MedusaScore: an accurate force field-based scoring function for virtual drug screening

Computational insights into the aggregation mechanism and amyloidogenic core of aortic amyloid medin polypeptide

Medin amyloid, prevalent in the vessel walls of 97 % of individuals over 50, contributes to arterial stiffening and cerebrovascular dysfunction, yet our understanding of its aggregation mechanism remains limited. Dividing the full-length 50-amino-acid medin peptide into five 10-residue segments, we conducted individual investigations on each segment's self-assembly dynamics via microsecond-timescale atomistic discrete molecular dynamics (DMD) simulations. Our findings showed that medin1-10 and medin11-20 segments predominantly existed as isolated unstructured monomers, unable to form stable oligomers. Medin31-40 exhibited moderate aggregation, forming dynamic β-sheet oligomers with frequent association and dissociation. Conversely, medin21-30 and medin41-50 segments demonstrated significant self-assembly capability, readily forming stable β-sheet-rich oligomers. Residue pairwise contact frequency analysis highlighted the critical roles of residues 22-26 and 43-49 in driving the self-assembly of medin21-30 and medin41-50, acting as the β-sheet core and facilitating β-strand formation in other regions within medin monomers, expecting to extend to oligomers and fibrils. Regions containing residues 22-26 and 43-49, with substantial self-assembly abilities and assistance in β-sheet formation, represent crucial targets for amyloid inhibitor drug design against aortic medial amyloidosis (AMA). In summary, our study not only offers deep insights into the mechanism of medin amyloid formation but also provides crucial theoretical and practical guidance for future treatments of AMA.

Identification of a Novel Protein Phosphatase 2A Activator, PPA24, as a Potential Therapeutic for FOLFOX-Resistant Colorectal Cancer

A series of compounds were designed utilizing molecular modeling and fragment-based design based upon the known protein phosphatase 2A (PP2A) activators, NSC49L and iHAP1, and evaluated for their ability to inhibit the viability of colorectal cancer (CRC) and folinic acid, 5-fluorouracil, and oxaliplatin (FOLFOX)-resistant CRC cells. PPA24 (19a) was identified as the most cytotoxic compound with IC50 values in the range of 2.36-6.75 μM in CRC and FOLFOX-resistant CRC cell lines. It stimulated PP2A activity to a greater extent, displayed lower binding energies through molecular docking, and showed higher binding affinity through surface plasmon resonance for PP2A catalytic subunit α than the known PP2A activators. PPA24 dose-dependently induced apoptosis and oxidative stress, decreased the level of c-Myc expression, and synergistically potentiated cytotoxicity when combined with gemcitabine and cisplatin. Furthermore, a PPA24-encapsulated nanoformulation significantly inhibited the growth of CRC xenografts without systemic toxicities. Together, these results signify the potential of PPA24 as a novel PP2A activator and a prospective therapeutic for CRC and FOLFOX-resistant CRC.

Uncovering Intermolecular Interactions Driving the Liquid-Liquid Phase Separation of the TDP-43 Low-Complexity Domain via Atomistic Dimerization Simulations

Liquid-liquid phase separation (LLPS) of transactive response DNA-binding protein of 43 kDa (TDP-43), which exerts multiple functions in the splicing, trafficking, and stabilization of RNA, mediates the formation of membraneless condensates with crucial physiological roles, while its aberrant LLPS is linked to multiple neurodegenerative diseases. However, due to the heterogeneous and dynamic nature of LLPS, major gaps remain in understanding the precise intermolecular interactions driving LLPS and how specific mutations alter LLPS dynamics. Here, we investigated the molecular mechanisms underlying the LLPS of the TDP-43 low-complexity domain (LCD) by simulating the dimerization process using all-atom discrete molecular dynamics with microsecond-long simulations. Our results showed that the TDP-43 LCD was intrinsically disordered, with helical structures consistent with prior nuclear magnetic resonance studies. Phase separation propensity was assessed by simulating the dimerization of the TDP-43 LCD and four mutants, showing that A321G, W334G, and M337V inhibited self-association, while G335D promoted it, fully consistent with experimental reports. During the dimerization process, two peptides experienced both elastic and nonelastic collisions, and the self-associated dimer featured both high- and low-contact states. These results suggested that the dimerization process of the TDP-43 LCD was accordingly dynamic and heterogeneous. Additionally, we identified crucial regions containing hydrophobic clusters and aromatic residues in the N-terminus, central region, and C-terminus that were essential for the self-association of the TDP-43 LCD. These residues with high binding affinities can act as stickers to form peptide networks in LLPS. Together, our simulation provides a comprehensive picture of the intermolecular interactions driving the phase separation of the TDP-43 LCD, offering insights into both physiological functions and pathological mechanisms.

@TOME 3.0: Interfacing Protein Structure Modeling and Ligand Docking

Knowledge of protein-ligand complexes is essential for efficient drug design. Virtual docking can bring important information on putative complexes but it is still far from being simultaneously fast and accurate. Receptors are flexible and adapt to the incoming small molecules while docking is highly sensitive to small conformational deviations. Conformation ensemble is providing a mean to simulate protein flexibility. However, modeling multiple protein structures for many targets is seldom connected to ligand screening in an efficient and straightforward manner. @TOME-3 is an updated version of our former pipeline @TOME-2, in which protein structure modeling is now directly interfaced with flexible ligand docking. Sequence-sequence profile comparisons identify suitable PDB templates for structure modeling and ligands from these templates are used to deduce binding sites to be screened. In addition, bound ligand can be used as pharmacophoric restraint during the virtual docking. The latter is performed by PLANTS while the docking poses are analysed through multiple chemoinformatics functions. This unique combination of tools allows rapid and efficient ligand docking on multiple receptor conformations in parallel. @TOME-3 is freely available on the web at https://atome.cbs.cnrs.fr.

Computational Investigation of Coaggregation and Cross-Seeding between Aβ and hIAPP Underpinning the Cross-Talk in Alzheimer's Disease and Type 2 Diabetes

The coexistence of amyloid-β (Aβ) and human islet amyloid polypeptide (hIAPP) in the brain and pancreas is associated with an increased risk of Alzheimer's disease (AD) and type 2 diabetes (T2D) due to their coaggregation and cross-seeding. Despite this, the molecular mechanisms underlying their interaction remain elusive. Here, we systematically investigated the cross-talk between Aβ and hIAPP using atomistic discrete molecular dynamics (DMD) simulations. Our results revealed that the amyloidogenic core regions of both Aβ (Aβ10-21 and Aβ30-41) and hIAPP (hIAPP8-20 and hIAPP22-29), driving their self-aggregation, also exhibited a strong tendency for cross-interaction. This propensity led to the formation of β-sheet-rich heterocomplexes, including potentially toxic β-barrel oligomers. The formation of Aβ and hIAPP heteroaggregates did not impede the recruitment of additional peptides to grow into larger aggregates. Our cross-seeding simulations demonstrated that both Aβ and hIAPP fibrils could mutually act as seeds, assisting each other's monomers in converting into β-sheets at the exposed fibril elongation ends. The amyloidogenic core regions of Aβ and hIAPP, in both oligomeric and fibrillar states, exhibited the ability to recruit isolated peptides, thereby extending the β-sheet edges, with limited sensitivity to the amino acid sequence. These findings suggest that targeting these regions by capping them with amyloid-resistant peptide drugs may hold potential as a therapeutic approach for addressing AD, T2D, and their copathologies.

Predicting Drug-Target Interactions Via Dual-Stream Graph Neural Network

Drug target interaction prediction is a crucial stage in drug discovery. However, brute-force search over a compound database is financially infeasible. We have witnessed the increasing measured drug-target interactions records in recent years, and the rich drug/protein-related information allows the usage of graph machine learning. Despite the advances in deep learning-enabled drug-target interaction, there are still open challenges: (1) rich and complex relationship between drugs and proteins can be explored; (2) the intermediate node is not calibrated in the heterogeneous graph. To tackle with above issues, this paper proposed a framework named DSG-DTI. Specifically, DSG-DTI has the heterogeneous graph autoencoder and heterogeneous attention network-based Matrix Completion. Our framework ensures that the known types of nodes (e.g., drug, target, side effects, diseases) are precisely embedded into high-dimensional space with our pretraining skills. Also, the attention-based heterogeneous graph-based matrix completion achieves highly competitive results via effective long-range dependencies extraction. We verify our model on two public benchmarks. The result of two publicly available benchmark application programs show that the proposed scheme effectively predicts drug-target interactions and can generalize to newly registered drugs and targets with slight performance degradation, outperforming the best accuracy compared with other baselines.

Molecular Insights into the Effects of F16L and F19L Substitutions on the Conformation and Aggregation Dynamics of Human Calcitonin

Human calcitonin (hCT) regulates calcium-phosphorus metabolism, but its amyloid aggregation disrupts physiological activity, increases thyroid carcinoma risk, and hampers its clinical use for bone-related diseases like osteoporosis and Paget's disease. Improving hCT with targeted modifications to mitigate amyloid formation while maintaining its function holds promise as a strategy. Understanding how each residue in hCT's amyloidogenic core affects its structure and aggregation dynamics is crucial for designing effective analogues. Mutants F16L-hCT and F19L-hCT, where Phe residues in the core are replaced with Leu as in nonamyloidogenic salmon calcitonin, showed different aggregation kinetics. However, the molecular effects of these substitutions in hCT are still unclear. Here, we systematically investigated the folding and self-assembly conformational dynamics of hCT, F16L-hCT, and F19L-hCT through multiple long-time scale independent atomistic discrete molecular dynamics (DMD) simulations. Our results indicated that the hCT monomer primarily assumed unstructured conformations with dynamic helices around residues 4-12 and 14-21. During self-assembly, the amyloidogenic core of hCT14-21 converted from dynamic helices to β-sheets. However, substituting F16L did not induce significant conformational changes, as F16L-hCT exhibited characteristics similar to those of wild-type hCT in both monomeric and oligomeric states. In contrast, F19L-hCT exhibited substantially more helices and fewer β-sheets than did hCT, irrespective of their monomers or oligomers. The substitution of F19L significantly enhanced the stability of the helical conformation for hCT14-21, thereby suppressing the helix-to-β-sheet conformational conversion. Overall, our findings elucidate the molecular mechanisms underlying hCT aggregation and the effects of F16L and F19L substitutions on the conformational dynamics of hCT, highlighting the critical role of F19 as an important target in the design of amyloid-resistant hCT analogs for future clinical applications.

Enhancing Generalizability in Protein-Ligand Binding Affinity Prediction with Multimodal Contrastive Learning

Improving the generalization ability of scoring functions remains a major challenge in protein-ligand binding affinity prediction. Many machine learning methods are limited by their reliance on single-modal representations, hindering a comprehensive understanding of protein-ligand interactions. We introduce a graph-neural-network-based scoring function that utilizes a triplet contrastive learning loss to improve protein-ligand representations. In this model, three-dimensional complex representations and the fusion of two-dimensional ligand and coarse-grained pocket representations converge while distancing from decoy representations in latent space. After rigorous validation on multiple external data sets, our model exhibits commendable generalization capabilities compared to those of other deep learning-based scoring functions, marking it as a promising tool in the realm of drug discovery. In the future, our training framework can be extended to other biophysical- and biochemical-related problems such as protein-protein interaction and protein mutation prediction.

Drugging the entire human proteome: Are we there yet?

Each of the ∼20,000 proteins in the human proteome is a potential target for compounds that bind to it and modify its function. The 3D structures of most of these proteins are now available. Here, we discuss the prospects for using these structures to perform proteome-wide virtual HTS (VHTS). We compare physics-based (docking) and AI VHTS approaches, some of which are now being applied with large databases of compounds to thousands of targets. Although preliminary proteome-wide screens are now within our grasp, further methodological developments are expected to improve the accuracy of the results.

Computational insights into the cross-talk between medin and Aβ: implications for age-related vascular risk factors in Alzheimer's disease

The aggregation of medin forming aortic medial amyloid is linked to arterial wall degeneration and cerebrovascular dysfunction. Elevated levels of arteriolar medin are correlated with an increased presence of vascular amyloid-β (Aβ) aggregates, a hallmark of Alzheimer's disease (AD) and vascular dementia. The cross-interaction between medin and Aβ results in the formation of heterologous fibrils through co-aggregation and cross-seeding processes both in vitro and in vivo. However, a comprehensive molecular understanding of the cross-interaction between medin and Aβ-two intrinsically disordered proteins-is critically lacking. Here, we employed atomistic discrete molecular dynamics simulations to systematically investigate the self-association, co-aggregation and also the phenomenon of cross-seeding between these two proteins. Our results demonstrated that both Aβ and medin were aggregation prone and their mixture tended to form β-sheet-rich hetero-aggregates. The formation of Aβ-medin hetero-aggregates did not hinder Aβ and medin from recruiting additional Aβ and medin peptides to grow into larger β-sheet-rich aggregates. The β-barrel oligomer intermediates observed in the self-aggregations of Aβ and medin were also present during their co-aggregation. In cross-seeding simulations, preformed Aβ fibrils could recruit isolated medin monomers to form elongated β-sheets. Overall, our comprehensive simulations suggested that the cross-interaction between Aβ and medin may contribute to their pathological aggregation, given the inherent amyloidogenic tendencies of both medin and Aβ. Targeting medin, therefore, could offer a novel therapeutic approach to preserving brain function during aging and AD by improving vascular health.

Deciphering the influence of Y12L and N17H substitutions on the conformation and oligomerization of human calcitonin

The abnormal aggregation of human calcitonin (hCT) hormone peptides impairs their physiological function, leading to harmful immune responses and cytotoxicity, which limits their clinical utility. Interestingly, a representative hCT analog incorporating Y12L and N17H substitutions (DM-hCT) has shown reduced aggregation tendencies while maintaining bioactivity. But the molecular mechanism of Y12L and N17H substitutions on the conformational dynamics of hCT remains unclear. Here, we systematically investigated the folding and self-assembly dynamics of hCT and DM-hCT using atomistic discrete molecular dynamics (DMD) simulations. Our findings revealed that hCT monomers predominantly adopted unstructured conformations with dynamic helices. Oligomerization of hCT resulted in the formation of β-sheet-rich aggregates and β-barrel intermediates. The Y12L and N17H substitutions enhanced helical conformations and suppressed β-sheet formation in both monomers and oligomers. These substitutions stabilized the dynamic helices and disrupted aromatic interactions responsible for β-sheet formation at residue 12. Notably, DM-hCT assemblies still exhibited β-sheets in phenylalanine-rich and C-terminal hydrophobic regions, suggesting that future optimizations should focus on these areas. Our simulations provide insights into the molecular mechanisms underlying hCT aggregation and the amyloid-resistant effects of Y12L and N17H substitutions. These findings have valuable implications for the development of clinical hCT analogs.

Structure-based, deep-learning models for protein-ligand binding affinity prediction

The launch of AlphaFold series has brought deep-learning techniques into the molecular structural science. As another crucial problem, structure-based prediction of protein-ligand binding affinity urgently calls for advanced computational techniques. Is deep learning ready to decode this problem? Here we review mainstream structure-based, deep-learning approaches for this problem, focusing on molecular representations, learning architectures and model interpretability. A model taxonomy has been generated. To compensate for the lack of valid comparisons among those models, we realized and evaluated representatives from a uniform basis, with the advantages and shortcomings discussed. This review will potentially benefit structure-based drug discovery and related areas.

Integration of Computational Docking into Anti-Cancer Drug Response Prediction Models

Cancer is a heterogeneous disease in that tumors of the same histology type can respond differently to a treatment. Anti-cancer drug response prediction is of paramount importance for both drug development and patient treatment design. Although various computational methods and data have been used to develop drug response prediction models, it remains a challenging problem due to the complexities of cancer mechanisms and cancer-drug interactions. To better characterize the interaction between cancer and drugs, we investigate the feasibility of integrating computationally derived features of molecular mechanisms of action into prediction models. Specifically, we add docking scores of drug molecules and target proteins in combination with cancer gene expressions and molecular drug descriptors for building response models. The results demonstrate a marginal improvement in drug response prediction performance when adding docking scores as additional features, through tests on large drug screening data. We discuss the limitations of the current approach and provide the research community with a baseline dataset of the large-scale computational docking for anti-cancer drugs.

Co-aggregation of α-synuclein with amyloid-β stabilizes β-sheet-rich oligomers and enhances the formation of β-barrels

Alzheimer's disease (AD) and Parkinson's disease (PD) are the two most common neurodegenerative diseases with markedly different pathological features of β-amyloid (Aβ) plaques and α-synuclein (αS) Lewy bodies (LBs), respectively. However, clinical overlaps in symptoms and pathologies between AD and PD are commonly observed caused by the cross-interaction between Aβ and αS. To uncover the molecular mechanisms behind their overlapping symptoms and pathologies, we computationally investigated the impact of αS on an Aβ monomer and dimerization using atomistic discrete molecular dynamics simulations (DMD). Our results revealed that αS could directly interact with Aβ monomers and dimers, thus forming β-sheet-rich oligomers, including potentially toxic β-barrel intermediates. The binding hotspot involved the second half of the N-terminal domain and NAC region in αS, along with residues 10-21 and 31-42 in Aβ. In their hetero-complex, the binding hotspot primarily assumed a β-sheet core buried inside, which was dynamically shielded by the highly charged, amyloid-resistant C-terminus of αS. Because the amyloid prion region was the same as the binding hotspot being buried, their fibrillization may be delayed, causing the toxic oligomers to increase. This study sheds light on the intricate relationship between Aβ and αS and provides insights into the overlapping pathology of AD and PD.

Unveiling Medin Folding and Dimerization Dynamics and Conformations via Atomistic Discrete Molecular Dynamics Simulations

Medin is a principal component of localized amyloid found in the vasculature of individuals over 50 years old. Its amyloid aggregation has been linked to endothelial dysfunction and vascular inflammation, contributing to the pathogenesis of various vascular diseases. Despite its significance, the structures of the medin monomer, oligomer, and fibril remain elusive, and the dynamic processes of medin aggregation are not fully understood. In this study, we comprehensively investigated the medin folding and dimerization dynamics and conformations using atomistic discrete molecular dynamics simulations. Our simulation results suggested that the folding initiation of the medin involved the formation of β-sheets around medin30-41 and medin42-50, with subsequent capping of other segments to their β-sheet edges. Medin monomers typically consisted of three or four β-strands, along with a dynamic N-terminal helix. Two isolated medin peptides readily aggregated into a β-sheet-rich dimer, displaying a strong aggregation propensity. Dimerization of medin not only enhanced the β-sheet conformations but also led to the formation of β-barrel oligomers. The aggregation tendencies of medin1-18 and medin19-29 were relatively weak. However, the segments of medin30-41 and medin42-50 played a crucial role as they primarily formed a β-sheet core and facilitated medin1-18 and medin19-29 to form intra- and interpeptide β-sheets. The findings highlight the critical role of the medin30-41 and medin42-50 regions in stabilizing the monomer structure and driving the medin amyloid aggregation. These regions could potentially serve as promising targets for designing antiamyloid inhibitors against amyloid aggregation of medin. Additionally, our study provides a full picture of the monomer conformations and dimerization dynamics for medin, which will help better understand the pathology of medin aggregation.

Septins Enable T Cell Contact Guidance via Amoeboid-Mesenchymal Switch

Lymphocytes exit circulation and enter in-tissue guided migration toward sites of tissue pathologies, damage, infection, or inflammation. By continuously sensing and adapting to the guiding chemo-mechano-structural properties of the tissues, lymphocytes dynamically alternate and combine their amoeboid (non-adhesive) and mesenchymal (adhesive) migration modes. However, which mechanisms guide and balance different migration modes are largely unclear. Here we report that suppression of septins GTPase activity induces an abrupt amoeboid-to-mesenchymal transition of T cell migration mode, characterized by a distinct, highly deformable integrin-dependent immune cell contact guidance. Surprisingly, the T cell actomyosin cortex contractility becomes diminished, dispensable and antagonistic to mesenchymal-like migration mode. Instead, mesenchymal-like T cells rely on microtubule stabilization and their non-canonical dynein motor activity for high fidelity contact guidance. Our results establish septin's GTPase activity as an important on/off switch for integrin-dependent migration of T lymphocytes, enabling their dynein-driven fluid-like mesenchymal propulsion along the complex adhesion cues.

SEVI Inhibits Aβ Amyloid Aggregation by Capping the β-Sheet Elongation Edges

Inhibiting the aggregation of amyloid peptides with endogenous peptides has broad interest due to their intrinsically high biocompatibility and low immunogenicity. Here, we investigated the inhibition mechanism of the prostatic acidic phosphatase fragment SEVI (semen-derived enhancer of viral infection) against Aβ42 fibrillization using atomistic discrete molecular dynamic simulations. Our result revealed that SEVI was intrinsically disordered with dynamic formation of residual helices. With a high positive net charge, the self-aggregation tendency of SEVI was weak. Aβ42 had a strong aggregation propensity by readily self-assembling into β-sheet-rich aggregates. SEVI preferred to interact with Aβ42, rather than SEVI themselves. In the heteroaggregates, Aβ42 mainly adopted β-sheets buried inside and capped by SEVI in the outer layer. SEVI could bind to various Aβ aggregation species─including monomers, dimers, and proto-fibrils─by capping the exposed β-sheet elongation edges. The aggregation processes Aβ42 from the formation of oligomers to conformational nucleation into fibrils and fibril growth should be inhibited as their β-sheet elongation edges are being occupied by the highly charged SEVI. Overall, our computational study uncovered the molecular mechanism of experimentally observed inhibition of SEVI against Aβ42 aggregation, providing novel insights into the development of therapeutic strategies against Alzheimer's disease.

Dissecting the Self-assembly Dynamics of Imperfect Repeats in α-Synuclein

The pathological aggregation of α-synuclein (αS) into amyloid fibrils is the hallmark of Parkinson's disease (PD). The self-assembly and membrane interactions of αS are mainly governed by the seven imperfect 11-residue repeats of the XKTKEGVXXXX motif around residues 1-95. However, the particular role of each repeat in αS fibrillization remains unclear. To answer this question, we studied the aggregation dynamics of each repeat with up to 10 peptides in silico by conducting multiple independent micro-second atomistic discrete molecular dynamics simulations. Our simulations revealed that only repeats R3 and R6 readily self-assembled into β-sheet-rich oligomers, while the other repeats remained as unstructured monomers with weak self-assembly and β-sheet propensities. The self-assembly process of R3 featured frequent conformational changes with β-sheet formation mainly in the non-conserved hydrophobic tail, whereas R6 spontaneously self-assembled into extended and stable cross-β structures. These results of seven repeats are consistent with their structures and organization in recently solved αS fibrils. As the primary amyloidogenic core, R6 was buried inside the central cross-β core of all αS fibrils, attracting the hydrophobic tails of adjacent R4, R5, and R7 repeats forming β-sheets around R6 in the core. Further away from R6 in the sequence but with a moderate amyloid aggregation propensity, the R3 tail could serve as a secondary amyloidogenic core and form independent β-sheets in the fibril. Overall, our results demonstrate the critical role of R3 and R6 repeats in αS amyloid aggregation and suggest their potential as targets for the peptide-based and small-molecule amyloid inhibitors.

emPDBA: protein-DNA binding affinity prediction by combining features from binding partners and interface learned with ensemble regression model

Protein-deoxyribonucleic acid (DNA) interactions are important in a variety of biological processes. Accurately predicting protein-DNA binding affinity has been one of the most attractive and challenging issues in computational biology. However, the existing approaches still have much room for improvement. In this work, we propose an ensemble model for Protein-DNA Binding Affinity prediction (emPDBA), which combines six base models with one meta-model. The complexes are classified into four types based on the DNA structure (double-stranded or other forms) and the percentage of interface residues. For each type, emPDBA is trained with the sequence-based, structure-based and energy features from binding partners and complex structures. Through feature selection by the sequential forward selection method, it is found that there do exist considerable differences in the key factors contributing to intermolecular binding affinity. The complex classification is beneficial for the important feature extraction for binding affinity prediction. The performance comparison of our method with other peer ones on the independent testing dataset shows that emPDBA outperforms the state-of-the-art methods with the Pearson correlation coefficient of 0.53 and the mean absolute error of 1.11 kcal/mol. The comprehensive results demonstrate that our method has a good performance for protein-DNA binding affinity prediction. Availability and implementation: The source code is available at https://github.com/ChunhuaLiLab/emPDBA/.

Structural and dynamic insights into α-synuclein dimer conformations

Parkinson disease is associated with the aggregation of the protein α-synuclein. While α-synuclein can exist in multiple oligomeric states, the dimer has been a subject of extensive debates. Here, using an array of biophysical approaches, we demonstrate that α-synuclein in vitro exhibits primarily a monomer-dimer equilibrium in nanomolar concentrations and up to a few micromolars. We then use spatial information from hetero-isotopic cross-linking mass spectrometry experiments as restrains in discrete molecular dynamics simulations to obtain the ensemble structure of dimeric species. Out of eight structural sub-populations of dimers, we identify one that is compact, stable, abundant, and exhibits partially exposed β-sheet structures. This compact dimer is the only one where the hydroxyls of tyrosine 39 are in proximity that may promote dityrosine covalent linkage upon hydroxyl radicalization, which is implicated in α-synuclein amyloid fibrils. We propose that this α-synuclein dimer features etiological relevance to Parkinson disease.

Application of computational methods for class A GPCR Ligand discovery

G protein-coupled receptors (GPCR) are integral membrane proteins of considerable interest as targets for drug development due to their role in transmitting cellular signals in a multitude of biological processes. Of the six classes categorizing GPCR (A, B, C, D, E, and F), class A contains the largest number of therapeutically relevant GPCR. Despite their importance as drug targets, many challenges exist for the discovery of novel class A GPCR ligands serving as drug precursors. Though knowledge of the structural and functional characteristics of GPCR has grown significantly over the past 20 years, a large portion of GPCR lack reported, experimentally determined structures. Furthermore, many GPCR have no known endogenous and/or synthetic ligands, limiting further exploration of their biochemical, cellular, and physiological roles. While many successes in GPCR ligand discovery have resulted from experimental high-throughput screening, computational methods have played an increasingly important role in GPCR ligand identification in the past decade. Here we discuss computational techniques applied to GPCR ligand discovery. This review summarizes class A GPCR structure/function and provides an overview of many obstacles currently faced in GPCR ligand discovery. Furthermore, we discuss applications and recent successes of computational techniques used to predict GPCR structure as well as present a summary of ligand- and structure-based methods used to identify potential GPCR ligands. Finally, we discuss computational hit list generation and refinement and provide comprehensive workflows for GPCR ligand identification.

Machine Learning Scoring Functions for Drug Discovery from Experimental and Computer-Generated Protein-Ligand Structures: Towards Per-Target Scoring Functions

In recent years, machine learning has been proposed as a promising strategy to build accurate scoring functions for computational docking finalized to numerically empowered drug discovery. However, the latest studies have suggested that over-optimistic results had been reported due to the correlations present in the experimental databases used for training and testing. Here, we investigate the performance of an artificial neural network in binding affinity predictions, comparing results obtained using both experimental protein-ligand structures as well as larger sets of computer-generated structures created using commercial software. Interestingly, similar performances are obtained on both databases. We find a noticeable performance suppression when moving from random horizontal tests to vertical tests performed on target proteins not included in the training data. The possibility to train the network on relatively easily created computer-generated databases leads us to explore per-target scoring functions, trained and tested ad-hoc on complexes including only one target protein. Encouraging results are obtained, depending on the type of protein being addressed.

Computer-Aided Drug Design towards New Psychotropic and Neurological Drugs

Central nervous system (CNS) disorders are a therapeutic area in drug discovery where demand for new treatments greatly exceeds approved treatment options. This is complicated by the high failure rate in late-stage clinical trials, resulting in exorbitant costs associated with bringing new CNS drugs to market. Computer-aided drug design (CADD) techniques minimise the time and cost burdens associated with drug research and development by ensuring an advantageous starting point for pre-clinical and clinical assessments. The key elements of CADD are divided into ligand-based and structure-based methods. Ligand-based methods encompass techniques including pharmacophore modelling and quantitative structure activity relationships (QSARs), which use the relationship between biological activity and chemical structure to ascertain suitable lead molecules. In contrast, structure-based methods use information about the binding site architecture from an established protein structure to select suitable molecules for further investigation. In recent years, deep learning techniques have been applied in drug design and present an exciting addition to CADD workflows. Despite the difficulties associated with CNS drug discovery, advances towards new pharmaceutical treatments continue to be made, and CADD has supported these findings. This review explores various CADD techniques and discusses applications in CNS drug discovery from 2018 to November 2022.

Hierarchical analysis of the target-based scoring function modification for the example of selected class A GPCRs

Computational methods, especially molecular docking-based calculations, have become indispensable in the modern drug discovery workflow. The constantly increasing chemical space requires fast, robust but most of all highly predictive methods to search for new bioactive agents. Thus, the scoring function (SF) is a useful and broadly applied energy-based element of docking software, allowing quick and effective evaluation of a ligand's propensity to bind to selected protein targets. Despite many spectacular successes of molecular docking applications in virtual screening (VS), the obtained results are often far from ideal, leading to incorrect selection of hit molecules and poor pose prediction. In our study we focused on docking calculation for the selected class A G-protein coupled receptors (GPCRs), with experimentally determined 3D structures and a sufficient set of known ligands with affinity values reported in the ChEMBL database. Our goal is to investigate how much the energy-based scoring function for this particular target class changes when changing from the default to the re-estimated weighting scheme on the specified energy terms in the SF definition. Additionally, we want to verify if indeed more accurate results are obtained when considering different levels of the biological hierarchy, namely: the whole class A GPCRs, sub-subfamilies, or just the individual proteins while applying default or specifically designed weighting coefficients. The performed calculation and evaluation factor values suggest a significant improvement of docking results for the designed SF definition. This individual approach improves the accuracy of binding affinity prediction and active compound recognition. The designed scoring function for classes, sub-subfamilies, or proteins leads to a significant improvement of molecular docking performance, especially at the level of individual proteins. Our results show that to increase the efficiency and predictive power of molecular docking calculations applied in classical VS, the strategy based on the individual approach for scoring function definition for selected proteins should be considered.

Structural Perturbation of Monomers Determines the Amyloid Aggregation Propensity of Calcitonin Variants

Human calcitonin (hCT) is a polypeptide hormone that participates in calcium-phosphorus metabolism. Irreversible aggregation of 32-amino acid hCT into β-sheet-rich amyloid fibrils impairs physiological activity and increases the risk of medullary carcinoma of the thyroid. Amyloid-resistant hCT derivatives substituting critical amyloidogenic residues are of particular interest for clinical applications as therapeutic drugs against bone-related diseases. Uncovering the aggregation mechanism of hCT at the molecular level, therefore, is important for the design of amyloid-resistant hCT analogues. Here, we investigated the aggregation dynamics of hCT, non-amyloidogenic salmon calcitonin (sCT), and two hCT analogues with reduced aggregation tendency─TL-hCT and phCT─using long timescale discrete molecular dynamics simulations. Our results showed that hCT monomers mainly adopted unstructured conformations with dynamically formed helices around the central region. hCT self-assembled into helix-rich oligomers first, followed by a conformational conversion into β-sheet-rich oligomers with β-sheets formed by residues 10-30 and stabilized by aromatic and hydrophobic interactions. Our simulations confirmed that TL-hCT and phCT oligomers featured more helices and fewer β-sheets than hCT. Substitution of central aromatic residues with leucine in TL-hCT and replacing C-terminal hydrophobic residue with hydrophilic amino acid in phCT only locally suppressed β-sheet propensities in the central region and C-terminus, respectively. Having mutations in both central and C-terminal regions, sCT monomers and dynamically formed oligomers predominantly adopted helices, confirming that both central aromatic and C-terminal hydrophobic residues played important roles in the fibrillization of hCT. We also observed the formation of β-barrel intermediates, postulated as the toxic oligomers in amyloidosis, for hCT but not for sCT. Our computational study depicts a complete picture of the aggregation dynamics of hCT and the effects of mutations. The design of next-generation amyloid-resistant hCT analogues should consider the impact on both amyloidogenic regions and also take into account the amplification of transient β-sheet population in monomers upon aggregation.

Insights into small molecule inhibitor bindings to PD-L1 with residue-specific binding free energy calculation

Targeting the immunological checkpoint PD-1/PD-L1 with antibodies has shown opportunities to improve cancer treatment in recent years. However, antibody therapy is a double-edged sword with high cost, low patient tolerance, lack of oral bioavailability, and a reaction to most solid tumors that prevents the adoption of antibodies. Advancement of small-molecule PD-1/PD-L1 inhibitors that could overwhelm these drawbacks is sluggish because of the poor pharmacodynamic properties and shallow pocket of the PD-1/PD-L1 binding interface. Recently, a number of compounds have been discovered to bind the PD-L1/PD-L1 dimer interface, providing an excellent alternative to inhibit the interaction between PD-1/PD-L1 and small molecules. Quantitative characterization of PD-L1 interactions with these inhibitors will advance the design of novel and efficient inhibitors in the future. Here, the binding free energies of 35 PD-L1 dimer inhibitors have been calculated using the alanine-scanning-interaction-entropy (AS-IE) method. Hotspot residues on PD-L1 and potential modification groups on the inhibitors were identified. The experimental results for the AS-IE method were better correlated than the classical MM/GBSA method. These results may set the stage for the design the more powerful PD-L1 inhibitors.Communicated by Ramaswamy H. Sarma.

Molecular Insights into the Misfolding and Dimerization Dynamics of the Full-Length α-Synuclein from Atomistic Discrete Molecular Dynamics Simulations

The misfolding and pathological aggregation of α-synuclein forming insoluble amyloid deposits is associated with Parkinson's disease, the second most common neurodegenerative disease in the world population. Characterizing the self-assembly mechanism of α-synuclein is critical for discovering treatments against synucleinopathies. The intrinsically disordered property, high degrees of freedom, and macroscopic timescales of conformational conversion make its characterization extremely challenging in vitro and in silico. Here, we systematically investigated the dynamics of monomer misfolding and dimerization of the full-length α-synuclein using atomistic discrete molecular dynamics simulations. Our results suggested that both α-synuclein monomers and dimers mainly adopted unstructured formations with partial helices around the N-terminus (residues 8-32) and various β-sheets spanning the residues 35-56 (N-terminal tail) and residues 61-95 (NAC region). The C-terminus mostly assumed an unstructured formation wrapping around the lateral surface and the elongation edge of the β-sheet core formed by an N-terminal tail and NAC regions. Dimerization enhanced the β-sheet formation along with a decrease in the unstructured content. The inter-peptide β-sheets were mainly formed by the N-terminal tail and NACore (residues 68-78) regions, suggesting that these two regions played critical roles in the amyloid aggregation of α-synuclein. Interactions of the C-terminus with the N-terminal tail and the NAC region were significantly suppressed in the α-synuclein dimer, indicating that the interaction of the C-terminus with the N-terminal tail and NAC regions could prevent α-synuclein aggregation. These results on the structural ensembles and early aggregation dynamics of α-synuclein will help understand the nucleation and fibrillization of α-synuclein.

Computational methods for calculation of protein-ligand binding affinities in structure-based drug design

Drug design is an expensive and time-consuming process. Any method that allows reducing the time the costs of the drug development project can have great practical value for the pharmaceutical industry. In structure-based drug design, affinity prediction methods are of great importance. The majority of methods used to predict binding free energy in protein-ligand complexes use molecular mechanics methods. However, many limitations of these methods in describing interactions exist. An attempt to go beyond these limits is the application of quantum-mechanical description for all or only part of the analyzed system. However, the extensive use of quantum mechanical (QM) approaches in drug discovery is still a demanding challenge. This chapter briefly reviews selected methods used to calculate protein-ligand binding affinity applied in virtual screening (VS), rescoring of docked poses, and lead optimization stage, including QM methods based on molecular simulations.

Molecular insights into the oligomerization dynamics and conformations of amyloidogenic and non-amyloidogenic amylin from discrete molecular dynamics simulations

The amyloid aggregation of human islet amyloid polypeptide (hIAPP) is associated with pancreatic β-cell death in type 2 diabetes. The S20G substitution of hIAPP (hIAPP(S20G)), found in Japanese and Chinese people, is more amyloidogenic and cytotoxic than wild-type hIAPP. Rat amylin (rIAPP) does not have aggregation propensity or cytotoxicity. Mounting evidence suggests that soluble low-molecular-weight amyloid oligomers formed during early aggregation are more cytotoxic than mature fibrils. The self-assembly dynamics and oligomeric conformations remain unknown because the oligomers are heterogeneous and transient. The molecular mechanism of sequence-variation rendering dramatically different aggregation propensity and cytotoxicity is also elusive. Here, we investigate the oligomerization dynamics and conformations of amyloidogenic hIAPP, hIAPP(S20G), and non-amyloidogenic rIAPP using atomistic discrete molecular dynamics (DMD) simulations. Our simulation results demonstrated that all three monomeric amylin peptides mainly adopted an unstructured formation with partial dynamical helices near the N-terminus. Relatively transient β-hairpins were more abundant in hIAPP and hIAPP(S20G) than in rIAPP. The S20G-substituting mutant of hIAPP altered the turn region of the β-hairpin motif, resulting in more hydrophobic residue-pairwise contacts within the β-hairpin. Oligomerization dynamic investigation revealed that all three peptides spontaneously accumulated into helix-populated oligomers. The conformational conversion to form β-sheet-rich oligomers was only observed in hIAPP and hIAPP(S20G). The population of high-β-sheet-content oligomers was enhanced by S20G substitution. Interestingly, both hIAPP and hIAPP(S20G) could form β-barrel formations, and the β-barrel propensity of hIAPP(S20G) was three times larger than that of hIAPP. No β-sheet-rich or β-barrel formations were observed in rIAPP. Our direct observation of the correlation between β-barrel oligomer formation and cytotoxicity suggests that β-barrels might play a critically important role in the cytotoxicity of amyloidosis.

Carbon Nanodots from an In Silico Perspective

Carbon nanodots (CNDs) are the latest and most shining rising stars among photoluminescent (PL) nanomaterials. These carbon-based surface-passivated nanostructures compete with other related PL materials, including traditional semiconductor quantum dots and organic dyes, with a long list of benefits and emerging applications. Advantages of CNDs include tunable inherent optical properties and high photostability, rich possibilities for surface functionalization and doping, dispersibility, low toxicity, and viable synthesis (top-down and bottom-up) from organic materials. CNDs can be applied to biomedicine including imaging and sensing, drug-delivery, photodynamic therapy, photocatalysis but also to energy harvesting in solar cells and as LEDs. More applications are reported continuously, making this already a research field of its own. Understanding of the properties of CNDs requires one to go to the levels of electrons, atoms, molecules, and nanostructures at different scales using modern molecular modeling and to correlate it tightly with experiments. This review highlights different in silico techniques and studies, from quantum chemistry to the mesoscale, with particular reference to carbon nanodots, carbonaceous nanoparticles whose structural and photophysical properties are not fully elucidated. The role of experimental investigation is also presented. Hereby, we hope to encourage the reader to investigate CNDs and to apply virtual chemistry to obtain further insights needed to customize these amazing systems for novel prospective applications.

An integrated understanding of the evolutionary and structural features of the SARS-CoV-2 spike receptor binding domain (RBD)

Conventional drug development strategies typically use pocket in protein structures as drug-target sites. They overlook the plausible effects of protein evolvability and resistant mutations on protein structure which in turn may impair protein-drug interaction. In this study, we used an integrated evolution and structure guided strategy to develop potential evolutionary-escape resistant therapeutics using receptor binding domain (RBD) of SARS-CoV-2 spike-protein/S-protein as a model. Deploying an ensemble of sequence space exploratory tools including co-evolutionary analysis and deep mutational scans we provide a quantitative insight into the evolutionarily constrained subspace of the RBD sequence-space. Guided by molecular simulation and structure network analysis we highlight regions inside the RBD, which are critical for providing structural integrity and conformational flexibility. Using fuzzy C-means clustering we combined evolutionary and structural features of RBD and identified a critical region. Subsequently, we used computational drug screening using a library of 1615 small molecules and identified one lead molecule, which is expected to target the identified region, critical for evolvability and structural stability of RBD. This integrated evolution-structure guided strategy to develop evolutionary-escape resistant lead molecules have potential general applications beyond SARS-CoV-2.

Predicting Protein-Ligand Docking Structure with Graph Neural Network

Modern day drug discovery is extremely expensive and time consuming. Although computational approaches help accelerate and decrease the cost of drug discovery, existing computational software packages for docking-based drug discovery suffer from both low accuracy and high latency. A few recent machine learning-based approaches have been proposed for virtual screening by improving the ability to evaluate protein-ligand binding affinity, but such methods rely heavily on conventional docking software to sample docking poses, which results in excessive execution latencies. Here, we propose and evaluate a novel graph neural network (GNN)-based framework, MedusaGraph, which includes both pose-prediction (sampling) and pose-selection (scoring) models. Unlike the previous machine learning-centric studies, MedusaGraph generates the docking poses directly and achieves from 10 to 100 times speedup compared to state-of-the-art approaches, while having a slightly better docking accuracy.

Pose Classification Using Three-Dimensional Atomic Structure-Based Neural Networks Applied to Ion Channel-Ligand Docking

The identification of promising lead compounds showing pharmacological activities toward a biological target is essential in early stage drug discovery. With the recent increase in available small-molecule databases, virtual high-throughput screening using physics-based molecular docking has emerged as an essential tool in assisting fast and cost-efficient lead discovery and optimization. However, the best scored docking poses are often suboptimal, resulting in incorrect screening and chemical property calculation. We address the pose classification problem by leveraging data-driven machine learning approaches to identify correct docking poses from AutoDock Vina and Glide screens. To enable effective classification of docking poses, we present two convolutional neural network approaches: a three-dimensional convolutional neural network (3D-CNN) and an attention-based point cloud network (PCN) trained on the PDBbind refined set. We demonstrate the effectiveness of our proposed classifiers on multiple evaluation data sets including the standard PDBbind CASF-2016 benchmark data set and various compound libraries with structurally different protein targets including an ion channel data set extracted from Protein Data Bank (PDB) and an in-house KCa3.1 inhibitor data set. Our experiments show that excluding false positive docking poses using the proposed classifiers improves virtual high-throughput screening to identify novel molecules against each target protein compared to the initial screen based on the docking scores.

NeuralDock: Rapid and Conformation-Agnostic Docking of Small Molecules

Virtual screening is a cost- and time-effective alternative to traditional high-throughput screening in the drug discovery process. Both virtual screening approaches, structure-based molecular docking and ligand-based cheminformatics, suffer from computational cost, low accuracy, and/or reliance on prior knowledge of a ligand that binds to a given target. Here, we propose a neural network framework, NeuralDock, which accelerates the process of high-quality computational docking by a factor of 106, and does not require prior knowledge of a ligand that binds to a given target. By approximating both protein-small molecule conformational sampling and energy-based scoring, NeuralDock accurately predicts the binding energy, and affinity of a protein-small molecule pair, based on protein pocket 3D structure and small molecule topology. We use NeuralDock and 25 GPUs to dock 937 million molecules from the ZINC database against superoxide dismutase-1 in 21 h, which we validate with physical docking using MedusaDock. Due to its speed and accuracy, NeuralDock may be useful in brute-force virtual screening of massive chemical libraries and training of generative drug models.

Hydrophobic/Hydrophilic Ratio of Amphiphilic Helix Mimetics Determines the Effects on Islet Amyloid Polypeptide Aggregation

Amyloid depositions of human islet amyloid polypeptides (hIAPP) are associated with type II diabetes (T2D) impacting millions of people globally. Accordingly, strategies against hIAPP aggregation are essential for the prevention and eventual treatment of the disease. Helix mimetics, which modulate the protein-protein interaction by mimicking the side chain residues of a natural α-helix, were found to be a promising strategy for inhibiting hIAPP aggregation. Here, we applied molecular dynamics simulations to investigate two helix mimetics reported to have opposite effects on hIAPP aggregation in solution, the oligopyridylamide-based scaffold 1e promoted, whereas naphthalimide-appended oligopyridylamide scaffold DM 1 inhibited the aggregation of hIAPP in solution. We found that 1e promoted hIAPP aggregation because of the recruiting effects through binding with the N-termini of hIAPP peptides. In contrast, DM 1 with a higher hydrophobic/hydrophilic ratio effectively inhibited hIAPP aggregation by strongly binding with the C-termini of hIAPP peptides, which competed for the interpeptide contacts between amyloidogenic regions in the C-termini and impaired the fibrillization of hIAPP. Structural analyses revealed that DM 1 formed the core of hIAPP-DM 1 complexes and stabilized the off-pathway oligomers, whereas 1e formed the corona outside the hIAPP-1e complexes and remained active in recruiting free hIAPP peptides. The distinct interaction mechanisms of DM 1 and 1e, together with other reported potent antagonists in the literature, emphasized the effective small molecule-based amyloid inhibitors by disrupting peptide interactions that should reach a balanced hydrophobic/hydrophilic ratio, providing a viable and generic strategy for the rational design of novel anti-amyloid nanomedicine.

Protein-ligand binding affinity prediction based on profiles of intermolecular contacts

As a key element in structure-based drug design, binding affinity prediction (BAP) for putative protein-ligand complexes can be efficiently achieved by the incorporation of structural descriptors and machine-learning models. However, developing concise descriptors that will lead to accurate and interpretable BAP remains a difficult problem in this field. Herein, we introduce the profiles of intermolecular contacts (IMCPs) as descriptors for machine-learning-based BAP. IMCPs describe each group of protein-ligand contacts by the count and average distance of the group members, and collaborate closely with classical machine-learning models. Performed on multiple validation sets, IMCP-based models often result in better BAP accuracy than those originating from other similar descriptors. Additionally, IMCPs are simple and concise, and easy to interpret in model training. These descriptors highly conclude the structural information of protein-ligand complexes and can be easily updated with personalized profile features. IMCPs have been implemented in the BAP Toolkit on github ( https://github.com/debbydanwang/BAP).

Graphene quantum dots obstruct the membrane axis of Alzheimer's amyloid beta

Alzheimer's disease (AD) is a primary form of dementia with debilitating consequences, but no effective cure is available. While the pathophysiology of AD remains multifactorial, the aggregation of amyloid beta (Aβ) mediated by the cell membrane is known to be the cause for the neurodegeneration associated with AD. Here we examined the effects of graphene quantum dots (GQDs) on the obstruction of the membrane axis of Aβ in its three representative forms of monomers (Aβ-m), oligomers (Aβ-o), and amyloid fibrils (Aβ-f). Specifically, we determined the membrane fluidity of neuroblastoma SH-SY5Y cells perturbed by the Aβ species, especially by the most toxic Aβ-o, and demonstrated their recovery by GQDs using confocal fluorescence microscopy. Our computational data through discrete molecular dynamics simulations further revealed energetically favorable association of the Aβ species with the GQDs in overcoming peptide-peptide aggregation. Overall, this study positively implicated GQDs as an effective agent in breaking down the membrane axis of Aβ, thereby circumventing adverse downstream events and offering a potential therapeutic solution for AD.

Virtual Screening with Gnina 1.0

Virtual screening-predicting which compounds within a specified compound library bind to a target molecule, typically a protein-is a fundamental task in the field of drug discovery. Doing virtual screening well provides tangible practical benefits, including reduced drug development costs, faster time to therapeutic viability, and fewer unforeseen side effects. As with most applied computational tasks, the algorithms currently used to perform virtual screening feature inherent tradeoffs between speed and accuracy. Furthermore, even theoretically rigorous, computationally intensive methods may fail to account for important effects relevant to whether a given compound will ultimately be usable as a drug. Here we investigate the virtual screening performance of the recently released Gnina molecular docking software, which uses deep convolutional networks to score protein-ligand structures. We find, on average, that Gnina outperforms conventional empirical scoring. The default scoring in Gnina outperforms the empirical AutoDock Vina scoring function on 89 of the 117 targets of the DUD-E and LIT-PCBA virtual screening benchmarks with a median 1% early enrichment factor that is more than twice that of Vina. However, we also find that issues of bias linger in these sets, even when not used directly to train models, and this bias obfuscates to what extent machine learning models are achieving their performance through a sophisticated interpretation of molecular interactions versus fitting to non-informative simplistic property distributions.

Mechanical Counterbalance of Kinesin and Dynein Motors in a Microtubular Network Regulates Cell Mechanics, 3D Architecture, and Mechanosensing

Microtubules (MTs) and MT motor proteins form active 3D networks made of unstretchable cables with rod-like bending mechanics that provide cells with a dynamically changing structural scaffold. In this study, we report an antagonistic mechanical balance within the dynein-kinesin microtubular motor system. Dynein activity drives the microtubular network inward compaction, while isolated activity of kinesins bundles and expands MTs into giant circular bands that deform the cell cortex into discoids. Furthermore, we show that dyneins recruit MTs to sites of cell adhesion, increasing the topographic contact guidance of cells, while kinesins antagonize it via retraction of MTs from sites of cell adhesion. Actin-to-microtubule translocation of septin-9 enhances kinesin-MT interactions, outbalances the activity of kinesins over that of dyneins, and induces the discoid architecture of cells. These orthogonal mechanisms of MT network reorganization highlight the existence of an intricate mechanical balance between motor activities of kinesins and dyneins that controls cell 3D architecture, mechanics, and cell-microenvironment interactions.

Using diverse potentials and scoring functions for the development of improved machine-learned models for protein-ligand affinity and docking pose prediction

The advent of computational drug discovery holds the promise of significantly reducing the effort of experimentalists, along with monetary cost. More generally, predicting the binding of small organic molecules to biological macromolecules has far-reaching implications for a range of problems, including metabolomics. However, problems such as predicting the bound structure of a protein-ligand complex along with its affinity have proven to be an enormous challenge. In recent years, machine learning-based methods have proven to be more accurate than older methods, many based on simple linear regression. Nonetheless, there remains room for improvement, as these methods are often trained on a small set of features, with a single functional form for any given physical effect, and often with little mention of the rationale behind choosing one functional form over another. Moreover, it is not entirely clear why one machine learning method is favored over another. In this work, we endeavor to undertake a comprehensive effort towards developing high-accuracy, machine-learned scoring functions, systematically investigating the effects of machine learning method and choice of features, and, when possible, providing insights into the relevant physics using methods that assess feature importance. Here, we show synergism among disparate features, yielding adjusted R² with experimental binding affinities of up to 0.871 on an independent test set and enrichment for native bound structures of up to 0.913. When purely physical terms that model enthalpic and entropic effects are used in the training, we use feature importance assessments to probe the relevant physics and hopefully guide future investigators working on this and other computational chemistry problems.

FFENCODER-PL: Pair Wise Energy Descriptors for Protein-Ligand Pose Selection

Scoring functions are the essential component in molecular docking methods. An accurate scoring function is expected to distinguish the native ligand pose from decoy poses. Our previous experience (Pei et al. J. Chem. Inf. Model. 2019, 59 (7), 3305-3315) proved that combining the random forest (RF) algorithm with knowledge-based potential functions can emphasize germane pair wise interactions and improve the performance of original knowledge-based potential functions on protein-ligand decoy detection. One of the most important potential function classes is the force field (FF) potential with one example being the Amber collection of FFs, which are widely available in the AMBER suite of simulation programs. However, for use in RF modeling studies, one needs pair wise energies that are hard to directly extract from Amber. To address this issue, FFENCODER-PL was constructed to calculate the pair wise energies based on the FF14SB and GAFF2 FFs in Amber. FFENCODER-PL was validated using 275 ligand and 21 protein-ligand structures. RF models were built by combining an RF classification algorithm with the pair wise energies calculated from FFENCODER-PL. CASF-2016 (Su et al. J. Chem. Inf. Model. 2019, 59, 895-913) was employed to test the performance of the resultant RF models, which outperformed 33 scoring functions on accuracy and native ranking tests. For the best decoy RMSD test, RF models give a best decoy with an RMSD of around 2 Å from the native pose after including the best decoy-decoy comparisons in the RF model. The relative importance of the RF algorithm and force field potentials was also tested with the results suggesting that both the RF algorithm and force field potentials are important and combining them is the only way to achieve high accuracy. Finally, FFENCODER-PL makes force field-based pair wise energies available for further development of machine learning-based scoring functions. The codes and data used in this paper can be found at https://github.com/JunPei000/Amber_protein_ligand_encoding.

The structural heterogeneity of α-synuclein is governed by several distinct subpopulations with interconversion times slower than milliseconds

α-Synuclein plays an important role in synaptic functions by interacting with synaptic vesicle membrane, while its oligomers and fibrils are associated with several neurodegenerative diseases. The specific monomer structures that promote its membrane binding and self-association remain elusive due to its transient nature as an intrinsically disordered protein. Here, we use inter-dye distance distributions from bulk time-resolved Förster resonance energy transfer as restraints in discrete molecular dynamics simulations to map the conformational space of the α-synuclein monomer. We further confirm the generated conformational ensemble in orthogonal experiments utilizing far-UV circular dichroism and cross-linking mass spectrometry. Single-molecule protein-induced fluorescence enhancement measurements show that within this conformational ensemble, some of the conformations of α-synuclein are surprisingly stable, exhibiting conformational transitions slower than milliseconds. Our comprehensive analysis of the conformational ensemble reveals essential structural properties and potential conformations that promote its various functions in membrane interaction or oligomer and fibril formation.

Inhibition of Amyloid Aggregation and Toxicity with Janus Iron Oxide Nanoparticles

Amyloid aggregation is a ubiquitous form of protein misfolding underlying the pathologies of Alzheimer's disease (AD), Parkinson's disease (PD) and type 2 diabetes (T2D), three primary forms of human amyloid diseases. While much has been learned about the origin, diagnosis and management of these neurological and metabolic disorders, no cure is currently available due in part to the dynamic and heterogeneous nature of the toxic oligomers induced by amyloid aggregation. Here we synthesized beta casein-coated iron oxide nanoparticles (βCas IONPs) via a BPA-P(OEGA-b-DBM) block copolymer linker. Using a thioflavin T kinetic assay, transmission electron microscopy, Fourier transform infrared spectroscopy, discrete molecular dynamics simulations and cell viability assays, we examined the Janus characteristics and the inhibition potential of βCas IONPs against the aggregation of amyloid beta (Aβ), alpha synuclein (αS) and human islet amyloid polypeptide (IAPP) which are implicated in the pathologies of AD, PD and T2D. Incubation of zebrafish embryos with the amyloid proteins largely inhibited hatching and elicited reactive oxygen species, which were effectively rescued by the inhibitor. Furthermore, Aβ-induced damage to mouse brain was mitigated in vivo with the inhibitor. This study revealed the potential of Janus nanoparticles as a new nanomedicine against a diverse range of amyloid diseases.

Identification and characterization of novel candidate compounds targeting 6- and 7-transmembrane μ-opioid receptor isoforms

BACKGROUND AND PURPOSE

The μ-opioid receptor (μ receptor) is the primary target for opioid analgesics. The 7-transmembrane (TM) and 6TM μ receptor isoforms mediate inhibitory and excitatory cellular effects. Here, we developed compounds selective for 6TM- or 7TM-μ receptors to further our understanding of the pharmacodynamic properties of μ receptors.

EXPERIMENTAL APPROACH

We performed virtual screening of the ZINC Drug Now library of compounds using in silico 7TM- and 6TM-μ receptor structural models and identified potential compounds that are selective for 6TM- and/or 7TM-μ receptors. Subsequently, we characterized the most promising candidate compounds in functional in vitro studies using Be2C neuroblastoma transfected cells, behavioural in vivo pain assays using various knockout mice and in ex vivo electrophysiology studies.

KEY RESULTS

Our virtual screen identified 30 potential candidate compounds. Subsequent functional in vitro cellular assays shortlisted four compounds (#5, 10, 11 and 25) that demonstrated 6TM- or 7TM-μ receptor-dependent NO release. In in vivo pain assays these compounds also produced dose-dependent hyperalgesic responses. Studies using mice that lack specific opioid receptors further established the μ receptor-dependent nature of identified novel ligands. Ex vivo electrophysiological studies on spontaneous excitatory postsynaptic currents in isolated spinal cord slices also validated the hyperalgesic properties of the most potent 6TM- (#10) and 7TM-μ receptor (#5) ligands.

CONCLUSION AND IMPLICATIONS

Our novel compounds represent a new class of ligands for μ receptors and will serve as valuable research tools to facilitate the development of opioids with significant analgesic efficacy and fewer side-effects.

Ultrasmall Molybdenum Disulfide Quantum Dots Cage Alzheimer's Amyloid Beta to Restore Membrane Fluidity

Alzheimer's disease (AD) is a major cause of dementia characterized by the overexpression of transmembrane amyloid precursor protein and its neurotoxic byproduct amyloid beta (Aβ). A small peptide of considerable hydrophobicity, Aβ is aggregation prone catalyzed by the presence of cell membranes, among other environmental factors. Accordingly, current AD mitigation strategies often aim at breaking down the Aβ-membrane communication, yet no data is available concerning the cohesive interplay of the three key entities of the cell membrane, Aβ, and its inhibitor. Using a lipophilic Laurdan dye and confocal fluorescence microscopy, we observed cell membrane perturbation and actin reorganization induced by Aβ oligomers but not by Aβ monomers or amyloid fibrils. We further revealed recovery of membrane fluidity by ultrasmall MoS2 quantum dots, also shown in this study as a potent inhibitor of Aβ amyloid aggregation. Using discrete molecular dynamics simulations, we uncovered the binding of MoS2 and Aβ monomers as mediated by hydrophilic interactions between the quantum dots and the peptide N-terminus. In contrast, Aβ oligomers and fibrils were surface-coated by the ultrasmall quantum dots in distinct testudo-like, reverse protein-corona formations to prevent their further association with the cell membrane and adverse effects downstream. This study offers a crucial new insight and a viable strategy for regulating the amyloid aggregation and membrane-axis of AD pathology with multifunctional nanomedicine.

DeepBSP-a Machine Learning Method for Accurate Prediction of Protein-Ligand Docking Structures

In recent years, machine-learning-based scoring functions have significantly improved the scoring power. However, many of these methods do not perform well in distinguishing the native structure from docked decoy poses due to the lack of decoy structural information in their training data. Here, we developed a machine-learning model, named DeepBSP, that can directly predict the root mean square deviation (rmsd) of a ligand docking pose with reference to its native binding pose. Unlike the binding affinity, the rmsd between the docking poses with reference to their native structures can be straightforwardly determined. By training on a generated data set with 11,925 native complexes and more than 165,000 docked poses, our model shows excellent docking power on our test set and also on the CASF-2016 docking decoy set compared to other major scoring functions. Thus, by combining molecular dockings that generate many poses with the application of DeepBSP, one can more accurately predict the best binding pose that is closest to the native complex structure. This DeepBSP model shall be very useful in picking out poses close to their natives from many poses generated from a dock application.

Ensemble Switching Unveils a Kinetic Rheostat Mechanism of the Eukaryotic Thiamine Pyrophosphate Riboswitch

Thiamine pyrophosphate (TPP) riboswitches regulate thiamine metabolism by inhibiting the translation of enzymes essential to thiamine synthesis pathways upon binding to thiamine pyrophosphate in cells across all domains of life. Recent work on the Arabidopsis thaliana TPP riboswitch suggests a multi-step TPP binding process involving multiple riboswitch configurational ensembles and that Mg2+ dependence underlies the mechanism of TPP recognition and subsequent transition to the expression-inhibiting state of the aptamer domain followed by changes in the expression platform. However, details of the relationship between TPP riboswitch conformational changes and interactions with TPP and Mg2+ ¬¬in the aptamer domain constituting this mechanism are unknown. Therefore, we integrated single-molecule multiparameter fluorescence and force spectroscopy with atomistic molecular dynamics simulations and found that conformational transitions within the aptamer domain's sensor helices associated with TPP and Mg2+ ligand binding occurred between at least five different ensembles on timescales ranging from µs to ms. These dynamics are orders of magnitude faster than the 10 second-timescale folding kinetics associated with expression-state switching in the switch sequence. Together, our results show that a TPP and Mg2+ dependent mechanism determines dynamic configurational state ensemble switching of the aptamer domain's sensor helices that regulates the stability of the switch helix, which ultimately may lead to the expression-inhibiting state of the riboswitch. Additionally, we propose that two pathways exist for ligand recognition and that this mechanism underlies a kinetic rheostat-like behavior of the Arabidopsis thaliana TPP riboswitch.

Modifications to the Framework Regions Eliminate Chimeric Antigen Receptor Tonic Signaling

Chimeric antigen receptor (CAR) tonic signaling, defined as spontaneous activation and release of proinflammatory cytokines by CAR-T cells, is considered a negative attribute because it leads to impaired antitumor effects. Here, we report that CAR tonic signaling is caused by the intrinsic instability of the mAb single-chain variable fragment (scFv) to promote self-aggregation and signaling via the CD3ζ chain incorporated into the CAR construct. This phenomenon was detected in a CAR encoding either CD28 or 4-1BB costimulatory endodomains. Instability of the scFv was caused by specific amino acids within the framework regions (FWR) that can be identified by computational modeling. Substitutions of the amino acids causing instability, or humanization of the FWRs, corrected tonic signaling of the CAR, without modifying antigen specificity, and enhanced the antitumor effects of CAR-T cells. Overall, we demonstrated that tonic signaling of CAR-T cells is determined by the molecular instability of the scFv and that computational analyses of the scFv can be implemented to correct the scFv instability in CAR-T cells with either CD28 or 4-1BB costimulation.

GPU-Accelerated Flexible Molecular Docking

Virtual screening is a key enabler of computational drug discovery and requires accurate and efficient structure-based molecular docking. In this work, we develop algorithms and software building blocks for molecular docking that can take advantage of graphics processing units (GPUs). Specifically, we focus on MedusaDock, a flexible protein-small molecule docking approach and platform. We accelerate the performance of the coarse docking phase of MedusaDock, as this step constitutes nearly 70% of total running time in typical use-cases. We perform a comprehensive evaluation of the quality and performance with single-GPU and multi-GPU acceleration using a data set of 3875 protein-ligand complexes. The algorithmic ideas, data structure design choices, and performance optimization techniques shed light on GPU acceleration of other structure-based molecular docking software tools.