Molecular Structure Prediction Research Articles

Comprehensive Structural and Functional Characterization of a Seed γ-thionin as a Potent Bioactive Molecule Against Fungal Pathogens and Insect Pests.

Fungi and insect pests ruin stored crop grain, which results in millions of dollars of damage, presenting an ongoing challenge for farmers in addition to diminishing the safety of stored food. A wide-range defensive system against pathogens is needed to reduce or even eliminate the dependence of the crop yield upon the use of pesticides. Plant defensins (γ-thionins) are antimicrobial peptides (AMPs) that are a component of the host defense system. They are known to interact with cell membranes to exhibit antifungal and insecticidal activity. They exhibit a broad range of activities against fungi and insects and are effective at low concentrations. Thionins act on membranes, greatly reducing the development of pathogen resistance. The aim of this study is to investigate a bioactive molecule that acts against fungal pathogens and stored grain insect pests. γ-thionin protein was extracted from Brassica oleracea L. var. capitata f. alba (white cabbage) seed powder in phosphate buffer (100 mM, pH 7.0) and was identified by MALDI-TOF/TOF. The crude extract was subjected to 70% ammonium sulfate saturation followed by gel filtration chromatography. The disc diffusion assay along with a microtiter bioassay was used to determine the antifungal activity of the protein against phytopathogenic fungi. The insecticidal efficacy was evaluated by feeding insect pests with food contaminated with the purified protein. Additionally, an in silico molecular structure prediction study of the protein was performed using Auto Dock Vina for molecular docking of the protein with either fungal membrane moieties or α-amylase from Tenebrio molitor L. MD simulations of protein-ligand complexes were conducted using Schrodinger's Desmond module. γ-Thionin (BoT) was purified from white cabbage seeds and showed 100% homology with thionin (Brassica oleracea L. var. viridis) and 80% homology with defensin-like protein 1 (Raphanus sativus L.), respectively. BoT significantly inhibited the mycelial growth of Aspergillus niger van Tieghem and Aspergillus flavus Link at a concentration of 2 μM. Similarly, 0.12 μM BoT treatment resulted in significant mortality of Tribolium castaneum Herbst and Sitophilus oryzae L. Molecular docking and MD simulation of BoT confirmed the strong binding affinity with fungal membrane moieties (phosphatidylinositol 4,5-bisphosphate and phosphatidic acid), which causes disruption of the cell membrane and leakage of the cellular contents, leading to cell death. BoT blocked the active site of α-amylase, and as a result of the inactivation of this gut enzyme, the digestive systems of insects were disturbed, resulting in their deaths. This study revealed that γ-thionin is a good antifungal and insecticidal agent that could be used as an alternate to fungicides and insecticides.

Analysis of a thermostable lipase from CTG-clade yeast: Molecular expression, characterization and structure prediction

Analysis of a thermostable lipase from CTG-clade yeast: Molecular expression, characterization and structure prediction

Molecular Phylogeny, DNA Barcoding, and ITS2 Secondary Structure Predictions in the Medicinally Important Eryngium Genotypes of East Coast Region of India.

Commercial interest in the culinary herb, Eryngium foetidum L., has increased worldwide due to its typical pungency, similar to coriander or cilantro, with immense pharmaceutical components. The molecular delimitation and taxonomic classification of this lesser-known medicinal plant are restricted to conventional phenotyping and DNA-based marker evaluation, which hinders accurate identification, genetic conservation, and safe utilization. This study focused on species discrimination using DNA sequencing with chloroplast–plastid genes (matK, Kim matK, and rbcL) and the nuclear ITS2 gene in two Eryngium genotypes collected from the east coast region of India. The results revealed that matK discriminated between two genotypes, however, Kim matK, rbcL, and ITS2 identified these genotypes as E. foetidum. The ribosomal nuclear ITS2 region exhibited significant inter- and intra-specific divergence, depicted in the DNA barcodes and the secondary structures derived based on the minimum free energy. Although the efficiency of matK genes is better in species discrimination, ITS2 demonstrated polyphyletic phylogeny, and could be used as a reliable marker for genetic divergence studies understanding the mechanisms of RNA molecules. The results of this study provide insights into the scientific basis of species identification, genetic conservation, and safe utilization of this important medicinal plant species.

HATS5m as an Example of GETAWAY Molecular Descriptor in Assessing the Similarity/Diversity of the Structural Features of 4-Thiazolidinone.

Among the various methods for drug design, the approach using molecular descriptors for quantitative structure–activity relationships (QSAR) bears promise for the prediction of innovative molecular structures with bespoke pharmacological activity. Despite the growing number of successful potential applications, the QSAR models often remain hard to interpret. The difficulty arises from the use of advanced chemometric or machine learning methods on the one hand, and the complexity of molecular descriptors on the other hand. Thus, there is a need to interpret molecular descriptors for identifying the features of molecules crucial for desirable activity. For example, the development of structure–activity modeling of different molecule endpoints confirmed the usefulness of H-GETAWAY (H-GEometry, Topology, and Atom-Weights AssemblY) descriptors in molecular sciences. However, compared with other 3D molecular descriptors, H-GETAWAY interpretation is much more complicated. The present study provides insights into the interpretation of the HATS5m descriptor (H-GETAWAY) concerning the molecular structures of the 4-thiazolidinone derivatives with antitrypanosomal activity. According to the published study, an increase in antitrypanosomal activity is associated with both a decrease and an increase in HATS5m (leverage-weighted autocorrelation with lag 5, weighted by atomic masses) values. The substructure-based method explored how the changes in molecular features affect the HATS5m value. Based on this approach, we proposed substituents that translate into low and high HATS5m. The detailed interpretation of H-GETAWAY descriptors requires the consideration of three elements: weighting scheme, leverages, and the Dirac delta function. Particular attention should be paid to the impact of chemical compounds’ size and shape and the leverage values of individual atoms.

Reliable crystal structure predictions from first principles

An inexpensive and reliable method for molecular crystal structure predictions (CSPs) has been developed. The new CSP protocol starts from a two-dimensional graph of crystal’s monomer(s) and utilizes no experimental information. Using results of quantum mechanical calculations for molecular dimers, an accurate two-body, rigid-monomer ab initio-based force field (aiFF) for the crystal is developed. Since CSPs with aiFFs are essentially as expensive as with empirical FFs, tens of thousands of plausible polymorphs generated by the crystal packing procedures can be optimized. Here we show the robustness of this protocol which found the experimental crystal within the 20 most stable predicted polymorphs for each of the 15 investigated molecules. The ranking was further refined by performing periodic density-functional theory (DFT) plus dispersion correction (pDFT+D) calculations for these 20 top-ranked polymorphs, resulting in the experimental crystal ranked as number one for all the systems studied (and the second polymorph, if known, ranked in the top few). Alternatively, the polymorphs generated can be used to improve aiFFs, which also leads to rank one predictions. The proposed CSP protocol should result in aiFFs replacing empirical FFs in CSP research.

Characterization of the novel HLA-A*24:191 allele and analysis of its MHC molecular modeling structure

To characterize a novel HLA allele, A*24:191, its DNA sequence, MHC modeling structure, and the possible influence of the amino-acid residue variations on the molecule. The HLA sequence was determined by Luminex PCR-SSO and PCR-SBT. Its MHC molecular structure and the possible effects of the amino-acid residue variations were modeled and analyzed with Phyre2, RCSB PDB and HistoCheck software. The PCR-SBT revealed the novel A*24:191 differs from A*24:02 in exon 2 at position 256, 265, 270 with G>C, G>C, A>T. The MHC molecular structure prediction showed that, compared with A*24:02, the 62nd residue of A*24:191 changed from the acidic E to a neutral Q, both with the side chain extending outside the α helix pointing forward the groove, (Risler's score, R=2), the 65th changed from the smaller neutral G extending inside the helix to a basic R with a long-chain extending upward outside the helix (R=52), and the 66th changed from the basic K to a neutral N both with a long side chain extending inside the groove (R=31). The above residues are located on the α helix of the α 1 domain which constituting the side wall of the peptide-binding groove. The DSS Score=3.85. From the surface image of the molecule, it can be clearly seen that the variations of the properties, sizes and configurations of the residues caused significant changes in the shape of the surface structure of the α helix. It suggested that the residue variations are likely to change the peptide binding properties as well as the TCR and antibody binding characteristics of the molecule.

DFT Quantum-chemical prediction of molecular structure of iron(VI) macrocyclic complex with phthalocyanine and two oxo ligands

The possibility of the existence of the unknown iron complex containing phthalocyanine and two oxygen atoms in the inner coordination sphere was predicted using DFT quantum chemical calculation (OPBE/TZVP and B3PW91/TZVP levels). The structural parameters, the ground state multiplicity, the NBO analysis results, and the standard thermodynamic parameters for complex formation (standard enthalpy [Formula: see text], entropy [Formula: see text] and Gibbs’s energy [Formula: see text] are represented.

Barriers to predictive high-throughput screening for spin-crossover

Current spin-crossover (SCO) energy calculations depend on nearly artisanal skill in picking quantum mechanical approximations and computational methods. That is incompatible with automated (work-flow-driven) screening. An acceptable methodology must be quantum mechanically sound, yield both basic structure and property values, and accurate ΔEHL without steering or tuning. Cost vs. accuracy causes focus on density functional theory (DFT). We show by a near-exhaustive study of schemes for calculating the basic molecular high-low spin energy difference, ΔEHL, that presently there is no combination of a constraint-based, non-empirical density functional approximation (DFA) and a set of well-defined semi-empirical corrections to it adequate for such a protocol. Somewhat successful hybrid DFA calculations of ΔEHL are too costly for high-throughput screening of condensed phases. Lower-cost alternatives combine a generalized gradient approximation (GGA) DFA with a Hubbard-U correction (DFT + U). But we show that neither U=0 nor any currently available unsteered U calculation gives a decent ΔEHL value for [Mn(taa)] without also degrading molecular structure or property predictions. Moving to the SCAN meta-GGA does not solve the problem. The revised-restored SCAN (r2SCAN) meta-GGA together with its deorbitalized version r2SCAN-L give improved but not wholly satisfactory results. We also document and diagnose several non-obvious technical and procedural sensitivities and inter-code differences. In addition to being a formidable challenge to DFA development, the lack we delineate is a major impediment to progress in the development of quantum materials and spintronics.

Synthesis, characterization and relaxivity validations of Gd(III) complex of DOTA tetrahydrazide as MRI contrast agent

We report a novel Gadolinium(III) complex of (1,4,7,10-tetraazacyclododecane-1,4,7,10-tetrayl)tetra(acetohydrazide) (Gd(III)L1) as a potential contrast agent for high resolution images in MRI and its molecular structure prediction by DFT calculations. The DOTA tetrahydrazide ligand framework (L1) was synthesized by the reaction of 1,4,7,10-tetraazacyclododecane (cyclen) with bromo derivative of conjugated t-butyl carbazate and followed by the subsequent cleavage of Boc moieties. The corresponding Gd(III) complex was prepared in an aqueous medium with the adjustment of pH to 7.0. The Gd(III)L1 complex shows the longitudinal relaxivity of 4.9 mM−1s−1 and transverse relaxivity of 6.2 mM−1s−1 at 37 ⁰C. The as-prepared complex Gd(III)L1 displays the enhanced intensity of T1 weighted phantom MR images with various concentrations at 7.0 T MRI scanner, which pave the way to achieve higher resolution images in MRI.

MRI-based radiomics and molecular classification of hepatocellular carcinoma — is there a correlation between them?

Introduction: Hepatocellular carcinoma (HCC) is the most common primary malignant liver tumor, usually developing in the context of chronic liver disease, most commonly associated with hepatitis B (HBV) or C (HCV) infection, alcohol use, or metabolic syndrome. Hepatocellular carcinoma is a very heterogeneous cancer, both at the histological and molecular level. According to recent scientific studies based on gene expression profiling, regardless of the nomenclature used by different authors, HCC can be divided into 2 main subgroups: the non-proliferative type and the proliferative type.Purpose. A review and analysis of the available foreign literature on the textural analysis of magnetic resonance imaging in noninvasive prediction of molecular classification and molecular structure of hepatocellular carcinoma.Material and methods. We searched for scientific publications and clinical guidelines in the PubMed information and analytical system for 2016–2021 using the keywords: «MRI», «radiomics», «texture analysis», «radiogenomics», «HCC», «proliferative», «nonproliferative», «molecular». After excluding studies devoted to technical aspects and the description of individual clinical observations, 16 articles were selected for analyzes.Results. The presented review has demonstrated the broad possibilities and prospects of the use of texture analysis of magnetic resonance imaging in the study of hepatocellular cancer, including the first results in the study of molecular characteristics (signatures) of this tumor. Correlation of textural characteristics with the expression of genes of immunotherapeutic targets CTLA-4 and PD-1 was shown, and a correlation was also shown between ring enhancement of a tumor in the arterial phase by MRI with gadoxetic acid (hepatospecific contrast agent) and proliferative HCC. Texture analysis scores were the predominant independent predictor of microvascular invasion, which was the main independent risk factor for postoperative recurrence.Conclusion. Although texture analysis in the study of HCC is an actively developing field, further study of the structure of this tumor at the molecular level is necessary to develop an personalized treatment, supplement or replace tumor biopsy, as well as to develop new prognostic biomarkers in patients with HCC.

Molecule‐Specific Uncertainty Quantification in Quantum Chemical Studies

AbstractSolving the electronic Schrödinger equation for changing nuclear coordinates provides access to the Born‐Oppenheimer potential energy surface. This surface is the key starting point for almost all theoretical studies of chemical processes in electronic ground and excited states (including molecular structure prediction, reaction mechanism elucidation, molecular property calculations, quantum and molecular dynamics). Electronic structure models aim at a sufficiently accurate approximation of this surface. They have therefore become a cornerstone of theoretical and computational chemistry, molecular physics, and materials science. In this work, we elaborate on general features of approximate electronic structure models such as accuracy, efficiency, and general applicability in order to arrive at a perspective for future developments, of which a vanguard has already arrived. Our quintessential proposition is that meaningful quantum mechanical predictions for chemical phenomena require system‐specific uncertainty information for each and every electronic structure calculation, if objective conclusions shall be drawn with confidence.

Multiscale Modeling of Amyloid Fibrils Formed by Aggregating Peptides Derived from the Amyloidogenic Fragment of the A-Chain of Insulin.

Computational prediction of molecular structures of amyloid fibrils remains an exceedingly challenging task. In this work, we propose a multi-scale modeling procedure for the structure prediction of amyloid fibrils formed by the association of ACC1-13 aggregation-prone peptides derived from the N-terminal region of insulin’s A-chain. First, a large number of protofilament models composed of five copies of interacting ACC1-13 peptides were predicted by application of CABS-dock coarse-grained (CG) docking simulations. Next, the models were reconstructed to all-atom (AA) representations and refined during molecular dynamics (MD) simulations in explicit solvent. The top-scored protofilament models, selected using symmetry criteria, were used for the assembly of long fibril structures. Finally, the amyloid fibril models resulting from the AA MD simulations were compared with atomic force microscopy (AFM) imaging experimental data. The obtained results indicate that the proposed multi-scale modeling procedure is capable of predicting protofilaments with high accuracy and may be applied for structure prediction and analysis of other amyloid fibrils.

Evaluation of cooking performance, structural properties, storage stability and shelf life prediction of high-moisture wet starch noodles

Cooking performance, micro- and molecular structure, storage stability and shelf-life prediction of high-moisture wet starch noodles (SN) were investigated. SEM images revealed that compared to dried SN, cooked wet SN had more evenly honeycomb-like network with smaller size of pores, indicating stronger interaction among molecules and causing favorable cooking performance. XRD and ATR-FTIR results evidenced that wet SN contained more complete crystallites and higher proportion of crystalline region. During storage, the quality decay of wet SN was mainly associated to the increment of total aerobic viable count (TAVC), titrable acidity and amylase, as well as the decreased textural hardness, overall acceptability and lightness. Based on TAVC, titrable acidity and overall acceptability, predicted shelf-life of vacuum-packed wet SN at 25 °C was 15.31, 21.54 and 16.65 weeks respectively, with relative error all within 20%, proving that the validated model could be an effective tool for monitoring the shelf-life of wet SN.

Selectivity and potency of natural product PIM kinase inhibitors identified by in silico docking

PIM3 (Proviral Integration site for Maloney murine leukemia virus kinase 3) is a proto-oncogene with serine/threonine kinase activity that prevents apoptosis, promotes cell survival, and stimulates protein translation. In addition, PIM3 functions in inflammation and immunity pathways. PIM3 inhibitors are being developed to treat cancer and inflammation-related disorders. Here we screen a 98,000 compound virtual library of natural products to identify those that are predicted to fit in the ATP site of PIM3. Since the structure of PIM3 has not been determined experimentally, we performed molecular structure prediction using the SWISS-MODEL tool to generate a PIM3 model structure for in silico screening. Compounds predicted to fit the ATP binding site of PIM3 were validated using biochemical assays, revealing activity against PIM3 for all eight candidates, with potencies mostly in the micromolar range. We tested several analogs of two validated candidates, the diosgenin glycoside dioscin and the biflavonoid hinokiflavone. Among five dioscin analogs, three exhibit similar potency against PIM3, and with some selectivity for PIM3 versus PIM1 and 2. Meanwhile, three of seven biflavonoid analogs exhibit sub-micromolar IC50 potency against PIM3, but with less selectivity for PIM3 versus PIM1 and 2.

Spin-Coupled Generalized Valence Bond Theory: New Perspectives on the Electronic Structure of Molecules and Chemical Bonds

Spin-Coupled Generalized Valence Bond (SCGVB) theory provides the foundation for a comprehensive theory of the electronic structure of molecules. SCGVB theory offers a compelling orbital description of the electronic structure of molecules as well as an efficient and effective zero-order wave function for calculations striving for quantitative predictions of molecular structures, energetics, and other properties. The orbitals in the SCGVB wave function are usually semilocalized, and for most molecules, they can be interpreted using concepts familiar to all chemists (hybrid orbitals, localized bond pairs, lone pairs, etc.). SCGVB theory also provides new perspectives on the nature of the bonds in molecules such as C2, Be2 and SF4/SF6. SCGVB theory contributes unparalleled insights into the underlying cause of the first-row anomaly in inorganic chemistry as well as the electronic structure of organic molecules and the electronic mechanisms of organic reactions. The SCGVB wave function accounts for nondynamical correlation effects and, thus, corrects the most serious deficiency in molecular orbital (RHF) wave functions. Dynamical correlation effects, which are critical for quantitative predictions, can be taken into account using the SCGVB wave function as the zero-order wave function for multireference configuration interaction or coupled cluster calculations.

Advances of bioelectrocatalysis by oxidoreductases

<p indent=0mm>Bioelectrocatalysis attracts much attention with the advantages of high selectivity, high efficiency, no byproduct and environmental friendliness. Electrification of industrial biotechnology is challenging for the different operating conditions and catalysts of electrochemical and biochemical reactions. In this review, we present the state-of-the-art information on enzymatic electrosynthesis for the production of fine chemicals. Multidisciplinary strategies of electrode surface modification and enzyme immobilization have been integrated in electrocatalytic devices with various applications in biosensors and bioelectrocatalysis. However, the efficiency of bioelectrosynthesis is determined by the intramolecular electron transfer rate, and interfacial electron transfer rate between the oxidoreductases and electrode. This review aims to summarize recent progresses on the mechanism studies of intramolecular electron transferof redox proteins, including tunneling, hopping, proton-coupled electron transfer, etc. Modular origin of electron transfer chain was also discussed from the perspective of the evolution of spatial adjacency network. Novel and efficient oxidoreductases with enhanced intramolecular and interfacial electron transfer rate can be obtained by protein engineering. The strategies for facilitating intramolecular electron transfer are addressed, which includes the regulation of co-evolved electron transfer network, the designs of molecular switch and the assembly of conductive modules. Oxidoreductases are engineered to improve their biocatalytic performance by using the tools of molecular evolution, modeling, structure prediction, and mutation. Co-evolved molecular switches controll proton-coupled electron transfer and regulate electron transfer inside the multi-center redox proteins. Assembly of surface-binding conductive peptides to oxidoreductases facilitates electricity-driven catalysis. Moreover, the modifications of oxidoreductases allow their predictable immobilization on functionalized electrode surface with improved stability, controlled orientation, surface coverage and enhanced electron transfer. After that, electron transfer within a series of well-defined peptides with orientation-controlled and surface-confined enzymes was addressed. Strategies also developed to increase the biocompatible active surface of electrodes and to promote charge transfer reactions. Nano- and macroporous electrodes based on metal nanoparticles, nanocarbon tubes, graphene, and metallic inverse opals have been designed and fabricated with predictable surface functionalities and electrochemical properties. Electrode/mediator/cofactor/enzyme conjugates can enhance the <italic>in vitro</italic> bioelectricity generation of cofactor. Screening of redox mediators for electrochemical regeneration of NADH was also summarized. Multifunctional surfaces of nano or meso-porous electrodes, where oxidoreductases were bounded to structured electrodes with necessary components (e.g., mediators, cofactor, etc.) with tailored properties, were summarized. Optimizations of functional electrode materials and surfaces can improve the efficiency of cofactor regeneration in the electrochemical reactor modules. Key challenges and future research for better bioelectrocatalysis are discussed and expected briefly. Mining and engineering of novel and robust enzymes from genomic and metagenomic libraries would benefit the bioelectrocatalysis. Based on genome mining, discovery of novel oxidoreductases from extreme microorganism provide thermophilic, alkalophilic, halophilic and organic solvent tolerant oxidoreductases. Reactors for bioelectrocatalysis are optimized to provide a platform for the production of chiral chemicals, which brings great promises for biomanufacturing. Further study of bio-inspired multi-enzyme immobilization which mimics quick electron transfer in multi-enzyme complex is suggested. A variety of approaches for bioelectrocatalysis have been successfully applied for the synthesis of chiral chemicals which are intermediates of fine chemicals and pharmaceuticals. Bioelectrosynthesis processes could replace many chemically catalyzed routes in the future, and aims to construct a new platform for more efficient and green biomanufacturing.

MLIMC: Machine Learning-Based Implicit-Solvent Monte Carlo.

Monte Carlo (MC) methods are important computational tools for molecular structure optimizations and predictions. When solvent effects are explicitly considered, MC methods become very expensive due to the large degree of freedom associated with the water molecules and mobile ions. Alternatively implicit-solvent MC can largely reduce the computational cost by applying a mean field approximation to solvent effects and meanwhile maintains the atomic detail of the target molecule. The two most popular implicit-solvent models are the Poisson-Boltzmann (PB) model and the Generalized Born (GB) model in a way such that the GB model is an approximation to the PB model but is much faster in simulation time. In this work, we develop a machine learning-based implicit-solvent Monte Carlo (MLIMC) method by combining the advantages of both implicit solvent models in accuracy and efficiency. Specifically, the MLIMC method uses a fast and accurate PB-based machine learning (PBML) scheme to compute the electrostatic solvation free energy at each step. We validate our MLIMC method by using a benzene-water system and a protein-water system. We show that the proposed MLIMC method has great advantages in speed and accuracy for molecular structure optimization and prediction.

Advances in Antimicrobial Peptides with an approach to molecular structure prediction

Antimicrobial peptides (AMP) are molecules involved in the innate immune system of almost all living organisms. They are potent agents with diverse structural and antimicrobial properties, which represent one of the most promising future drug candidate for combating infections and antimicrobial drug resistance. They have also a wide spectrum of activity against a large number of pathogenic microorganisms, multiple mechanisms of action and a low potential for resistance. Numerous AMPs have been isolated from natural sources and many others have been de novo designed and synthetically produced. This review provide insight into antimicrobials properties of AMPs, focusing in antibacterial molecules, that provides better understanding of versatile biological properties of such peptides. Also, provide a short insight into chemical modifications, characterization and the principal methods for protein structure prediction of AMPs.

Reactivity of arsenoplatin complex versus water and thiocyanate: a DFT benchmark study

Seven different density functionals, including GGAs, meta-GGAs, hybrids and range-separated hybrids, and considering Grimme’s empirical dispersion correction (M06-L, M06-2X, PBE0, B3LYP, B3LYP-D3, CAM-B3LYP, ωB97X) have been tested for their performance in the prediction of molecular structures, energies and energy barriers for a class of newly developed antitumor platinum complexes involving main group heavy elements such as arsenic. The calculated structural parameters, energies and energy barriers have been compared to the available experimental data. The results show that range-separated hybrid functionals CAM-B3LYP and ωB97X give good results in predicting both geometrical parameters and isomerization energies and barrier heights and are promising new tools for the theoretical study of novel platinum(II) arsenic compounds.

Drug discovery with explainable artificial intelligence

Deep learning bears promise for drug discovery, including advanced image analysis, prediction of molecular structure and function, and automated generation of innovative chemical entities with bespoke properties. Despite the growing number of successful prospective applications, the underlying mathematical models often remain elusive to interpretation by the human mind. There is a demand for ‘explainable’ deep learning methods to address the need for a new narrative of the machine language of the molecular sciences. This Review summarizes the most prominent algorithmic concepts of explainable artificial intelligence, and forecasts future opportunities, potential applications as well as several remaining challenges. We also hope it encourages additional efforts towards the development and acceptance of explainable artificial intelligence techniques.