Main Content Region

Drug Discovery - A Rational Approach

In-silico high throughput screening

Dr. David Lloyd & Dr. Darren Fayne, Molecular Design Group, School of Biochemistry, Trinity College Dublin

The discovery of new medicines is a complex process – hugely expensive in terms of money and time. Advances in computation, biology and chemistry now allow scientists to model and understand drug targets like never before. In the same way that we are familiar with the use of computer imaging in the design cars or buildings – cancer, Alzheimer’s, malaria – these diseases all have specific biomolecular mechanisms – Achilles’ heels which we can first model and then exploit to provide therapies and drug candidates for the future. Adopting a multi-disciplinary approach to rationally design, drugs tailored to specific diseases will facilitate the development of therapies effectively and quickly, giving us medicines whose properties are better understood, have less side effects and which come to market within a reasonable time. Underpinning research and development with robust enabling technologies makes practical sense – productivity is increased, processes are streamlined, resultant products are of better quality. The pharmaceutical industry has long recognised this edict and has introduced cutting-edge technologies where available, to facilitate the drug discovery process, moving from target to hit to lead as quickly as possible so as to reduce the irrefutably enormous costs of developing and delivering a viable therapeutic candidate. The use of computational software in drug discovery is now growing in popularity; validation cases and reviews have recently been published highlighting its utility in the discovery process, and much effort has been invested in the formulation of in silico screening routines to complement traditional wet lab in vitro screening. ‘Scaffold-hopping’ or chemotype switching is a relatively new technique in computational biology and drug design where the pharmacologically relevant features of an active ligand are used for in silico virtual screening to identify alternate chemotypes with equivalent potency and mechanism of action [1].

The Molecular Design Group (MDG) in Trinity College Dublin is Ireland's largest and most successful rational drug discovery research group. Dr David Lloyd is the group leader, with 2 postdocs (Dr Darren Fayne and Dr Andrew Knox) and 5 PhD students. The MDG focuses on the integration of in silico and traditional wet lab drug discovery techniques, applying chemical biology and rational drug design strategies to a number of key therapeutic targets. Our core computational efforts are in the areas of virtual high-throughput screening and structure- and ligand-based design approaches, where in silico predictions are validated using wet lab chemistries and in vitro assay.

The primary interest of the group lies in the application and development of computational modelling to augment the drug design process. The MDG utilise structure and ligand-based drug design (SBDD and LBDD) technologies together with virtual screening approaches to rationally tailor small molecules for the modulation of therapeutically relevant biological macromolecules - enzymes and receptors [2]. LBDD uses physical and chemical traits of a known ligand to identify similar compounds that might also interact with the biological target. Likewise, SBDD software can perform virtual screening docking calculations on a database of molecules and output those judged most likely to bind to the protein of interest (Figure 1).

Figure 1
Figure 1. X-ray crystal structure of antagonist Raloxifene - ERα (PDB ID: 1ERE) Purple: protein lone pair electrons. Green: hydrophobic regions.

All designs are validated through wet chemical exemplification and biological assay. With a specific disease focus on oncology, the MDG targets the nuclear hormone receptor family to identify modulators of transcriptional control and are engaged on several funded projects on apoptosis - examining (de)stabilisers of tubulin and the cell cycle. The MDG has an established reputation in rational computational SBDD and LBDD of chemotherapeutic drug candidates, particularly in the production of ligands targeting the estrogen receptor (ER), and are also studying peroxisome proliferator-activated receptors (PPAR), glucocorticoid receptors (GR) and tubulin [3-5]. Our interest in oncology targets is spreading to include other protein superfamilies to be targeted by novel chemotherapeutic compounds.

A wide variety of 2D and 3D descriptors can be calculated for ligands which are active on a particular target. This information can be utilised to determine where in chemical space the ligands reside. A new molecule is more likely to be active if it is also located in the same region of chemical space as known active molecules [6]. This is an important step in the computational search for new therapeutic modulators as chemical space has been predicted to consist of in the region of 1060 molecules which necessitates focusing our research toward preferential regions of space (Figure 2).

Figure 2. 3D illustration of where molecules reside in chemical space. Green spheres – all cancer actives, red cloud – cancer medicinal chemistry space, blue cloud – traditional ‘drug-like’ space.

Once a specific region of space has been identified a modified version of the Lipinski rule of five is generally applied to the virtual molecules so as to increase the likelihood that they will be orally bioavailable. The accessibility of a tablet application of the drug increases patient compliance and, thus, the effectiveness of the therapy. Before virtually screening the potential active molecules through the 3D structure of their target protein it is important to accurately model the inherent flexibility of the molecules. Sampling of different attainable conformations of each ligand needs to be performed and we have detailed a faster and more efficient procedure for examining a greater amount of conformational space (Figure 3) [7, 8].

Figure 3. Conformational space. Violet: 280 conformers produced from the canonical SMILES string of a ligand. Green: conformers generated from 8 permutation of the same string. Red: the crystal structure of the ligand.

The computational expertise developed in the study of these proteins is readily applicable to other identified targets. Key to the successful translation of our computational design efforts to potential therapeutics is an integrated approach to discovery science, closely linked to clinical and basic research.


This research was funded by HEA, PRTLI, IITAC, Health Research Board Ireland and Science Foundation Ireland.


  1. Lloyd DG, Buenemann CL, Todorov NP, Manallack DT, Dean PM. Scaffold hopping in de novo design - ligand generation in the absence of receptor information Journal of Medicinal Chemistry 2004 Jan 29;47(3):493-96
  2. Dean PM, Lloyd DG, Todorov NP. De novo drug design: integration of structure-based and ligand-based methods. Current Opinion in Drug Discovery and Development 2004 7(3):347-353
  3. Knox AJ, Meegan MJ, Lloyd DG. Estrogen receptors: molecular interactions, virtual screening and future prospects. Curr Top Med Chem. 2006;6(3):217-43.
  4. Lloyd DG, Smith HM, O'Sullivan T, Knox AS, Zisterer DM, Meegan MJ. Antiestrogenically active 2-benzyl-1,1-diarylbut-2-enes: Synthesis, Structure-Activity Relationships and Molecular Modelling Study for Flexible Estrogen Receptor Antagonists. Medicinal Chemistry, 2006, in press
  5. Lloyd DG, Hughes RB, Zisterer DM, Williams DC, Fattorusso C, Catalanotti B, Campiani G, Meegan MJ. Benzoxepin Derived Estrogen Receptor Modulators: A Novel Molecular Scaffold for the Estrogen Receptor. Journal of Medicinal Chemistry 2004 47(23):5612-5615
  6. Lloyd DG, Golfis G, Knox AJ, Fayne D, Meegan MJ, Oprea TI. Oncology exploration: charting cancer medicinal chemistry space. Drug Discov Today. 2006 Feb;11(3-4):149-59
  7. Knox AJ, Meegan MJ, Carta G, Lloyd DG. Considerations in compound database preparation--"hidden" impact on virtual screening results. J Chem Inf Model. 2005 Nov-Dec;45(6):1908-19
  8. Carta G, Onnis V, Knox AJ, Fayne D, Lloyd DG. Permuting input for more effective sampling of 3D conformer space. J Comput Aided Mol Des. 2006, Epub.