DoGSiteScorer is a grid-based method which uses a Difference of Gaussian filter to detect potential binding pockets - solely based on the 3D structure of the protein - and splits them into subpockets. Global properties, describing the size, shape and chemical features of the predicted (sub)pockets are calculated. Per default, a simple druggability score is provided for each (sub)pocket, based on a linear combination of the three descriptors describing volume, hydrophobicity and enclosure. Furthermore, a subset of meaningful descriptors is incorporated in a support vector machine (libsvm) to predict the (sub)pocket druggability score (values are between zero and one). The higher the score the more druggable the pocket is estimated to be. <sup>1

1. A. Volkamer, D. Kuhn, T. Grombacher, F. Rippmann, M. Rarey. Combining global and local measures for structure-based druggability predictions. J. Chem. Inf. Model. 2012,52,360-372.</sup>

DoGSiteScorer

PoseView automatically creates two-dimensional diagrams of complexes with known 3D structure according to chemical drawing conventions.¹ Directed bonds between protein and ligand are drawn as dashed lines and the interacting protein residues and the ligand are visualized as structure diagrams. Hydrophobic contacts are represented more indirectly by means of spline sections highlighting the hydrophobic parts of the ligand and the label of the contacting amino acid. The generation of structure diagrams and their layout modifications are based on the library 2Ddraw ². Interactions between the molecules are estimated by a builtin interaction model that is based on atom types and simple geometric criteria.

<sup>1. Stierand, K., Maaß, P., Rarey, M. (2006) Molecular Complexes at a Glance: Automated Generation of two-dimensional Complex Diagrams. Bioinformatics, 22, 1710-1716.
2. Fricker, P., Gastreich, M., and Rarey, M. (2004) Automated Generation of Structural Molecular Formulas under Constraints. Journal of Chemical Information and Computer Sciences, 44, 1065-1078.</sup>

PoseView

SIENA has been developed for the automated assembly and preprocessing of protein binding site ensembles. Starting with a single binding site, SIENA searches the PDB for alternative conformations of the same or sequentially closely related binding sites. The method is based on an indexed database for identifying of perfect k-mer matches and a new algorithm for the detection of protein binding site conformations. Furthermore SIENA provides a variety of different filters for pruning the resulting conformational ensemble in order to meet a user’s case specific requirements. This involves a new algorithm for the interaction-based selection of binding site conformations as well as RMSD-based clustering for ensemble reduction. SIENA provides the user with a sequence alignment of the binding site as well as superimposed PDB structures which are, apart of the transferred coordinates, equal to the original files from the PDB and thus contain all structural details and further information. <sup>1 2

1. Bietz, S.; Rarey, M. (2015). ASCONA: Rapid Detection and Alignment of Protein Binding Site Conformations. Journal of Chemical Information and Modeling, 55(8):1747–1756.
2. Bietz, S.; Rarey, M. (2015). SIENA: Efficient Compilation of Selective Protein Binding Site Ensembles. Journal of Chemical Information and Modeling, 56(1): 248–259.</sup>

SIENA

HyPPI Prediction Server classifies a protein-protein complex concerning its interaction type into permanent, transient or crystal artifact. Permanent protein-protein complexes are only stable in their complexed state and the subunits would denature upon complex dissociation. Transient protein-protein complexes are stable in the complexed form as well as in the monomeric depending of the necessary function of the complex. Crystal artifacts have no biological function and are artifically formed during crystallization process. The discrimination is performed using two characteristics of the protein-protein complex, the hydrophobicity of the interface (ΔGHydrophobic) and the quotient of interface area ratios (IF-quotient). ΔGHydrophobic represents the energy emerging exclusively from the hydrophobic effect upon binding of two protein subunits and was calculated according to the desolvation term of the HYDE scoring function1. The IF-quotient takes the symmetry of the protein-protein interface into account. <sup>1

1.Schneider, N., Lange, G., Hindle, S., Klein, R., Rarey, M. (2013). A consistent description of HYdrogen bond and DEhydration energies in protein–ligand complexes: methods behind the HYDE scoring function. Journal of Computer-Aided Molecular Design, 27(1):15-29.</sup>

HyPPI

The electron density score for individual atoms (EDIA) quantifies the electron density fit of an atom. Multiple EDIA can be combined with the help of the power mean to compute EDIA_m, the electron density score for multiple atoms to score a set of atoms such as a ligand or residue. Substructures such as residues and ligands can be automatically analyzed with EDIA_m. Due to the power mean, an EDIA_m below 0.8 has at least three atoms with an EDIA below 0.8.
EDIA is calculated in computing the weighted sum over all relevant grid points in the sphere of interest around the atom. The radius of the sphere of interest is two times the resolution dependent electron density sphere radius. More details can be found in our soon to be published publication. EDIA above 0.8 mark well supported, EDIA in range 0.4 to 0.8 mark medium supported and EDIA below 0.4 mark badly supported atoms. Due to the power mean, an EDIA_m below 0.8 is based on at least three atoms with EDIA below 0.8.<sup>1

1. Meyder, A.; Nittinger, E.; Lange, G.; Klein, R.; Rarey, M. (2017). Estimating Electron Density Support for Individual Atoms and Molecular Fragments in X-ray Structures. Journal of Chemical Information and Modeling, 57(10): 2437–2447.</sup>

EDIA

METALizer predicts the coordination geometry of metals in metalloproteins. Potential coordination geometries of metals are matched onto the found metal interactions in the examined structure. The predicted coordination geometries and the observed metal interaction distances can be compared interactively to statistics calculated on the PDB.
Furthermore, METALizer is combined with other tools in the ProteinsPlus server: Using SIENA¹, ensembles of proteins with sequentially and structurally closely related metal binding sites can be retrieved from the PDB, superimposed and visualized. This allows the comparison of the predicted coordination geometries and metal interaction distances to statistics calculated only on related metal binding sites. Furthermore, different binding modes of ligands to the metal and of the metal within the protein can be explored.
Another option is the EDIA² filter to detect atoms that are poorly supported by electron density. These are then excluded from the METALizer analysis.

<sup>1. Bietz, S.; Rarey, M. (2016). SIENA: Efficient Compilation of Selective Protein Binding Site Ensembles. Journal of Chemical Information and Modeling, 56 (1), pp 248–259.
2. Meyder A.; Nittinger, E.; Lange, G.; Klein, R.; Rarey, M. (2017). Estimating Electron Density Support for Individual Atoms and Molecular Fragments in X-ray Structures. Journal of Chemical Information and Modeling, 57 (10), pp 2437-2447.</sup>

METALizer

The activity finder establishes a connection between crystallographic data stored in the PDB database and the activity values that can be found in the ChEMBL database. The activity finder links structural data of the PDB to activity values stored in the ChEMBL database. It utilizes information published by the platforms Ligand Expo, Swiss-Prot and ChEMBL. Ligands are extracted from the PDB and stored as unique SMILES (uSMILES). The ChEMBL ligand information is translated to uSMILES and matched with the data from PDB. Entries for which a link between PDB id, UNIPROT id and ChEMBL target id exists are retained and saved to a SQLite database. Version 23 of ChEMBL was used. PDB and Swiss-Prot data are only as up to date as the published files (access date: 15.7.2017).

ActivityFinder

WarPP¹ places water molecules for the active site of a given PDB file. Based on interaction geometries previously derived from protein crystal structures potential water positions are generated for free interaction directions (lone pairs of acceptors or hydrogen atoms of donors). Every small molecule present in the PDB structure is used to generate an active site of 6.5 Å radius around each atom. Within these active sites water molecules are placed.
In the corresponding publication we showed that WarPP is able to place 80% of water molecules within 1.0 Å distance to a crystallographically observed one.

^{1. Nittinger, E.; Flachsenberg, F.; Bietz, S.; Lange, G.; Klein, R.; Rarey, M.(2018). Placement of Water Molecules in Protein Structures: From Large-Scale Evalutations to Single-Case Examples. Journal of Chemical Information and Modeling, 58 (8), pp 1625-1637.}

WarPP

StructureProfiler was developed as an all-in-one tool to screen structures based on selection criteria typically used upon dataset assembly for structure-based design methods.¹ Test configurations based on either the Astex², the Iridium HT³, the Platinum⁴, or the combination of all three criteria catalogs. (Note that RSCC as an electron density validation criteria was replaced by EDIA_m⁵ scoring).

<sup>1. Meyder, A.; Kampen, S.; Sieg, J.; Fährrolfes, R.; Friedrich, N.-O.; Flachsenberg, F.; Rarey, M. (2018). StructureProfiler: An all-in-one Tool for 3D Protein Structure Profiling. Bioinformatics, 35(5): 874-876.
2. Hartshorn, M. J. et al. (2007) Diverse, high-quality test set for the validation of protein-ligand docking performance. Journal of Medicinal Chemistry, 50(4): 726–741.
3. Warren, G. L.; Do, T. D.; Kelley, B. P.; Nicholls, A.; Warren, S. D.; (2012). Essential considerations for using protein-ligand structures in drug discovery. Drug Discovery Today, 17(23-24), 1270–1281.
4. Friedrich, N.-O. et al. (2017). High-Quality Dataset of Protein-Bound Ligand Conformations and Its Application to Benchmarking Conformer Ensemble Generators. Journal of Chemical Information and Modeling, 57(3): 529–539.
5. Meyder, A.; Nittinger, E.; Lange, G.; Klein, R.; Rarey, M. (2017). Estimating Electron Density Support for Individual Atoms and Molecular Fragments in X-ray Structures. Journal of Chemical Information and Modeling, 57(10): 2437–2447.</sup>

StructureProfiler

GeoMine¹ enables textual, numerical and 3D searching with full chemical awareness in protein-ligand interfaces of the entire PDB dataset. The tool is based on the desktop application Pelikan². Data from PDB files is preprocessed and stored in a PostgreSQL database.

<sup>1. Diedrich, K.; Graef, J.; Schöning-Stierand, K.; Rarey, M.(2020). GeoMine: interactive pattern mining of proteinligand interfaces in the Protein Data Bank. Bioinformatics, btaa693.
2. Inhester, T.; Bietz, S.; Hilbig, M.; Schmidt, R.; Rarey, M.(2017). Index-Based Searching of Interaction Patterns in Large Collections of Protein-Ligand Interfaces. Journal of Chemical Information and Modeling, 57, 2, 148-158.</sup>

GeoMine

JAMDA is a novel and fully automated protein-ligand docking tool. It combines the TrixX docking algorithm^{1, 2} with the JAMDA scoring function³. The docking and scoring performance of JAMDA is in line with the state of the art in the field. JAMDA has an unprecedented level of automation enabling molecular docking for everyone on the web. Furthermore, the JAMDA scoring function is combined with a new gradient-based optimizer achieving high numerical stability. Starting from a protein structure and the molecules to be docked, the structures are preprocessed and docked automatically. The docking site can be defined by a reference ligand or by pocket residues (either manual or predicted by DoGSiteScorer⁴). The molecules for docking can be uploaded without any preprocessing, i.e. neither coordinates nor protonation states are required.

This is a beta version of JAMDA and docking results may change in the future.

<sup>1. Schellhammer, I.; Rarey, M. (2007). TrixX: structure-based molecule indexing for large-scale virtual screening in sublinear time. J. Comput. Aided Mol. Des., 21 (5), pp 223–238.
2. Henzler, A.M.,; Urbaczek, S.; Hilbig, M.; Rarey, M. (2014). An integrated approach to knowledge-driven structure-based virtual screening. J. Comput. Aided Mol. Des., 28 (9), pp 927–939.
3. Flachsenberg, F.; Meyder, A.; Penner, P.; Sommer, K.; Rarey, M. (2020). A Consistent Scheme for Gradient-Based Optimization of Protein-Ligand Poses. J. Chem. Inf. Model., doi: 10.1021/acs.jcim.0c01095.
4. Volkamer, A.; Griewel, A., Grombacher; Rarey. M.; (2010) Analyzing the topology of active sites: on the prediction of pockets and subpockets. J. Chem. Inf. Model., 50 (11), pp 2041-2052.</sup>

JAMDA