A fragment-based approach to the SAMPL3 Challenge

Water Network in the Binding Pocket of Fluorinated BPTI-Trypsin Complexes─Insights from Simulation and Experiment

Structural waters in the S1 binding pocket of β-trypsin are critical for the stabilization of the complex of β-trypsin with its inhibitor bovine pancreatic trypsin inhibitor (BPTI). The inhibitor strength of BPTI can be modulated by replacing the critical lysine residue at the P1 position by non-natural amino acids. We study BPTI variants in which the critical Lys15 in BPTI has been replaced by α-aminobutyric acid (Abu) and its fluorinated derivatives monofluoroethylglycine (MfeGly), difluoroethylglycine (DfeGly), and trifluoroethylglycine (TfeGly). We investigate the hypothesis that additional water molecules in the binding pocket can form specific noncovalent interactions with the fluorinated side chains and thereby act as an extension of the inhibitors. We report potentials of mean force (PMF) of the unbinding process for all four complexes and enzyme activity inhibition assays. Additionally, we report the protein crystal structure of the Lys15MfeGly-BPTI-β-trypsin complex (pdb: 7PH1). Both experimental and computational data show a stepwise increase in inhibitor strength with increasing fluorination of the Abu side chain. The PMF additionally shows a minimum for the encounter complex and an intermediate state just before the bound state. In the bound state, the computational analysis of the structure and dynamics of the water molecules in the S1 pocket shows a highly dynamic network of water molecules that does not indicate a rigidification or stabilizing trend in regard to energetic properties that could explain the increase in inhibitor strength. The analysis of the energy and the entropy of the water molecules in the S1 binding pocket using grid inhomogeneous solvation theory confirms this result. Overall, fluorination systematically changes the binding affinity, but the effect cannot be explained by a persistent water network in the binding pocket. Other effects, such as the hydrophobicity of fluorinated amino acids and the stability of the encounter complex as well as the additional minimum in the potential of mean force in the bound state, likely influence the affinity more directly.

SAMPL7 protein-ligand challenge: A community-wide evaluation of computational methods against fragment screening and pose-prediction

A novel crystallographic fragment screening data set was generated and used in the SAMPL7 challenge for protein-ligands. The SAMPL challenges prospectively assess the predictive power of methods involved in computer-aided drug design. Application of various methods to fragment molecules are now widely used in the search for new drugs. However, there is little in the way of systematic validation specifically for fragment-based approaches. We have performed a large crystallographic high-throughput fragment screen against the therapeutically relevant second bromodomain of the Pleckstrin-homology domain interacting protein (PHIP2) that revealed 52 different fragments bound across 4 distinct sites, 47 of which were bound to the pharmacologically relevant acetylated lysine (Kac) binding site. These data were used to assess computational screening, binding pose prediction and follow-up enumeration. All submissions performed randomly for screening. Pose prediction success rates (defined as less than 2 Å root mean squared deviation against heavy atom crystal positions) ranged between 0 and 25% and only a very few follow-up compounds were deemed viable candidates from a medicinal-chemistry perspective based on a common molecular descriptors analysis. The tight deadlines imposed during the challenge led to a small number of submissions suggesting that the accuracy of rapidly responsive workflows remains limited. In addition, the application of these methods to reproduce crystallographic fragment data still appears to be very challenging. The results show that there is room for improvement in the development of computational tools particularly when applied to fragment-based drug design.

Evaluation of log P, pKa, and log D predictions from the SAMPL7 blind challenge

The Statistical Assessment of Modeling of Proteins and Ligands (SAMPL) challenges focuses the computational modeling community on areas in need of improvement for rational drug design. The SAMPL7 physical property challenge dealt with prediction of octanol-water partition coefficients and pK_a for 22 compounds. The dataset was composed of a series of N-acylsulfonamides and related bioisosteres. 17 research groups participated in the log P challenge, submitting 33 blind submissions total. For the pK_a challenge, 7 different groups participated, submitting 9 blind submissions in total. Overall, the accuracy of octanol-water log P predictions in the SAMPL7 challenge was lower than octanol-water log P predictions in SAMPL6, likely due to a more diverse dataset. Compared to the SAMPL6 pK_a challenge, accuracy remains unchanged in SAMPL7. Interestingly, here, though macroscopic pK_a values were often predicted with reasonable accuracy, there was dramatically more disagreement among participants as to which microscopic transitions produced these values (with methods often disagreeing even as to the sign of the free energy change associated with certain transitions), indicating far more work needs to be done on pK_a prediction methods.

Fragment-based design of small molecule PCSK9 inhibitors using simulated annealing of chemical potential simulations

PCSK9 is a protein secreted by the liver that binds to the low-density lipoprotein receptor (LDLR), causing LDLR internalization, decreasing the clearance of circulating LDL particles. Mutations in PCSK9 that strengthen its interactions with LDLR result in familial hypercholesterolemia (FH) and early onset atherosclerosis, while nonsense mutations of PCSK9 result in cardio-protective hypocholesterolemia. These observations led to PCSK9 inhibition for cholesterol lowering becoming a high-interest therapeutic target, with antibody drugs reaching the market. An orally-available small molecule drug is highly desirable, but inhibiting the PCSK9/LDLR protein-protein interaction (PPI) has proven challenging. Alternate approaches to finding good lead candidates are needed. Motivated by the FH mutation data on PCSK9, we found that modeling the PCSK9/LDLR interface revealed extensive electron delocalization between and within the protein partners. Based on this, we hypothesized that compounds assembled from chemical fragments could achieve the affinity required to inhibit the PCSK9/LDLR PPI if they were selected to interact with PCSK9 in a way that, like LDLR, also involves significant fractional charge transfer to form partially covalent bonds. To identify such fragments, Simulated Annealing of Chemical Potential (SACP) fragment simulations were run on multiple PCSK9 structures, using optimized partial charges for the protein. We designed a small molecule, composed of several fragments, predicted to interact at two sites on the PCSK9. This compound inhibits the PPI with 1 μM affinity. Further, we designed two similar small molecules where one allows charge delocalization though a linker and the other doesn’t. The first inhibitor with charge delocalization enhances LDLR surface expression by 60% at 10 nM, two orders of magnitude more potent than the EGF domain of LDLR. The other enhances LDLR expression by only 50% at 1 μM. This supports our conjecture that fragments can have surprisingly outsized efficacy in breaking PPI’s by achieving fractional charge transfer leading to partially covalent bonding.

Computational design of environmental sensors for the potent opioid fentanyl

We describe the computational design of proteins that bind the potent analgesic fentanyl. Our approach employs a fast docking algorithm to find shape complementary ligand placement in protein scaffolds, followed by design of the surrounding residues to optimize binding affinity. Co-crystal structures of the highest affinity binder reveal a highly preorganized binding site, and an overall architecture and ligand placement in close agreement with the design model. We use the designs to generate plant sensors for fentanyl by coupling ligand binding to design stability. The method should be generally useful for detecting toxic hydrophobic compounds in the environment.

Many small molecules, including toxins and some medicines, have flexible structures, which makes it difficult to detect and/or neutralize them. The pain medication fentanyl, for example, can rotate to adopt many shapes. In recent years, fentanyl drug abuse has become increasingly common, and the drug is often illegally produced. The number of deaths caused by fentanyl has risen greatly, which provides a strong reason to find new ways to detect this and other drugs.

Now, Baker et al. have created new sensors that are able to detect fentanyl. First, the 11 most likely shapes that fentanyl could adopt were identified based on known information about the structure of the molecule. Then, a computer program was used to design proteins that were predicted to strongly bind to these most common shapes. Next, genes that coded for these proteins were synthesized in the laboratory and introduced into bacteria, which read the genes to build the proteins.

Similar to a well-fitted lock and key, the shape of the newly designed protein had to complement a likely shape of the fentanyl molecule. Baker et al. used a technique called X-ray crystallography to visualize the proteins in atomic detail and confirm that these fentanyl-binders matched their corresponding computational models. Those proteins that bound fentanyl best were then engineered into plant cells, and later into whole plants, together with reporter systems that gave signals when the sensors detected fentanyl.

In future, these specifically synthesized proteins could be integrated into entire panels of plants or other systems to detect toxins and other harmful chemicals. Such systems would be of interest in a medical setting and for detecting environmental contamination.

Hot-spot identification on a broad class of proteins and RNA suggest unifying principles of molecular recognition

Chemically diverse fragments tend to collectively bind at localized sites on proteins, which is a cornerstone of fragment-based techniques. A central question is how general are these strategies for predicting a wide variety of molecular interactions such as small molecule-protein, protein-protein and protein-nucleic acid for both experimental and computational methods. To address this issue, we recently proposed three governing principles, (1) accurate prediction of fragment-macromolecule binding free energy, (2) accurate prediction of water-macromolecule binding free energy, and (3) locating sites on a macromolecule that have high affinity for a diversity of fragments and low affinity for water. To test the generality of these concepts we used the computational technique of Simulated Annealing of Chemical Potential to design one small fragment to break the RecA-RecA protein-protein interaction and three fragments that inhibit peptide-deformylase via water-mediated multi-body interactions. Experiments confirm the predictions that 6-hydroxydopamine potently inhibits RecA and that PDF inhibition quantitatively tracks the water-mediated binding predictions. Additionally, the principles correctly predict the essential bound waters in HIV Protease, the surprisingly extensive binding site of elastase, the pinpoint location of electron transfer in dihydrofolate reductase, the HIV TAT-TAR protein-RNA interactions, and the MDM2-MDM4 differential binding to p53. The experimental confirmations of highly non-obvious predictions combined with the precise characterization of a broad range of known phenomena lend strong support to the generality of fragment-based methods for characterizing molecular recognition.

Design, synthesis and biological evaluation of renin inhibitors guided by simulated annealing of chemical potential simulations

We have applied simulated annealing of chemical potential (SACP) to a diverse set of ∼150 very small molecules to provide insights into new interactions in the binding pocket of human renin, a historically difficult target for which to find low molecular weight (MW) inhibitors with good bioavailability. In one of its many uses in drug discovery, SACP provides an efficient, thermodynamically principled method of ranking chemotype replacements for scaffold hopping and manipulating physicochemical characteristics for drug development. We introduce the use of Constrained Fragment Analysis (CFA) to construct and analyze ligands composed of linking those fragments with predicted high affinity. This technique addresses the issue of effectively linking fragments together and provides a predictive mechanism to rank order prospective inhibitors for synthesis. The application of these techniques to the identification of novel inhibitors of human renin is described. Synthesis of a limited set of designed compounds provided potent, low MW analogs (IC₅₀s<100nM) with good oral bioavailability (F>20-58%).

Absolute binding free energy calculations of CBClip host-guest systems in the SAMPL5 blind challenge

Herein, we report the absolute binding free energy calculations of CBClip complexes in the SAMPL5 blind challenge. Initial conformations of CBClip complexes were obtained using docking and molecular dynamics simulations. Free energy calculations were performed using thermodynamic integration (TI) with soft-core potentials and Bennett's acceptance ratio (BAR) method based on a serial insertion scheme. We compared the results obtained with TI simulations with soft-core potentials and Hamiltonian replica exchange simulations with the serial insertion method combined with the BAR method. The results show that the difference between the two methods can be mainly attributed to the van der Waals free energies, suggesting that either the simulations used for TI or the simulations used for BAR, or both are not fully converged and the two sets of simulations may have sampled difference phase space regions. The penalty scores of force field parameters of the 10 guest molecules provided by CHARMM Generalized Force Field can be an indicator of the accuracy of binding free energy calculations. Among our submissions, the combination of docking and TI performed best, which yielded the root mean square deviation of 2.94 kcal/mol and an average unsigned error of 3.41 kcal/mol for the ten guest molecules. These values were best overall among all participants. However, our submissions had little correlation with experiments.

Designing a small molecule erythropoietin mimetic

Erythropoietin (EPO) is a protein made by the kidneys in response to low red blood cell count that is secreted into the bloodstream and binds to a receptor on hematopoietic stem cells in the bone marrow inducing them to become new red blood cells. EPO made with recombinant DNA technology was brought to market in the 1980s to treat anemia caused by kidney disease and cancer chemotherapy. Because EPO infusion was able to replace blood transfusions in many cases, it rapidly became a multibillion dollar per year drug and as the first biologic created with recombinant technology it launched the biotech industry. For many years intense research was focused on creating a small molecule orally available EPO mimetic. The Robert Wood Johnson (RWJ) group seemed to definitively establish that only large peptides with a minimum of 60 residues could replace EPO, as anything less was not a full agonist. An intense study of the published work led me to hypothesize that the size of the mimetic is not the real issue, but the symmetry making and breaking of the EPO receptor induced by the ligand is the key to activating the stem cells. This analysis meant that residues in the binding site of the receptor deemed absolutely essential for ligand binding and activation from mutagenesis experiments, were probably not really that important. My fundamental hypotheses were: (a) the symmetric state of the homodimeric receptor is the most stable state and thus must be the off-state, (b) a highly localized binding site exists at a pivot point where the two halves of the receptor meet, (c) small molecules can be created that have high potency for this site that will be competitive with EPO and thus can displace the protein-protein interaction, (d) small symmetric molecules will stabilize the symmetric off-state of the receptor, and (e) a key asymmetry in the small molecule will stabilize a mirror image asymmetry in the receptor resulting in the stabilization of the on-state and proliferation of the stem cells into red blood cells. Researchers at Amgen published a co-crystal structure of EPO bound to the EPO receptor, which has a beautiful twofold symmetry-it was argued that this is the active state of the receptor. Activating the EPO receptor with EPO induces an almost instantaneous shutdown mechanism to sharply curtail any proliferative signal transduction, and thus, my hypotheses lead to the conclusion that the Amgen co-crystal is actually the state after receptor downregulation and thus an off-state. To put these hypotheses to the test, my computational method of Simulated Annealing of Chemical Potential was run using the co-crystal created at RWJ, which is the receptor trapped in a partial agonist state. The simulations predicted a previously unknown high affinity binding site at the pivot point where the two halves of the dimeric receptor meet, and detailed analysis of the fragment patterns led to the prediction of a molecule less than 300 MW that is basically twofold symmetric with a chiral center on one side and not the other. Thus, to the degree that computer simulations can be taken seriously, these results support my hypotheses on small molecule receptor activation. When this small molecule was synthesized and tested it indeed induced human hematopoietic stems cells to become red blood cells. When the predicted chiral center of this molecule was removed eliminating its one asymmetric feature, the resulting molecule was an antagonist-it could potently displace hot EPO but could no longer induce stem cell proliferation and differentiation. These results provided strong support for my theories on how to create potent small molecule EPO agonists and were used to launch the new company Locus Pharmaceuticals. These molecules, however, required significant chemical changes in order to make them stable in other in vitro assays and to be in vivo active, but these alterations had to be done in a way that maintained the symmetry-asymmetry considerations that led to the creation of an in vitro active molecule. The combination of changing functional groups to enable good pharmacokinetics, while not changing the key intrinsic symmetry properties were never seriously pursued at Locus and the program died. Investigations into how red blood cells are created have occupied many prominent researchers for the entire twentieth century. In the second half of the century EPO was discovered and by the end of the century it became a blockbuster commercial product that launched the biotech revolution.

Designing an orally available nontoxic p38 inhibitor with a fragment-based strategy

The MAPK p38 became a focal point of inflammatory research when it was recognized that it played a key role in the production of the pro-inflammatory molecules TNF-alpha, IL-beta, and cyclooxygenase-2 (COX-2). The pharmaceutical industry devoted enormous efforts to creating p38 inhibitors, because blocking p38 had the potential of downregulating a group of pro-inflammatory mediators, and thus, one drug could have a cocktail effect. The market potential seemed to be clearly established (Bonafede et al., Clinicoecon Outcomes Res 6:381-388, 2014) with a multiplicity of TNF-alpha antibodies and a soluble receptor (Mewar and Wilson, Br J Pharmacol 162:785-791, 2011) already on the market, although the relationship between TNF-alpha production and p38 activation is a complicated two-way (Sabio and Davis, Semin Immunol 26:237-245, 2014) signal transduction process. With the discovery that activated p38 stabilizes (Mancini and Di Battista, Inflamm Res 60:1083-1092, 2011) COX-2 mRNA and upregulates expression of IL-beta (Bachstetter and Van Eldik, Aging Dis 1:199-211, 2010) probably in a similar manner, inhibiting p38 appeared to be a way of blocking TNF-alpha, COX-2, and IL-beta simultaneously. At Locus Pharmaceuticals we jumped on this opportunity, because we believed that our fragment-based drug discovery approach was ideally suited for making a potent small molecule p38 inhibitor that did not bind in the ATP site, but also had the solubility, lack of planarity, and low molecular weight required of a clinical candidate. Just to be clear, in our experience highly planar compounds often result in poor pharmacokinetics, because they tend to bind strongly to plasma proteins. At Locus we typically repeated assays by adding increasing amounts of plasma to check for potency degradation in the presence of blood. We found this to be a useful early indicator of pharmacokinetics and in vivo behavior. It became clear from our work and the work of others that binding to the ATP site resulted in nonspecific isoform toxicities, but binding in the adjacent allosteric DFG-site resulted in molecules that were too planar and too hydrophobic. Applying the computational method of Simulated Annealing of Chemical Potential (SACP) to this problem, we at Locus were able to come up with surprising fragment substitution patterns that led to potent non-ATP p38 inhibitors with the solubility and lack of planarity that resulted in potent in vivo efficacy in rodents with 33 % oral bioavailability. By using the simulations, we made only a small number of molecules and created a high quality clinical candidate. We also did extensive co-crystallography work, which demonstrated that the compounds bound in the mode predicted by the simulations. Unfortunately, all p38 programs ultimately shut down, because compelling evidence emerged that inhibiting p38 had no long-term clinical (Genovese, Arthritis Rheum 60:317-320, 2009) benefit. Devoting a large amount of limited resources to a target that ultimately turns out to be a mistake because it was not properly validated is a fatal error for a small company, and this is one of the reasons that Locus ultimately failed.

Computational methods for fragment-based ligand design: growing and linking

Fragment-based drug design has proved itself as a powerful technique for increasing the sampling and diversity of chemical space and enabling the design of novel leads and compounds. Computational techniques for identifying fragments, binding sites and particularly for linking, growing, and evolving fragments play a significant role in the process. Information from ADME studies and clustering property information in the form of toxicophores and chemotypes can play a significant role in aiding the design of novel, selective fragments with good activity profiles.

Blind prediction of HIV integrase binding from the SAMPL4 challenge

Here, we give an overview of the protein-ligand binding portion of the Statistical Assessment of Modeling of Proteins and Ligands 4 (SAMPL4) challenge, which focused on predicting binding of HIV integrase inhibitors in the catalytic core domain. The challenge encompassed three components--a small "virtual screening" challenge, a binding mode prediction component, and a small affinity prediction component. Here, we give summary results and statistics concerning the performance of all submissions at each of these challenges. Virtual screening was particularly challenging here in part because, in contrast to more typical virtual screening test sets, the inactive compounds were tested because they were thought to be likely binders, so only the very top predictions performed significantly better than random. Pose prediction was also quite challenging, in part because inhibitors in the set bind to three different sites, so even identifying the correct binding site was challenging. Still, the best methods managed low root mean squared deviation predictions in many cases. Here, we give an overview of results, highlight some features of methods which worked particularly well, and refer the interested reader to papers in this issue which describe specific submissions for additional details.

Multipose binding in molecular docking

Molecular docking has been extensively applied in virtual screening of small molecule libraries for lead identification and optimization. A necessary prerequisite for successful differentiation between active and non-active ligands is the accurate prediction of their binding affinities in the complex by use of docking scoring functions. However, many studies have shown rather poor correlations between docking scores and experimental binding affinities. Our work aimed to improve this correlation by implementing a multipose binding concept in the docking scoring scheme. Multipose binding, i.e., the property of certain protein-ligand complexes to exhibit different ligand binding modes, has been shown to occur in nature for a variety of molecules. We conducted a high-throughput docking study and implemented multipose binding in the scoring procedure by considering multiple docking solutions in binding affinity prediction. In general, improvement of the agreement between docking scores and experimental data was observed, and this was most pronounced in complexes with large and flexible ligands and high binding affinities. Further developments of the selection criteria for docking solutions for each individual complex are still necessary for a general utilization of the multipose binding concept for accurate binding affinity prediction by molecular docking.