BoBER: web interface to the base of bioisosterically exchangeable replacements
© The Author(s) 2017
Received: 5 June 2017
Accepted: 4 December 2017
Published: 12 December 2017
Bioisosterism and scaffold hopping are key concepts in the lead optimization stages of drug discovery [1, 2]. They can be defined as replacements of a part of a biologically active compound with a substructure that leads to a compound of the same or similar biological interaction. A bioisosteric replacement usually represents a functional group in a lead molecule that can be used in exchange of another functional group while the overall molecule retains similar non-covalent interactions towards a target. Bioisosteres are used to replace a functional group that is important for binding, but a new group in its place would improve the overall properties of a ligand, such as, lessen side-effects, improve pharmacokinetic properties, improve selectivity, simplify synthetic routes, increase metabolic stability or help avoid patent related issues . Moreover, scaffold hopping can be interpreted as a subclass of bioisosteric replacements, where a larger part of a ligand—the core scaffold—is replaced. This core scaffold is important due to formation of direct interactions with the target or alternatively, it may provide appropriate scaffolding that spatially arranges functional groups so that they are able to form the necessary interactions.
In the past, bioisosteric and scaffold hopping replacements were obtained experimentally using the trial-and-error approach, resulting in today’s extensive literature available to the medicinal chemistry community . The collected data can be used to create extensive digitized databases of bioisosteric replacements. BIOSTER , for instance, contains bioisosteric transformations collected from literature published in the last 40 years. ChEMBL  is a public domain database of over 1.5 million small molecules paired with associated bioactivity data mined from medicinal chemistry literature. The database enables identification of experimentally observed molecular substructures that exhibit bioisosteric characteristics. Based on these data, the Matched Molecular Pair (MMP) approach  enables the identification of molecules in ChEMBL that differ only in one functional group. This allows for the analysis of potential changes in biological properties that may be affected by such transformation. The MMP has been made freely available for non-commercial use on-line as the Swiss-Bioisostere database .
Rapidly growing freely available structural databases such as the Protein Data Bank (PDB)  offer another opportunity to obtain new bioisosteric and scaffold hopping replacements in a rigorous and automated way. Kennewell et al.  developed a method for comparison and superimposition of all holo proteins in the PDB based on protein backbone atoms, which allows ligands to be transposed between protein binding sites based on protein structure superimpositions. Fragments occupying the same geometric space are considered as potentially bioisosterically replaceable. Another method, KRIPO  quantifies similarities of binding site subpockets based on optimized pharmacophore based fingerprints, and enables both intra- and inter-family comparisons of proteins. Using this method, the complete PDB was converted into a database comprising of around 300,000 fingerprints of local binding sites together with their associated ligand fragments. The method enables the identification of bioisosteric replacements for ligand substructures based on local binding site similarities independently from the protein sequence or overall protein folding. Khashan  developed FragVLib, a virtual library of fragments which enables finding bioisosteric replacements based on a subgraph matching tool that identifies similar binding pockets according to their 3D structures and chemical composition. Further, sc-PDB-Frag  is an approach that considers bioisosteric searches with no a priori knowledge of either ligand (fragment) or protein (binding site) similarities. This can be achieved by converting protein–ligand interaction patterns to 1D or 3D graphs. Bioisosteres are then defined as any pair of ligands that share similar interaction patterns with their native target protein. Because the selection is directly based on protein–ligand interactions it does not require any pairwise similarity calculation between either ligands or binding sites. To extend the repertoire of methods for obtaining bioisosteric and scaffold replacements, we developed a freely available pre-calculated database of bioisostere replaceable fragments obtained with a rigorous all-against-all PDB local binding site alignments. Additionally, we developed a corresponding web interface, which enables easy acquisition of appropriate fragment replacements.
In this work we present Base of Bioisosterically Exchangeable Replacements (BoBER), a new web server for identification of bioisosteric and scaffold hopping replacements based on our PDB mining approach . In this approach, bioisosteric replacements are identified using local binding site alignment algorithm ProBiS [15–19], which enables identification of locally similar binding sites irrespective of proteins’ folds or amino acid sequences. It seeks for similar local spatial arrangements of physico-chemically similar surface functional groups in binding sites, enabling the detection of replaceable fragment pairs between distantly related protein structures. ProBiS was used to superimpose holo binding sites from the entire PDB, and pairs of bioisosterically replaceable fragments were collected in the BoBER database . The advantage of our method, which takes into account local neighborhood of fragments, is that it enables the distinction between different binding pockets in proteins with similar overall sequence identity, while recognizing similar binding pockets in proteins with very different sequences. This assures that identified bioisosteres will form similar interactions in the new environment of a possibly unrelated protein, while reducing the number of obtained bioisosteres that would not be able to form an appropriate interaction pattern with the protein’s binding site. BoBER web server is interactive and freely available at http://bober.insilab.org, and will benefit medicinal chemists in the lead optimization stage of the drug design process. The web server was tested in the Chrome and Firefox web browsers.
Design and implementation
Generation of database of bioisosteric replacements
BoBER web server
The details of the Loose and Rigorous filtering options
Bioisosteric fragment pair is from the same SCOP family
Both (inter- and intra-family)
Interchangeability of similar join atom types
Non-interchangeable join atom types
Consideration of join atoms
Use structures with common core as queries (ignore join atoms)
Use specific structure as query (consider join atoms)
The second query option is to Draw the core structure (without join atoms) of the fragment using the JSME. BoBER will output query fragments contained within the BoBER database that have the same substructure present within its core structure. Again, this fragment can be selected to find its bioisoteric replacements.
The third option is to specify the properties for the query fragment, for which we wish to find replacements. These properties include simple descriptors, such as the number of heavy atoms a fragment contains, the number of potential hydrogen bond donors and acceptors, number of atoms in rings or the number of core and/or join atoms.
In all three options, upon clicking the Submit query button, a Fragment selection panel is displayed, containing query fragments meeting the chosen criteria. A query fragment for which we wish to display its possible replacements can be selected, after which the user can define the Overall Hausdorff distance cutoff, which defines the extent of overlap between all of the database fragment atoms and the query ones. Lower HD requires better spatial overlap of corresponding ligand atoms in the superimposed binding sites. By visual inspection of a large number of pairs, we set the default value of Overall HD to 1.50 Å, which is loosely the maximum at which fragments can still be considered as replaceable. Fragments can also be filtered based on the superimposed proteins’ SCOP families . Choosing the Rigorous filtering radio button selects the Intra-family option which limits the best fragment pairs (lowest HD) to the part of the database obtained from superimposed protein structures belonging to the same SCOP family. The Inter-family option, available in the Custom options, outputs fragments that originate from proteins that are of different SCOP families or when one or both of the protein families are unspecified. When selecting Both we get the best fragment pairs independently of this criteria. The Both option is selected as default when using the Loose filtering radio button. The protein-family related criteria refers to the superimposed proteins within the BoBER database independent of the target family to which we want our changed ligand to bind, as the current version of BoBER does not yet support this specification.
Output of replaceable fragments
After query fragment selection and submission of HD-based criteria, a new Results tab opens. This tab contains a table, which displays the 2D structures of the query and reference fragment pairs and their corresponding HDs. Join atoms that are in spatial overlap between two fragments, that is corresponding join atoms, are shown with the same highlight colors. When using the Use structures with common core as queries option (part of the Loose filtering option), the bioisosteric fragments are shown reduced to their core structure. If the Use specific structure as query radio button has been selected, then the user can sort the bioisosteric pairs based on three different HD values (Overall, Core or H-bonding HD) in ascending or descending order. When Use structures with common core as queries is selected, the sorting can be done only based on the Core HD as the other HDs that are based on join atoms are not relevant in this case. By clicking on the structure image of a fragment, a new tab opens in the browser with the PDB web page of the protein–ligand complex from which the fragment was obtained.
Comparison of BoBER performance
We compared BoBER with SwissBioisostere database , which is another freely available tool for obtaining bioisosteric replacements. The two approaches differ significantly, therefore we expected different results. We queried the SwissBioisostere database with the three ring fragments (Table 2) previously used in BoBER, where we replaced join atoms with the R-groups indicating attachment points. For fragment 1, we obtained six suggested replacements, none of which exactly matched the bioisosteric fragment obtained with BoBER. However, the cyclopentathiophene-carbonitrile fragment suggested by the SwissBioisostere is similar to the fragment suggested by BoBER (first row, Table 2). For fragment 2, we obtained many potential replacements using both servers, and several similar bioisosteres were found. For example, compare bioisosteres of fragment 2 obtained by BoBER and SwissBioisostere (second row, Table 2), where a thiophene in BoBER fragment is replaced with a benzene. It is well known that thiophene is a bioisosteric replacement for benzene. Finally, no similar bioisostere was found among the results obtained with SwissBioisostere for BoBER’s bioisostere of fragment 3 in which an acidic moiety is bound to a furan ring. There seems to be some overlap between the bioisosteres found by BoBER and SwissBioisostere. BoBER also finds different replacements that might not have been included in SwissBioisostere as of yet.
We developed a new web server BoBER that enables the prediction of bioisosteric replacements given a query fragment or query small molecule as input based on our knowledge-based method that uses binding sites superimposition to identify possible bioisosteric and scaffold hopping replacements from existing ligands. The predicted bioisosteric replacements are obtained after the ProBiS-based superimposition of existing PDB crystal holo protein structures, which makes us confident that a significant proportion of newly generated compounds will retain activity. The database of bioisosteric pairs obtained with this method is implemented in a freely available web server BoBER, which enables medicinal chemists to quickly search and get new and unique ideas about possible bioisosteric or scaffold hoping replacements that could be used to improve hit or lead structures. We showed how the BoBER web server could be used on an inhibitor of MurF enzyme. In the future, the BoBER approach will be implemented in the ligand-based virtual screening software LiSiCA  to enable searching databases for similar ligands not only on the basis of atom type similarity but also based on possible bioisosteric or scaffold hopping replacements.
The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. All authors read and approved the final manuscript.
Financial support through Grants P1-0002, J1-6736, L7-8269, J1-6743 and L1-6745 of the Ministry of Higher Education, Science and Technology of Slovenia, and Slovenian Research Agency is acknowledged. All authors also gratefully acknowledge the support of NVIDIA Corporation with the donation of the Tesla K40 GPU used for this research.
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Langdon SR, Ertl P, Brown N (2010) Bioisosteric replacement and scaffold hopping in lead generation and optimization. Mol Inform 29:366–385View ArticleGoogle Scholar
- Brown N (2014) Bioisosteres and scaffold hopping in medicinal chemistry. Mol Inform 33:458–462View ArticleGoogle Scholar
- Sethy SP, Meher CP, Biswal S, Sahoo U, Patro SK (2013) The role of bioisosterism in molecular modification and drug design: a review. Asian J Pharm Sci Res 3:61–87Google Scholar
- Brown N (2012) Bioisosteres in medicinal chemistry. Wiley, HobokenView ArticleGoogle Scholar
- Ujváry I, Hayward J (2012) Bioster: a database of bioisosteres and bioanalogues. In: Brown N (ed) Bioisosteres medicinal chemistry. Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, pp 53–74View ArticleGoogle Scholar
- Gaulton A, Bellis LJ, Bento AP et al (2012) ChEMBL: a large-scale bioactivity database for drug discovery. Nucleic Acids Res 40:D1100–D1107View ArticleGoogle Scholar
- Dossetter AG, Griffen EJ, Leach AG (2013) Matched molecular pair analysis in drug discovery. Drug Discov Today 18:724–731View ArticleGoogle Scholar
- Wirth M, Zoete V, Michielin O, Sauer WHB (2013) SwissBioisostere: a database of molecular replacements for ligand design. Nucleic Acids Res 41:D1137–D1143View ArticleGoogle Scholar
- Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE (2000) The protein data bank. Nucleic Acids Res 28:235–242View ArticleGoogle Scholar
- Kennewell EA, Willett P, Ducrot P, Luttmann C (2006) Identification of target-specific bioisosteric fragments from ligand–protein crystallographic data. J Comput Aided Mol Des 20:385–394View ArticleGoogle Scholar
- Wood DJ, de Vlieg J, Wagener M, Ritschel T (2012) Pharmacophore fingerprint-based approach to binding site subpocket similarity and its application to bioisostere replacement. J Chem Inf Model 52:2031–2043View ArticleGoogle Scholar
- Khashan R (2012) FragVLib a free database mining software for generating “Fragment-based Virtual Library” using pocket similarity search of ligand-receptor complexes. J Cheminformatics 4:18View ArticleGoogle Scholar
- Desaphy J, Rognan D (2014) sc-PDB-Frag: a database of protein-ligand interaction patterns for bioisosteric replacements. J Chem Inf Model 54:1908–1918View ArticleGoogle Scholar
- Lešnik Samo, Konc Janez, Janežič Dušanka (2016) Scaffold hopping and bioisosteric replacements based on binding site alignments. Croat Chem Acta. https://doi.org/10.5562/cca2993 Google Scholar
- Konc J, Janežič D (2010) ProBiS algorithm for detection of structurally similar protein binding sites by local structural alignment. Bioinformatics 26:1160–1168View ArticleGoogle Scholar
- Konc J, Janežič D (2010) ProBiS: a web server for detection of structurally similar protein binding sites. Nucleic Acids Res 38:W436–W440View ArticleGoogle Scholar
- Konc J, Janežič D (2012) ProBiS-2012: web server and web services for detection of structurally similar binding sites in proteins. Nucleic Acids Res 40:W214–W221View ArticleGoogle Scholar
- Konc J, Janežič D (2014) ProBiS-ligands: a web server for prediction of ligands by examination of protein binding sites. Nucleic Acids Res 42:W215–W220View ArticleGoogle Scholar
- Konc J, Česnik T, Konc JT, Penca M, Janežič D (2012) ProBiS-database: precalculated binding site similarities and local pairwise alignments of PDB structures. J Chem Inf Model 52:604–612View ArticleGoogle Scholar
- Fox NK, Brenner SE, Chandonia J-M (2015) The value of protein structure classification information—surveying the scientific literature. Proteins Struct Funct Bioinform 83:2025–2038View ArticleGoogle Scholar
- Longenecker KL, Stamper GF, Hajduk PJ et al (2005) Structure of MurF from Streptococcus pneumoniae co-crystallized with a small molecule inhibitor exhibits interdomain closure. Protein Sci 14:3039–3047View ArticleGoogle Scholar
- Turk S, Hrast M, Sosič I, Barreteau H, Mengin-Lecreulx D, Blanot D, Gobec S (2013) Biochemical characterization of MurF from Streptococcus pneumoniae and the identification of a new MurF inhibitor through ligand-based virtual screening. Acta Chim Slov 60:294–299Google Scholar
- Hrast M, Turk S, Sosič I et al (2013) Structure–activity relationships of new cyanothiophene inhibitors of the essential peptidoglycan biosynthesis enzyme MurF. Eur J Med Chem 66:32–45View ArticleGoogle Scholar
- Hrast M, Anderluh M, Knez D, Randall CP, Barreteau H, O’Neill AJ, Blanot D, Gobec S (2014) Design, synthesis and evaluation of second generation MurF inhibitors based on a cyanothiophene scaffold. Eur J Med Chem 73:83–96View ArticleGoogle Scholar
- Comess KM, Schurdak ME, Voorbach MJ et al (2006) An ultraefficient affinity-based high-throughout screening process: application to bacterial cell wall biosynthesis enzyme MurF. J Biomol Screen 11:743–754View ArticleGoogle Scholar
- Lešnik S, Štular T, Brus B, Knez D, Gobec S, Janežič D, Konc J (2015) LiSiCA: a software for ligand-based virtual screening and its application for the discovery of butyrylcholinesterase inhibitors. J Chem Inf Model 55:1521–1528View ArticleGoogle Scholar