the use of alternative model sources such as DMPfold and Alphafold2, considering which model preparation protocols are optimal.

We also present MrParse, a tool to access structures deposited to the PDB and predicted structures deposited in databases in a single, unified interface. This resource, which will also enable facile access to the latest automated homology models, will also present model quality assessment scores to help assess the quality of database-derived models. Presenting PDB structures and database-derived models in a common dashboard will help the crystallographer maximise the chance of success through MR.

[1] Simpkin et al. (2019) Acta Cryst D75(12) 1051-1062

The molecular recognition of carbohydrates by proteins plays a key role in many biological processes including immune response, pathogen entry into a cell and cell-cell adhesion (e.g., in cancer metastasis). Carbohydrates interact with proteins mainly through hydrogen bonding, metal-ion-mediated interaction and non-polar dispersion interactions. The role of dispersion-driven CH-π interactions (stacking) in protein-carbohydrate recognition has been underestimated for a long time considering the polar interactions to be the main forces for saccharide interactions. However, over the last few years it turns out that non-polar interactions are equally important. Using the Protein Data Bank (PDB) structural data, we analyzed the CH-π interactions employing bioinformatics (data mining, structural analysis), several experimental (ITC, X-ray crystallography) and computational techniques. Within 12 000 protein complexes with carbohydrates from PDB, the stacking interactions were found in about 39% of them. The calculations and the ITC measurement results suggest that the CH-π stacking contribution to the overall binding energy ranges from 4 kcal/mol up to 8 kcal/mol. All the results show that the stacking CH-π interactions in protein-carbohydrate complexes can be considered to be a driving force of the binding in such complexes.

Protein folding relies on the formation of secondary structures as helices, beta-strands, turns, with specific values of the backbone torsion angles ϕ and ψ for each secondary structure.

β-turns represent the most prevalent type of nonrepetitive secondary structure in proteins. A β-turn is a region of four consecutive residues, where the polypeptide chain reverses its direction and the distance between the α-carbon atoms of the residues i and i+3 is less than 7 Å.

In 1968 Venkatachalam recognized the existence of β-turns as a result of a conformational study of four consecutive amino acid residues [1].He evidenced three distinct conformations characterized by specific values of the phi, psi torsion angles and by the presence of a hydrogen bond between the peptide backbone carbonyl group of the first residue C=O(i) and the backbone amino group of the fourth residue N-H(i+3).

In the next 50 years of research, several classifications of beta-turns were proposed, based exclusively on the evaluation of the dihedral angles phi and psi [2].

Recently, Newberry and Raines evidenced the importance of weak chemical interactions in the formation of protein secondary structures [3].

Thus, in this work we aimed to identify repeated patterns of n → π* interactions between carbonyl groups of successive residues in proteins and cyclic peptides. The survey considered 1424 X-ray protein structures in the Protein Data Bank with a resolution of 1.2 Å or better, R-factor of 0.2 or better and sequence identity of 50% or lower. We also performed a statistical analysis on the geometrical feature of CO…CO interactions in turn mimetic compounds as cyclic peptides, cyclic depsipeptides and cyclic peptoids, considering a total of 232 compounds.

The obtained results show that the n → π * interactions could allow to discriminate among different turn types and explain the peculiar differences from a chemical point of view.

References

[1] Venkatachalam, C. M. Stereochemical criteria for polypeptides and proteins. V. Conformation of a system of three linked peptide units. Biopolymers 1968, 6, 1425-1436.

[2] Shapovalov, M.; Vucetic, S.; Dunbrack, R.L. A new clustering and nomenclature for beta turns derived from high-resolution protein structures. PLoS Comput. Biol. 2019, 15, e1006844.

[3] Newberry, R. W.; Raines, R. T. Secondary Forces in Protein Folding. ACS Chem. Biol. 2019, 14, 8, 1677–1686. (https://pubs.acs.org/doi/10.1021/acschembio.9b00339

Breast cancer type 2 susceptibility (BRCA2) protein plays an essential role in the repair of DNA double-strand breaks and interstrand cross-links by Homologous recombination [1]. Germ-line mutations in BRCA2 confer an increased risk of hereditary breast and ovarian cancer [2]. A large number of missense mutations have been identified in the DNA binding carboxy-terminal domain of BRCA2 which is also known to interact with FANCD2 [3]. However, majority of these missense mutations are classified as variants of ‘Uncertain Significance’ due to lack of structural, functional, and clinical studies. Accurate and reliable methods to predict the pathogenicity of variants are utmost required for better clinical management of the disease. Here we present a multi-disciplinary approach to characterize a missense mutation identified in the C-terminal domain of BRCA2. Different functional domains of the wild-type and mutant BRCA2 protein were cloned and the proteins were expressed and purified in bacterial system. Circular -dichroism and Fluorescence spectroscopic techniques were employed to evaluate the differences between secondary and tertiary structures of wild-type and mutant protein. Molecular Dynamics Simulation was further utilized to measure the effect of mutations on the structural conformation of the protein.

References:[1] Xia, B. et al. (2006). Molecular Cell, Vol. 22, 719-729[2] Venkitaraman, A.R. (2002). Cell, Vol. 108, 171–182[3] Hussain S, et al. (2003). Human Molecular Genetics, Vol. 12, No. 19, 2503-2510

Keywords: BRCA2; Molecular Dynamics Simulation

As the number of macromolecular structures solved by the electron microscopy method (EM) rapidly increases, the need for improved methods for processing and improving all aspects of the EM structure determination methods also grows. One possible approach to improving the processing of experimental data is by using the symmetry information. To this end, all major cryo-EM software suites provide an option to use the symmetry information to improve the final resolution of the solved structure. This technique is based on averaging density over all asymmetric units, thus reducing noise and increasing signal to noise ratio in the data.

Nonetheless, all currently available EM suites do require the user to supply the symmetry of the structure and sometimes rotating the structure so that the symmetry axes are in a particular orientation relative to the system axes in order to work. In this contribution, we present a novel method for determining density map symmetries de novo as well as a new software tool called ProSHADE implementing this method.

The presented method relies on computing the optimised self-rotation function [1] using the spherical harmonics decomposition coefficients and converting these onto SO(3) space coefficients as described by [2]. By subsequently computing the inverse Fourier transform in SO(3) space, the self-rotation function is obtained. Next, the self-rotation function values are mapped onto a set of concentric spheres with radius equal to the angle of rotation in the axis-angle rotation representation, while the position on the sphere represents the rotation axis of the axis-angle rotation representation. This representation allows for fast detection of any axis which has high self-rotation function values along any particular set of angles. This in turn allows for detection of cyclic (C) symmetry groups as they are by definition a set of rotations, along the same axis, which do not change the shape (i.e. have high self-rotation function value). Once all C symmetries are detected, the dihedral (D), tetrahedral (T), octahedral (O) and icosahedral (I) symmetries can be detected by finding for the required C symmetries combinations forming the larger symmetry groups.

Since the self-rotation function values are proportional to real-space correlation between the original and rotated density map values, they are affected by the shape of the density map; this relationship is such that the more spherical the density map is, the higher the overall real-space correlation will be irrespective of the actual symmetry in the density. Therefore, to reduce the number of false positive results, the method also uses the Fourier Shell Correlation (FSC) to confirm any detected symmetry, increasing the reliability of the method.

The current implementation of this method in ProSHADE is capable of correctly detecting the symmetry type and fold for over 85% of symmetrised structures deposited in the EMDB database [3] with approximately half of the incorrectly determined symmetries being a subgroup of the originally reported symmetry. The symmetry detection does not require any user input and is therefore readily available for inclusion into cryo-EM structure solving pipelines. ProSHADE is an open source project available under the GPL version 3 license on all major operating systems either as stand-alone executable or as a python language module.

References:

[1] Navaza J. (1994). Acta Cryst, A50, 157-163.[2] Kostelec P.J., Rockmore D.N. (2008). Journal of Fourier Analysis and Applications, 14, 145–179.[3] Lawson C.L., Baker M.L., Best C., Bi C., Dougherty M., Feng P., van Ginkel G., Devkota B., Lagerstedt I., Ludtke S.J., Newman R.H., Oldfield T.J., Rees I., Sahni G., Sala R., Velankar S., Warren J., Westbrook J.D., Henrick K., Kleywegt G.J., Berman H.M., and Chiu W.C. (2011). Nucleic Acids Research, 39, 456-464.

Reliable determination of the quaternary structure of a protein is often crucial to a full understanding of its function. However, for decades, crystallographers have sometimes struggled to distinguish between biologically meaningful interfaces observed in a crystal structure, and the unnatural lattice contacts that allow for crystal formation. In order to solve this problem, Eugene Krissinel developed PISA, and later jsPISA, a CCP4 tool that sorts different candidate quaternary structures according to a likelihood obtained from results on dissociation free energy [1, 2]. Additionally, jsPISA incorporates an "interaction radar" that allows for a rapid visualisation of how likely an interface is according to a number of physico-chemical parameters.

In order to provide jsPISA with an additional -- independent -- source of information to determine the probability of a given interface to biologically exist, we propose the use of evolutionary covariance data. This proposal is based upon the observation that pairs of residues whose interaction contributes to a biologically important interface are constrained in their evolution [3, 4]. Consequently, the detection of a pair covariation signal points at the existence of a contact between the two residues contributing to the formation of a biologically important interface. The new extension, named PISACov, aims to enhance the results currently displayed by jsPISA with an additional score and new data based on evolutionary covariance analysis, thereby helping determine the relevant quaternary structure in difficult cases.

[1] Krissinel, E.; Henrick, K. J. (2007) Mol Biol., 372 , 774-797.
[2] Krissinel, E. (2015) Nucleic Acids Research, 43, W314–W319.
[3] Ovchinnikov, S.; Kamisetty, H.; Baker, D. (2014) eLife, 3:e02030.
[4] Hopf, T.A. et al. (2014) eLife, 3:e03430.

The Protein Data Bank in Europe - Knowledge Base (PDBe-KB, https://pdbe-kb.org) is a community-driven, collaborative data resource that provides literature-derived, curated and predicted structural and functional annotations of molecular structure data. Consortium members provide annotations including catalytic sites, ligand binding sites, protein flexibility, post-translational modification sites, and the effect of genetic variability or mutations. PDBe-KB aims to increase the visibility and reduce the fragmentation of these annotations and place macromolecular structure data in their biological context, thus facilitating their use by the broader scientific community in fundamental and applied research.

PDBe-KB currently collaborates with 31 resources from 11 countries, and we integrate their annotations with core PDB structural data in a novel and distributable PDBe graph database. Researchers can access all the annotations either by using the graph database or programmatically via API endpoints. We have also created web pages called “Aggregated Views” that provide an overview of all the structure data related to a full-length protein (i.e. UniProtKB accession). These views are better at displaying the biological context of proteins instead of the conventional PDBe entry-page focus on a single PDB entry.

We have been continuously improving and expanding PDBe-KB since its inception in 2018. Last year we rolled out a significant update to support the global effort to tackle the COVID-19 pandemic by creating dedicated pages gathering all the structural information for the new coronavirus SARS-CoV-2 (e.g. https://www.ebi.ac.uk/pdbe/pdbe-kb/protein/P0DTC2 ). We have also added new functionalities, such as viewing the superposition of every PDB chain that corresponds with a protein of interest. These views are instrumental in identifying ligand-binding hotspots and conformationally flexible regions.

Evolution has led to proteins being able to specifically bind molecules. They have evolved to bind a great variety of chemical substances, from carbohydrates to small organic molecules as well as ions or other macro molecules. This power of evolution can also be harnessed in silico, using applications like Rosetta, to engineer proteins so that they bind ligands with higher specificity.

However, these bioinformatical algorithms are limited as they often cannot take all aspects of molecular modeling, like protein flexibility, solvent simulation, or energy calculation, reliably into account. In addition, most complex applications are usually difficult to use for novices.

Here, we present EvoDock, a modular and easy-to-use pipeline which is capable of integrating multiple molecular modeling programs into an evolutionary algorithm, to predict optimized variants of ligand-binding proteins.

Furthermore, we demonstrate how its predictions could be confirmed by crystallographic structures and will discuss the potential usage of binding energy and the energy of the overall structure to determine a protein’s fitness.

N-methyl-D-aspartate receptors (NMDARs) are central to the pathophysiology of neurodegenerative diseases such as schizophrenia [1], however despite significant structural insights of the receptor [2,3,4,5] the importance of mutations in the NMDAR have been poorly described in the literature. Here we present molecular dynamics simulation data combined with modelling and binding free energy calculations to outline the effects of mutations [6] in the GluN1 subunit of the NMDAR on agonist binding affinity and ligand-receptor interactions. Our data demonstrates the changes caused by the positioning of an introduced tyrosine residue at the binding pocket and its associated changes in the conformation upon ligand binding. Furthermore, molecular dynamics simulations demonstrate the changes in ligand environment in the ligand-receptor complex leading to a loss of key interactions and an associated instability of the bound complex. Lastly, binding free energy calculations show that it is no longer energetically favourable for ionic interactions to form and an associated overall increase in Gibbs free energy for ligand binding. These data are important in explaining the changes in behaviour for mutations in the GluN1 ligand binding region and are consistent with previously reported experiments [7]. We are also pursuing experimental approaches to further understand the action of ligand binding.

[1] Coyle JT (2012). NMDA receptor and schizophrenia: a brief history. Schizophr Bull 38: 920-926.

[2] Amin JB, Gochman A, He M, Certain N, & Wollmuth LP (2021). NMDA Receptors Require Multiple Pre-opening Gating Steps for Efficient Synaptic Activity. Neuron 109: 488-501.e484.

[3] Yu A, & Lau AY (2018). Glutamate and Glycine Binding to the NMDA Receptor. Structure (London, England : 1993) 26: 1035-1043.e1032.

[4] Tajima N, Karakas E, Grant T, Simorowski N, Diaz-Avalos R, Grigorieff N, et al. (2016). Activation of NMDA receptors and the mechanism of inhibition by ifenprodil. Nature 534: 63-68.

[5] Furukawa H, & Gouaux E (2003). Mechanisms of activation, inhibition and specificity: crystal structures of the NMDA receptor NR1 ligand-binding core. The EMBO journal 22: 2873-2885.

[6] Zehavi Y, Mandel H, Zehavi A, Rashid MA, Straussberg R, Jabur B, et al. (2017). De novo GRIN1 mutations: An emerging cause of severe early infantile encephalopathy. Eur J Med Genet 60: 317-320.

[7] Skrenkova K, Song J-m, Kortus S, Kolcheva M, Netolicky J, Hemelikova K, et al. (2020). The pathogenic S688Y mutation in the ligand-binding domain of the GluN1 subunit regulates the properties of NMDA receptors. Sci Rep 10: 18576.

Nucleobases form base pairs, and the question of what are the main driving forces behind the base pair formation is a prevalent one. In our approach to tackle this question we used data from Cambridge Structural Database – we searched for base pair types found in small molecule crystal structures. Obtained base pair types, their frequencies of occurrence and protonation patterns lets us analyze the tendencies of nucleobases to form base pairs.

The reason of why such tendencies occurred varied by the nucleobase, but some general trend were present. The protonation patterns followed the pK_a values of particular nucleobases and the hydrogen bond lengths did not depend on the charge of nucleobases. We compared our findings with analogous data stored in the RNA Basepair Catalog – we found base pairs exclusive for small molecule crystal structures, exclusive for RNA crystal structures and these which were present in both environments. Basing on the frequencies of base pairs, we proven that the pairs often occurring in crystals of small molecules also often occur in RNA crystals [1].

Many base pairs can form higher order structures (Fig. 1) – either some larger aggregates, ribbons, tetramers or whole layers. We wanted to see if there are any rules that govern over the tendency to form particular higher order structure. Is it more of a question of strength of hydrogen bonds (when hydrogen bonds incorporating O or N atoms are favored over these with C), overall interaction energy, or is it rather determined by a simple geometry.

[1] Cabaj, M. K., Dominiak, P. M. (2020). NAR, 48, 8302.

Identifying the probable positions of the protein side-chains is one of the protein modelling steps that can greatly improve the prediction of protein-ligand, protein-protein interactions. With some exceptions, most of the strategies predicting the side-chain conformations use predetermined angles, also called rotamer libraries, that are usually generated from the subset of high-quality protein structures. Although, these libraries are very useful when selecting possible side-chain atom positions, the overall validity and usability with regard to specific protein structure should be studied further.

In order to get well-rounded rotamer library, there should be the balance between the coverage and the quantity of possible side-chain positions. The lack of possible side-chain rotamers will hinder the correct selection of atom positions and the over abundance – the fast selection for protein structure predictions.

We are suggesting the approach that would cover both the coverage and the accuracy of the rotamer library. The rotag software was developed in order to accommodate both these problems. It scans side-chain conformations using dead-end elimination strategy and evaluating potential energies on each calculation step.

The additional challenge that we faced was to have proper method to compare rotamer libraries. The best-case RMSD, best-case dihedral angles and average rotamer choice parameters were selected as good candidates for the comparisons. Multiple rotamer libraries were compared: Dunbrack, Dynameomics, Penultimate and those generated with rotag.

The comparisons revealed that the rotamer libraries that were created from the subset of existing protein structures sometimes lack rotamer positions for certain side-chains of the target proteins. Using more flexible methods, such as rotag, increases the probability of the inclusion of correct conformations. However, not in all cases these flexible methods produce the correct subset of potential candidates.

Many macromolecular data sets suffer from being more or less incomplete mainly as a result of experimental difficulties. Often axes are missing which should be inspected for systematic absences. Even data sets considered complete usually miss very low resolution reflections due to beam stop issues. Low resolution reflections are important as they to a large extent define the protein/solvent boundary. Seriously incomplete data sets can hamper many crystallographic calculations.

A direct space constraint on intensities is the positivity of the Patterson map. All intensities should also make statistical sense and conform to distributions based on observed intensities. Statistics include histograms of normalized and full intensities and scaling as a function of resolution or intensity. Here it is demonstrated how Patterson positivity combined with statistical matching can estimate unobserved intensities. Among applications are space group determination from observation of systematic absences on missing axes and classical rotation functions for molecular replacement. Also any ab initio phasing procedure based on intensities alone is expected to benefit from a complete data set.

The calculations towards a complete data set consist of flipping negative values in the Patterson map followed by histogram match and scaling of the back transformed intensities on order to conform with observed intensities. The generated intensities for observed reflections are in turn flipped relative to the true observations while the unobserved reflections are kept as is. The procedure is initiated by fitting the observed intensities as a function of resolution and determine F000 using knowledge of the solvent content. Initially fitted intensities are substituted in for unobserved data. The calculations are iterative gradually reducing the flipping factor in direct as well as reciprocal space. For cross validation a free data set with 5 % of the observed intensities are kept aside and treated as unobserved.

Figure 1. Structure factor amplitudes of Aspergillus aculeatus rhamnogalacturonan acetylesterase in space group P2₁2₁2₁
Left: 95.45 % complete data set, 0kl-section. Right: Data set completed by Patterson positivity and statistical matching. Systematic absences on missing l-axis are clearly visible.

As a further test of the procedure a virtually complete data set is subjected to various omissions. Omissions include: Low resolution cut off (beam stop issues), high resolution cut off (detector misplaced too far), thin shells of resolution (ice rings), the 10 % strongest intensities (overloads) and increasing omissions around axes and planes in reciprocal space.

In the order of minutes a completed data set can be produced with estimates of unobserved intensities along with estimated standard deviations based on how well the free intensities are reproduced by following the ideas of Read (1986) [1].

[1] Read, R. (1986). Acta Cryst. A42, 140-149

The NPC1 (Niemann-Pick type C1) is one of the main players of cholesterol control in the lysosome and almost its action is closed combined with NPC2 (Niemann-Pick type C2) protein. The dysfunction of one of the proteins can cause problems in overal chloesterol homestasis and leads to a disease, called the Niemann-Pick type C (NPC) disease. It has been reported that many mutations are responsible to the disease. The point mutation R518W or R518Q on the NPC1 is one of such examples. Even though many details on the cholesterol transport mechanism of NPC1 is elucidated especially with the full-length NPC1 structure obtained from cryo-EM study, it is not obvious how the simple mutation can leads such a big difference in proper function of NPC1. In this respect, the single mutation mentioned above could be a good candidate to relate the dynamical function of NPC1 to its structure in cholesterol transport.

In this presentation, we report how the corresponding mutation can induces the structural change in NPC1 by molecular dynamics simulations. Detailed analysis of the resulting simulation trajectory reveals important structural features that is essential for proper function of the NPC1 for cholesterol transport. It has been found that the mutation leads to structural change that is required for proper interaction with NPC2.

The current study can provides some insights into how the structure is closed related to the function of NPC1 in cholesterol transport in terms of its interaction with NPC2 protein.

References

[1] Saha, P., et al. (2020) Inter-domain dynamics drive cholesterol transport by NPC1 and NPC1L1 proteins. Elife 9, e57089.

[2] Dubey, V., et al. (2020) Cholesterol binding to the sterol-sensing region of Niemann Pick C1 protein confines dynamics of its N-terminal domain. PLoS Comput Biol 16, e1007554.

The increasing number of protein structures and the increasing size of protein structures calls for automated methods. The Protein Topology Graph Library (PTGL) [1, 2] models the topology of protein structures as graphs. PTGL supports three levels of abstraction: amino acids, secondary structure elements (SSEs) and chains. For each level of abstraction, the vertices correspond to the level, i.e., vertices correspond to amino acids on amino acid-level and so on. On all abstraction levels, edges denote spatial neighborhoods (contacts). Contacts are based on the computation of Euclidean atom-atom distances. On SSE-level, vertices are labeled as helix or strand. Edges are labeled by the orientation of their SSEs as parallel, antiparallel or mixed. On chain-level, edges are weighted with the number of residue-residue contacts (see Fig 1.).

We used chain-level Complex Graphs (CGs) as a highly abstracted and meaningful view on respiratory complex I. We compared the CGs of the core subunits of complex I between T. thermophilus and H. sapiens. The Complex Graphs shared 29 edges. Each CG had one edge that the other has not. Therefore, the CGs were able to capture the topology of structurally conserved regions.

We applied hierarchical clustering to the edges of a CG. We compared the resulting dendrogram with an assembly process proposed in the literature [4]. Solely based on the CG, we found similarities between the dendrogram and the proposed assembly process. We also applied graph clustering to investigate whether complex I’s modules could be extracted solely from the CG. We showed that CGs could identify modules and guide the finding of the assembly process for complexes.

Concluding, PTGL provides graphs modeling the topology of protein structures on different levels of abstraction for 151.837 PDB structures, including 921 large structures. The webserver provides a search for predefined motifs and user-defined arbitrary patterns. The computation is automated and the implementation publicly available. The representation of graphs enables the application of graph-theoretic methods, such as graph partitioning. This allows feasible analyses on the rapidly growing PDB.

[1] Wolf, J.N., Keßler, M., Ackermann, J. & Koch, I. (2020). Bioinformatics. 37(7), 1032-1034.

[2] May, P., Kreuschwig, A., Steinke, T. & Koch, I. (2009). Nucleic Acids Res. 38, D326-D330.

[3] Baradaran, R., Berrisford, J.M., Minhas, G.S. & Sazanov, L.A. (2013). Nature. 494, 443-448.

[4] Guerrero-Castillo, S., Baertling, F., Kownatzki, D., Wessels, H.J., Arnold, S., Brandt, U. & Nijtmans, L. (2017). Cell Metabolism. 25(1), 128-139.

SARS-CoV-2 (Severe Acute Respiratory Syndrome-Corona Virus 2) spike protein which is the viral protein that causes human cell infections by binding to host cell receptor ACE2 and initiates the membrane fusion. After the entry process, the S-protein needs to be called up and activated by the furin and TMPRSS2 which are the cellular proteases, which stimulates the virus entry into the human cells. By inhibiting the furin protease leads to suppress the spike protein activation in the host cell. The present study aims to understand the intermolecular interactions and binding affinity of furin with its inhibitors decanoyl-RVKR-chloromethylketone (CMK) and Naphthofluorescein which are reported experimentally. The molecular docking studies show the binding affinity of two inhibitors with furin; docking scores for CMK and Naphthofluorescein are -9.727 and -6.036 kcal/mol respectively. The docked complexes of both inhibitors form key interactions with furin and exhibits high docking scores. Further, the molecular dynamics (MD) simulation for both complexes have been performed to understand their stability, shows both inhibitors are stable in the active site region of furin. The RMSD and RMSF plots retrieved from the MD results confirm that CMK molecule having high stability on compare with the Naphthofluorescein. The investigation on furin inhibitors help to evaluate these drugs to be used as a repurposed drug for the SARS-CoV-2. The detailed study will be presented.

Proteins and nucleic acids evolved in the aqueous environment, and water is therefore deeply interrelated with both biomolecular structure and function. The first layer of water molecules around the biomolecular surface - the hydration shell - has properties different from the bulk water [1]. The dynamics of these water molecules is significantly reduced, and the shell mostly consists of ordered (localized) water molecules. However, the first shell water molecules do not have an ice-like structural properties. These ordered water molecules play significant role in recognition and binding of ligands.

In our work, we utilize crystallographic data to compile the average hydration patterns around biomolecules. Firstly, we investigated hydration of DNA building blocks [2, 3], and later hydration of amino acids in proteins as a function of their rotameric state and the secondary structure [4, 5]. Recently, we analyzed hydration of DNA dinucleotides as a function of their conformation and sequence [6]. Here, we present the overview of these results as well as the methodology we used to obtain the data and the potential application of the results.

REFERENCES
[1] Biedermannová L. & Schneider B.: Hydration of proteins and nucleic acids: Advances in experiment and theory. A review. BBA - Gen. 1860: 1821-1835 (2016).
[2] Schneider B. & Berman H.M.: Hydration of the DNA Bases Is Local. Biophys. J. 69: 2661-2669 (1995).
[3] Schneider B., Patel K. & Berman H.M.: Hydration of the Phosphate Group in Double-Helical DNA. Biophys. J. 75: 2422-2434 (1998).
[4] Biedermannová L. & Schneider B.: Structure of the ordered hydration of amino acids in proteins: analysis of crystal structures. Acta Cryst. D71: 2192-2202 (2015).
[5] Černý J., Schneider B. & Biedermannová L.: WatAA: Atlas of Protein Hydration. Exploring synergies between data mining and ab initio calculations. PCCP 19, 17094 (2017).
[6] Biedermannová L. et al., to be published (2021).

Lore of chemical biology guides us that drug discovery of protein binding relies on either optimize the active site complexity of lock and key or induced-fit with conformation selection dynamics; yet, the latter that often-coupled protein interior transport dynamics was much harder to study due to its lack of strong interactions in transient states. This study starts to make progress in using in-situ operando X-ray and neutron contrast variation techniques to depict the landscape of protein binding substrate dynamics in solution. We herein demonstrate, for the first time, the 3-D dynamical structures of hydrated CYP450 protein exterior surfaces to interior buried heme site by a distributed connection of channels that direct the reactant in and out. Using CYP450s of prostacyclin synthase (PGIS) and thromboxane synthase (TXAS) as prototypes we have unveiled the unique dynamics of P450 functional channels in/out the haem site, which drive a variety of water molecules motion, water density change and pre-organization toward the heme active site and hence harness the substrate-binding selectivity. The result is able to clarify how these two proteins catalyze the same substrate of prostaglandin H₂ by entirely different regio-chemical-selective pathways.