Recently much effort has been invested in using convolutional neural network (CNN) models trained on 3D structural images of protein-ligand complexes to distinguish binding from non-binding ligands for virtual screening. However, the dearth of reliable protein-ligand x-ray structures and binding affinity data has required the use of constructed datasets for the training and evaluation of CNN molecular recognition models. Here, we outline various sources of bias in one such widely-used dataset, the Directory of Useful Decoys: Enhanced (DUD-E). We have constructed and performed tests to investigate whether CNN models developed using DUD-E are properly learning the underlying physics of molecular recognition, as intended, or are instead learning biases inherent in the dataset itself. We find that superior enrichment efficiency in CNN models can be attributed to the analogue and decoy bias hidden in the DUD-E dataset rather than successful generalization of the pattern of protein-ligand interactions. Comparing additional deep learning models trained on PDBbind datasets, we found that their enrichment performances using DUD-E are not superior to the performance of the docking program AutoDock Vina. Together, these results suggest that biases that could be present in constructed datasets should be thoroughly evaluated before applying them to machine learning based methodology development.

The expulsion of water from surfaces upon molecular recognition and non-specific association makes a major contribution to the free energy changes of these processes. In order to facilitate the characterization of water structure and thermodynamics on surfaces, we have incorporated Grid Inhomogeneous Solvation Theory (GIST) into the CPPTRAJ toolset of AmberTools. GIST is a grid-based implementation of Inhomogeneous fluid Solvation Theory, which analyzes the output from molecular dynamics simulations to map out solvation thermodynamic and structural properties on a high-resolution, three-dimensional grid. The CPPTRAJ implementation, called GIST-cpptraj, has a simple, easy-to-use command line interface, and is open source and freely distributed. We have also developed a set of open-source tools, called GISTPP, which facilitate the analysis of GIST output grids. Tutorials for both GIST-cpptraj and GISTPP can be found at ambermd.org.

Over the past few decades, farmers have increasingly integrated cover crops into their cropping systems. Cover-crop benefits can help a farmer to achieve sustainability or reduce negative environmental externalities, such as soil erosion or chemical runoff. However, the impact on farm economics will likely be the strongest incentive to adopt cover crops. These impacts can include farm profits, cash crop yields or both. This paper provides a review of cover-crop adoption, production, risk and policy considerations from an economic perspective. These dimensions are examined through a review of cover-crop literature. This review was written to provide an overview of cover crops and their impacts on the farm business and the environment, especially with regard to economic considerations. Through increasing knowledge about cover crops, the intent here is to inform producers contemplating adoption and policy makers seeking to encourage adoption.

<p>Recently much effort has been invested in using convolutional neural network (CNN) models trained on 3D structural images of protein-ligand complexes to distinguish binding from non-binding ligands for virtual screening. However, the dearth of reliable protein-ligand x-ray structures and binding affinity data has required the use of constructed datasets for the training and evaluation of CNN molecular recognition models. Here, we outline various sources of bias in one such widely-used dataset, the Directory of Useful Decoys: Enhanced (DUD-E). We have constructed and performed tests to investigate whether CNN models developed using DUD-E are properly learning the underlying physics of molecular recognition, as intended, or are instead learning biases inherent in the dataset itself. We find that superior enrichment efficiency in CNN models can be attributed to the analogue and decoy bias hidden in the DUD-E dataset rather than successful generalization of the pattern of protein-ligand interactions. Comparing additional deep learning models trained on PDBbind datasets, we found that their enrichment performances using DUD-E are not superior to the performance of the docking program AutoDock Vina. Together, these results suggest that biases that could be present in constructed datasets should be thoroughly evaluated before applying them to machine learning based methodology development. </p>

The principles underlying water reorganization around simple non-polar solutes are well understood and provide the framework for classical hydrophobic effect, whereby water molecules structure themselves around solutes so that they maintain favorable energetic contacts with both the solute and with other water molecules. However, for certain solute surface topographies, water molecules, due to their geometry and size, are unable to simultaneously maintain favorable energetic contacts with both the surface and neighboring water molecules. In this study, we analyze the solvation of ligand-binding sites for six structurally diverse proteins using hydration site analysis and measures of local water structure, in order to identify surfaces at which water molecules are unable to structure themselves in a way that maintains favorable enthalpy relative to bulk water. These surfaces are characterized by a high degree of enclosure, weak solute-water interactions, and surface constraints that induce unfavorable pair interactions between neighboring water molecules. Additionally, we find that the solvation of charged side-chains in an active site generally results in favorable enthalpy but can also lead to pair interactions between neighboring water molecules that are significantly unfavorable relative to bulk water. We find that frustrated local structure can occur not only in apolar and weakly polar pockets, where overall enthalpy tends to be unfavorable, but also in charged pockets, where overall water enthalpy tends to be favorable. The characterization of local water structure in these terms may prove useful for evaluating the displacement of water from diverse protein active-site environments.

We have developed SSTMap, a software package for mapping structural and thermodynamic water properties in molecular dynamics trajectories. The package introduces automated analysis and mapping of local measures of frustration and enhancement of water structure. The thermodynamic calculations are based on Inhomogeneous Fluid Solvation Theory (IST), which is implemented using both site-based and grid-based approaches. The package also extends the applicability of solvation analysis calculations to multiple molecular dynamics (MD) simulation programs by using existing cross-platform tools for parsing MD parameter and trajectory files. SSTMap is implemented in Python and contains both command-line tools and a Python module to facilitate flexibility in setting up calculations and for automated generation of large data sets involving analysis of multiple solutes. Output is generated in formats compatible with popular Python data science packages. This tool will be used by the molecular modeling community for computational analysis of water in problems of biophysical interest such as ligand binding and protein function.

Serotonin receptors (5-HT3AR) play a crucial role in regulating gut movement, and are the principal target of setrons, a class of high-affinity competitive antagonists, used in the management of nausea and vomiting associated with radiation and chemotherapies. Structural insights into setron-binding poses and their inhibitory mechanisms are just beginning to emerge. Here, we present high-resolution cryo-EM structures of full-length 5-HT3AR in complex with palonosetron, ondansetron, and alosetron. Molecular dynamic simulations of these structures embedded in a fully-hydrated lipid environment assessed the stability of ligand-binding poses and drug-target interactions over time. Together with simulation results of apo- and serotonin-bound 5-HT3AR, the study reveals a distinct interaction fingerprint between the various setrons and binding-pocket residues that may underlie their diverse affinities. In addition, varying degrees of conformational change in the setron-5-HT3AR structures, throughout the channel and particularly along the channel activation pathway, suggests a novel mechanism of competitive inhibition.

In this study, we demonstrate a method to construct a water-based pharmacophore model which can be utilized in the absence of known ligands. This method utilizes waters found in the binding pocket, sampled through molecular dynamics. Screening of compound databases against this water-based pharmacophore model reveals that this approach can successfully identify known binders to a target protein. The method was tested by enrichment studies of 7 therapeutically important targets and compared favourably to screening-by-docking with Glide. Our results suggest that even without experimentally known binders, pharmacophore models can be generated using molecular dynamics with waters and used for virtual screening.