CSAR Benchmark of Flexible MedusaDock in Affinity Prediction and Nativelike Binding Pose Selection

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.

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.

Guiding Conventional Protein-Ligand Docking Software with Convolutional Neural Networks

The high-performance computational techniques have brought significant benefits for drug discovery efforts in recent decades. One of the most challenging problems in drug discovery is the protein-ligand binding pose prediction. To predict the most stable structure of the complex, the performance of conventional structure-based molecular docking methods heavily depends on the accuracy of scoring or energy functions (as an approximation of affinity) for each pose of the protein-ligand docking complex to effectively guide the search in an exponentially large solution space. However, due to the heterogeneity of molecular structures, the existing scoring calculation methods are either tailored to a particular data set or fail to exhibit high accuracy. In this paper, we propose a convolutional neural network (CNN)-based model that learns to predict the stability factor of the protein-ligand complex and exhibits the ability of CNNs to improve the existing docking software. Evaluated results on PDBbind data set indicate that our approach reduces the execution time of the traditional docking-based method while improving the accuracy. Our code, experiment scripts, and pretrained models are available at https://github.com/j9650/MedusaNet.

DUBS: A Framework for Developing Directory of Useful Benchmarking Sets for Virtual Screening

Benchmarking is a crucial step in evaluating virtual screening methods for drug discovery. One major issue that arises among benchmarking data sets is a lack of a standardized format for representing the protein and ligand structures used to benchmark the virtual screening method. To address this, we introduce the Directory of Useful Benchmarking Sets (DUBS) framework, as a simple and flexible tool to rapidly create benchmarking sets using the protein databank. DUBS uses a simple input text based format along with the Lemon data mining framework to efficiently access and organize data to the protein databank and output commonly used inputs for virtual screening software. The simple input format used by DUBS allows users to define their own benchmarking data sets and access the corresponding information directly from the software package. Currently, it only takes DUBS less than 2 min to create a benchmark using this format. Since DUBS uses a simple python script, users can easily modify this to create more complex benchmarks. We hope that DUBS will be a useful community resource to provide a standardized representation for benchmarking data sets in virtual screening. The DUBS package is available on GitHub at https://github.com/chopralab/lemon/tree/master/dubs.

CANDOCK: Chemical Atomic Network-Based Hierarchical Flexible Docking Algorithm Using Generalized Statistical Potentials

Small-molecule docking has proven to be invaluable for drug design and discovery. However, existing docking methods have several limitations such as improper treatment of the interactions of essential components in the chemical environment of the binding pocket (e.g., cofactors, metal ions, etc.), incomplete sampling of chemically relevant ligand conformational space, and the inability to consistently correlate docking scores of the best binding pose with experimental binding affinities. We present CANDOCK, a novel docking algorithm, that utilizes a hierarchical approach to reconstruct ligands from an atomic grid using graph theory and generalized statistical potential functions to sample biologically relevant ligand conformations. Our algorithm accounts for protein flexibility, solvent, metal ions, and cofactor interactions in the binding pocket that are traditionally ignored by current methods. We evaluate the algorithm on the PDBbind, Astex, and PINC proteins to show its ability to reproduce the binding mode of the ligands that is independent of the initial ligand conformation in these benchmarks. Finally, we identify the best selector and ranker potential functions such that the statistical score of the best selected docked pose correlates with the experimental binding affinities of the ligands for any given protein target. Our results indicate that CANDOCK is a generalized flexible docking method that addresses several limitations of current docking methods by considering all interactions in the chemical environment of a binding pocket for correlating the best-docked pose with biological activity. CANDOCK along with all structures and scripts used for benchmarking is available at https://github.com/chopralab/candock_benchmark.

Physical and Toxicological Profiles of Human IAPP Amyloids and Plaques

Although much has been learned about the fibrillization kinetics, structure and toxicity of amyloid proteins, the properties of amyloid fibrils beyond the saturation phase are often perceived as chemically and biologically inert, despite evidence suggesting otherwise. To fill this knowledge gap, we examined the physical and biological characteristics of human islet amyloid polypeptide (IAPP) fibrils that were aged up to two months. Not only did aging decrease the toxicity of IAPP fibrils, but the fibrils also sequestered fresh IAPP and suppressed their toxicity in an embryonic zebrafish model. The mechanical properties of IAPP fibrils in different aging stages were probed by atomic force microscopy and sonication, which displayed comparable stiffness but age-dependent fragmentation, followed by self-assembly of such fragments into the largest lamellar amyloid structures reported to date. The dynamic structural and toxicity profiles of amyloid fibrils and plaques suggest that they play active, long-term roles in cell degeneration and may be a therapeutic target for amyloid diseases.

Understanding Effects of PAMAM Dendrimer Size and Surface Chemistry on Serum Protein Binding with Discrete Molecular Dynamics Simulations

Polyamidoamine (PAMAM) dendrimers, a class of polymeric nanoparticles (NPs) with highly-controllable sizes and surface chemistry, are promising candidates for many biomedical applications, including drug and gene delivery, imaging, and inhibition of amyloid aggregation. In circulation, binding of serum proteins with dendritic NPs renders the formation of protein corona and alters the biological identity of the NP core, which may subsequently elicit immunoresponse and cytotoxicity. Understanding the effects of PAMAM size and surface chemistry on serum protein binding is, therefore, crucial to enable their broad biomedical applications. Here, by applying atomistic discrete molecular dynamics (DMD) simulations, we first uncovered the binding of PAMAM with HSA and Ig and detailed the dependences of such binding on PAMAM size and surface modification. Compared to either anionic or cationic surfaces, modifications with neutral phosphorylcholine (PC), polyethylene glycol (PEG), and hydroxyls (OH) significantly reduced binding with proteins. The relatively strong binding between proteins and PAMAM dendrimers with charged surface groups was mainly driven by electrostatic interactions as well as hydrophobic interactions. Using steered DMD (SDMD) simulations, we conducted a force-pulling experiment in silico estimating the critical forces separating PAMAM-protein complexes and deriving the corresponding free energy barriers for dissociation. The SDMD-derived HSA-binding affinities were consistent with existing experimental measurements. Our results highlighted the association dynamics of protein-dendrimer interactions and binding affinities, whose implications range from fundamental nanobio interfacial phenomena to the development of "stealth NPs".

Nanoscale inhibition of polymorphic and ambidextrous IAPP amyloid aggregation with small molecules

Understanding how small molecules interface amyloid fibrils on the nanoscale is of importance for developing therapeutic treatment against amyloid-based diseases. Here we show, for the first time, that human islet amyloid polypeptide (IAPP) in the fibrillar form is polymorphic and ambidextrous possessing multiple periodicities. Upon interfacing with small molecule epigallocatechin gallate (EGCG), IAPP aggregation was rendered off pathway assuming the form of soft and disordered clusters, while mature IAPP fibrils displayed kinks and branching but conserved the twisted fibril morphology. These nanoscale phenomena resulted from competitive interactions between EGCG and the IAPP amyloidogenic region, as well as end capping of the fibrils by the small molecule. This information is crucial to delineating IAPP toxicity implicated in type 2 diabetes and developing new inhibitors against amyloidogenesis.

Star Polymers Reduce Islet Amyloid Polypeptide Toxicity via Accelerated Amyloid Aggregation

Protein aggregation into amyloid fibrils is a ubiquitous phenomenon across the spectrum of neurodegenerative disorders and type 2 diabetes. A common strategy against amyloidogenesis is to minimize the populations of toxic oligomers and protofibrils by inhibiting protein aggregation with small molecules or nanoparticles. However, melanin synthesis in nature is realized by accelerated protein fibrillation to circumvent accumulation of toxic intermediates. Accordingly, we designed and demonstrated the use of star-shaped poly(2-hydroxyethyl acrylate) (PHEA) nanostructures for promoting aggregation while ameliorating the toxicity of human islet amyloid polypeptide (IAPP), the peptide involved in glycemic control and the pathology of type 2 diabetes. The binding of PHEA elevated the β-sheet content in IAPP aggregates while rendering a new morphology of "stelliform" amyloids originating from the polymers. Atomistic molecular dynamics simulations revealed that the PHEA arms served as rodlike scaffolds for IAPP binding and subsequently accelerated IAPP aggregation by increased local peptide concentration. The tertiary structure of the star nanoparticles was found to be essential for driving the specific interactions required to impel the accelerated IAPP aggregation. This study sheds new light on the structure-toxicity relationship of IAPP and points to the potential of exploiting star polymers as a new class of therapeutic agents against amyloidogenesis.

Implications of peptide assemblies in amyloid diseases

Neurodegenerative disorders and type 2 diabetes are global epidemics compromising the quality of life of millions worldwide, with profound social and economic implications. Despite the significant differences in pathology - much of which are poorly understood - these diseases are commonly characterized by the presence of cross-β amyloid fibrils as well as the loss of neuronal or pancreatic β-cells. In this review, we document research progress on the molecular and mesoscopic self-assembly of amyloid-beta, alpha synuclein, human islet amyloid polypeptide and prions, the peptides and proteins associated with Alzheimer's, Parkinson's, type 2 diabetes and prion diseases. In addition, we discuss the toxicities of these amyloid proteins based on their self-assembly as well as their interactions with membranes, metal ions, small molecules and engineered nanoparticles. Through this presentation we show the remarkable similarities and differences in the structural transitions of the amyloid proteins through primary and secondary nucleation, the common evolution from disordered monomers to alpha-helices and then to β-sheets when the proteins encounter the cell membrane, and, the consensus (with a few exceptions) that off-pathway oligomers, rather than amyloid fibrils, are the toxic species regardless of the pathogenic protein sequence or physicochemical properties. In addition, we highlight the crucial role of molecular self-assembly in eliciting the biological and pathological consequences of the amyloid proteins within the context of their cellular environments and their spreading between cells and organs. Exploiting such structure-function-toxicity relationship may prove pivotal for the detection and mitigation of amyloid diseases.

Mechanistic Insights from Discrete Molecular Dynamics Simulations of Pesticide-Nanoparticle Interactions

Nanoscale particles have the potential to modulate the transport, lifetimes, and ultimate uptake of pesticides that may otherwise be bound to agricultural soils. Engineered nanoparticles provide a unique platform for studying these interactions. In this study, we utilized discrete molecular dynamics (DMD) as a screening tool for examining nanoparticle-pesticide adsorptive interactions. As a proof-of-concept, we selected a library of 15 pesticides common in the United States and 4 nanomaterials with likely natural or incidental sources, and simulated all possible nanoparticle-pesticide pairs. The resulting adsorption coefficients derived from DMD simulations ranged over several orders of magnitude, and in many cases were significantly stronger than pesticide adsorption on clay surfaces, highlighting the significance of specific nanoscale phases as a preferential media with which pesticides may associate. Binding was found to be significantly enhanced by the capacity to form hydrogen bonds with slightly hydroxylated fullerols, highlighting the importance of considering the precise nature of weathered nanomaterials as opposed to pristine precursors. Results were compared to experimental adsorption studies using selected pesticides, with a Pearson correlation coefficient of 0.97.

CSAR 2014: A Benchmark Exercise Using Unpublished Data from Pharma

The 2014 CSAR Benchmark Exercise was the last community-wide exercise that was conducted by the group at the University of Michigan, Ann Arbor. For this event, GlaxoSmithKline (GSK) donated unpublished crystal structures and affinity data from in-house projects. Three targets were used: tRNA (m1G37) methyltransferase (TrmD), Spleen Tyrosine Kinase (SYK), and Factor Xa (FXa). A particularly strong feature of the GSK data is its large size, which lends greater statistical significance to comparisons between different methods. In Phase 1 of the CSAR 2014 Exercise, participants were given several protein-ligand complexes and asked to identify the one near-native pose from among 200 decoys provided by CSAR. Though decoys were requested by the community, we found that they complicated our analysis. We could not discern whether poor predictions were failures of the chosen method or an incompatibility between the participant's method and the setup protocol we used. This problem is inherent to decoys, and we strongly advise against their use. In Phase 2, participants had to dock and rank/score a set of small molecules given only the SMILES strings of the ligands and a protein structure with a different ligand bound. Overall, docking was a success for most participants, much better in Phase 2 than in Phase 1. However, scoring was a greater challenge. No particular approach to docking and scoring had an edge, and successful methods included empirical, knowledge-based, machine-learning, shape-fitting, and even those with solvation and entropy terms. Several groups were successful in ranking TrmD and/or SYK, but ranking FXa ligands was intractable for all participants. Methods that were able to dock well across all submitted systems include MDock,1 Glide-XP,2 PLANTS,3 Wilma,4 Gold,5 SMINA,6 Glide-XP2/PELE,7 FlexX,8 and MedusaDock.9 In fact, the submission based on Glide-XP2/PELE7 cross-docked all ligands to many crystal structures, and it was particularly impressive to see success across an ensemble of protein structures for multiple targets. For scoring/ranking, submissions that showed statistically significant achievement include MDock1 using ITScore1,10 with a flexible-ligand term,11 SMINA6 using Autodock-Vina,12,13 FlexX8 using HYDE,14 and Glide-XP2 using XP DockScore2 with and without ROCS15 shape similarity.16 Of course, these results are for only three protein targets, and many more systems need to be investigated to truly identify which approaches are more successful than others. Furthermore, our exercise is not a competition.

Stabilizing Off-pathway Oligomers by Polyphenol Nanoassemblies for IAPP Aggregation Inhibition

Experimental studies have shown that many naturally occurring polyphenols have inhibitory effect on the aggregation of several proteins. Here, we use discrete molecular dynamics (DMD) simulations and high-throughput dynamic light scattering (DLS) experiments to study the anti-aggregation effects of two polyphenols, curcumin and resveratrol, on the aggregation of islet amyloid polypeptide (IAPP or amylin). Our DMD simulations suggest that the aggregation inhibition is caused by stabilization of small molecular weight IAPP off-pathway oligomers by the polyphenols. Our analysis indicates that IAPP-polyphenol hydrogen bonds and π-π stacking combined with hydrophobic interactions are responsible for the stabilization of oligomers. The presence of small oligomers is confirmed with DLS measurements in which nanometer-sized oligomers are found to be stable for up to 7.5 hours, the time frame within which IAPP aggregates in the absence of polyphenols. Our study offers a general anti-aggregation mechanism for polyphenols, and further provides a computational framework for the future design of anti-amyloid aggregation therapeutics.

Combined Approach of Patch-Surfer and PL-PatchSurfer for Protein-Ligand Binding Prediction in CSAR 2013 and 2014

The Community Structure-Activity Resource (CSAR) benchmark exercise provides a unique opportunity for researchers to objectively evaluate the performance of protein-ligand docking methods. Patch-Surfer and PL-PatchSurfer, molecular surface-based methods for predicting binding ligands of proteins developed in our group, were tested on both CSAR 2013 and 2014 benchmark exercises in combination with an empirical scoring function-based method, AutoDock, while we only participated in CSAR 2013 using Patch-Surfer. The prediction results for Phase 1 task in CSAR 2013 showed that Patch-Surfer was able to rank all the four designed binding proteins within top ranks, outperforming AutoDock Vina. In Phase 2 of 2013, PL-PatchSurfer correctly selected the correct ligand pose for two target proteins. PL-PatchSurfer performed reasonably well in ranking ligands according to their binding affinity and in selecting near-native ligand poses in 2013 Phase 3 and 2014 Phase 1, respectively, although AutoDock Vina showed better performance. Lastly, in the 2014 Phase 2 exercise, the PL-PatchSurfer scores computed for ligands to target protein pairs correlated well with their pIC50 values, which was better or comparable to results by other participants. Overall, our methods showed fairly good performance in CSAR 2013 and 2014. Unique characteristics of the methods are discussed in comparison with AutoDock.