MolPort 3D molecular similarity search service

A frequent problem in drug discovery is that an interesting lead has been identified, but it is unsuitable for optimization

Most problems:

  • Difficult synthetic chemistry
  • Limited scope for analogues
  • Undesirable ADMET characteristics
  • IP issues
  • and more...
3D similarity search service

Jump to new series with
our 3D similarity search service

Contact MolPort

Background:

Molecular interactions between bioactive compounds and their target proteins are primarily driven by complementarity of shape and electrostatic potential between the binding partners with high complementarity leading to tight binding.

Example of protein-ligand complementarity. A) Hydrogen bond (yellow dashed lines) complementarity between thrombin and inhibitor. B) Shape complementarity.
Example of protein-ligand complementarity. A) Hydrogen bond (yellow dashed lines) complementarity between thrombin and inhibitor. B) Shape complementarity.

As a consequence, compounds with high shape and electrostatic similarity to an active “lead molecule” have an increased probability of sharing the biological activity profile of the lead compound. Similar shapes can occur in compounds with very different chemical structures, and in this way 3D searching can identify novel compounds, so called “scaffold hopping”.

3D molecular similarity method:

Our 3D molecular similarity methods quantify the overall shape and electrostatic similarity of a query to a database molecule. They return a rank-ordered list of the database, with a score of 1.0 indicting an identical structure and lower scores representing less similar compounds. A family of similarity measures are available. These differ in the details of the similarity measure. It is also possible to vary the relative contribution of shape and electrostatics, depending on the nature of the query and target sites.

Example: Zyprexa binds as an antagonist/inverse agonist at multiple G-protein coupled receptors:

Receptor Ki (nM) Result
dopamine D1 31 Zyprexa binds as an antagonist/inverse agonist at multiple G-protein coupled receptors
dopamine D2 11
dopamine D4 27
serotonin 5-HT2A 4
serotonin 5-HT2C 11
serotonin 5-HT3 57
muscarinic M1 26
adrenergic alpha1 19
histamine H1 7

To demonstrate the effectiveness of the technique, one needs to search a database of compounds with known activity. In this case we use the ChEMBL database. ChEMBL is a curated chemical database of bioactive molecules with drug-like properties maintained by the European Bioinformatics Institute (EBI), of the European Molecular Biology Laboratory (EMBL).

Top scoring hit compounds from searching the ChEMBL database using Zyprexa as a query compound
Top scoring hit compounds from searching the ChEMBL database using Zyprexa as a query compound

Some of the top scoring hits are clearly close structural analogues of Zyprexa, but others show little similarity at the 2D level.

The hits share common biological activities with Zyprexa, even when 2D similarity is low
ChEMBL Hit ChEMBL Targets 3D Similarity 2D Similarity
(MACCS Tanimoto)
CHEMBL273786 D1-D5,Muscarinic 0.985 0.847
CHEMBL14605 D1-D5,Muscarinic 0.958 0.847
CHEMBL327197 5HT-1-4 0.936 0.750
CHEMBL339438 Muscarinic M1, M2 0.934 0.530
CHEMBL124886 5HT-3a, 5HT-3b 0.932 0.783
CHEMBL341474 5HT-4 0.932 0.700
CHEMBL130978 5HT-1, 5HT-2, D2 0.927 0.576
CHEMBL336721 5HT-1, 5HT-2, D2 0.915 0.567
CHEMBL339852 5HT-1, 5HT-2, D 0.910 0.549
CHEMBL1093138 H4 0.910 0.682
CHEMBL132993 5HT-1, 5HT-2, D 0.908 0.493
CHEMBL128041 5HT-1, 5HT-2, D 0.907 0.493
CHEMBL327198 5HT-1a, 5HT-2a 0.904 0.606
CHEMBL541744 PARP1 0.899 0.585
CHEMBL107903 5HT-1A, D2 0.898 0.618

The majority of hits share biological activities with Zyprexa.

Superposition of Zyprexa (magenta) and structurally
distinct, but similarly shaped, hit compound.
Superposition of Zyprexa (magenta) and structurally <br />
		distinct, but similarly shaped, hit compound

Further information on the methods and example applications can be found in the references provided below.

Typical Project Workflow

Compounds provided by the client will be used as search queries, with a general protocol as follows:

  1. If the bioactive conformation of the query compound is known, for example from a crystal structure, then this will be used as the basis for the search. If the bioactive conformation is unknown then a representative set of low-energy conformations will be generated using molecular mechanics.
  2. These conformational models will be used as input to the query methods outlined above. Several searches will be run in each case to cover a range of similarity metrics to ensure a thorough search.
  3. The database to be searched consists of 3D models of commercially available screening compounds. The chemical structure of each compound is stored in the database, together with physicochemical properties such as molecular weight, lipophilicity, number of rotatable bonds, etc. For each drug-like compound, full three-dimensional conformational models, and shape and property descriptors have been created. The physicochemical properties can be used as filters, for example to focus on compounds that could be orally bioavailable or cross the blood-brain barrier.
  4. The resultant “hitlists” will be post-processed to remove duplicates, exclude undesirable compounds and ensure chemical diversity in the hits.
  5. Final selection will be made from an integrated combined list based on availability, price and logistical considerations, with the aim of obtaining a final list size of about 100 compounds (or as specified by the client).

Jump to new series with
our 3D similarity search service

Contact MolPort

References

Exploration of Piperidinols as Potential Antitubercular Agents. Abuhammad A, Fullam E, Bhakta S, Russell AJ, Morris GM, Finn PW and Sim E. Molecules 19, 16274-16290; doi:10.3390/molecules191016274 (2014).

Comparison of ultra-fast 2D and 3D ligand and target descriptors for side-effect prediction and network analysis in polypharmacology. Cortes-Cabrera A, Morris GM, Finn PW, Morreale A, Gago F, British Journal of Pharmacology, Br. J. Pharmacol. 170, 557-67 (2013).

Shape-based similarity searching in chemical databases. Finn PW, Morris GM, WIREs Comput Mol Sci. 2013, 3: 226-241. doi: 10.1002/wcms.1128 (2013),

Improving the accuracy of ultrafast ligand-based screening: incorporating lipophilicity into ElectroShape as an extra dimension. Armstrong MS, Finn PW, Morris GM, Richards WG. J Comput Aided Mol Des. 25, 785-790 (2011).

ElectroShape: fast molecular similarity calculations incorporating shape, chirality and electrostatics. Armstrong MS, Morris GM, Finn PW, Sharma R, Moretti L, Cooper RI, Richards WG. Journal of Computer- Aided Drug Design, 24, 789-801 (2010).

Molecular similarity including chirality. Armstrong MS, Morris GM, Finn PW, Sharma R and Richards WG. Journal of Molecular Graphics and Modelling 28, 368-370 (2009).

Radioreceptor binding profile of the atypical antipsychotic olanzapine. Bymaster FP, Calligaro DO, Falcone JF, Marsh RD, Moore NA, Tye NC, Seeman P, Wong DT. Neuropsychopharmacology 14, 87-96 (1996).

ChEMBL: a large-scale bioactivity database for drug discovery Gaulton A; et al. Nucleic Acids Research 40: D1100–7 (2011).

2017 MolPort, v2.24, release date 09-10-2017 13:00 (+0200). All rights reserved.