György G. Ferenczy

Thermodynamics guided lead discovery and optimization

The documented unfavorable changes of physicochemical properties during lead discovery and optimization prompted us to investigate the present practice of medicinal chemistry optimization from a thermodynamic perspective. Basic principles of binding thermodynamics suggest that discriminating between enthalpy-driven and entropy-driven optimizations could be beneficial. We hypothesize that entropy-driven optimizations might be responsible for the undesirable trend observed in physicochemical properties. Consequently, we suggest that enthalpy-driven optimizations are preferred because they provide better quality compounds. Monitoring binding thermodynamics during optimization programs initiated from thermodynamically characterized hits or leads, therefore, could improve the success of discovery programs. Here, we summarize common industry practices for tackling optimization challenges and review how the assessment of binding thermodynamics could support medicinal chemistry efforts.

The molecular structure of uracil: an electron diffraction study

Abstract The molecular structure of free uracil has been determined by electron diffraction. Planar or nearly planar models fit the experimental data. Differences between the four CN bond lengths as well as those between the two CO bond lengths were assumed from ab initio (4–21) calculations in the electron diffraction analysis. The following bond lengths (rg) and bond angles (rα) were determined; the estimated total errors are parenthesized as expressed in units of the last digit: r(CN)mean 1.399(6), r(CC) 1.462(8), r(CC) 1.343(24), r(CO)mean 1.212(3) A, C2N1C6 123.2(12), N3C4C5 115.5(18), C4C5C6 119.7(21), C5C6N1 122.1(22), N1C2O8 123.8(14), C5C4O10 124.3(20)°. The structural parameters by electron diffraction are in agreement with those by quantum chemical optimizations (ab initio 4–21, MINDO/3). The bond lengths indicate partial delocalization in the ring.

Enthalpic efficiency of ligand binding

The thermodynamics of ligand-protein binding has received much attention recently. In the present contribution we focus on the enthalpic component of binding. The dissociation constant, pK(d), was decomposed into enthalpic and entropic components (pK(d) = pK(H) + pK(S)), and pK(H), defined as pK(H) = -ΔH/(2.303·RT) was used to characterize the enthalpy contribution to binding. It was found that the maximal achievable pK(H) decreases with increasing molecular size. This is in contrast to maximal pK(d) that increases with molecular size until it achieves a plateau. Size-independent enthalpic efficiency (SIHE) was defined as SIHE = pK(H)/40·HA(0.3), with HA being the number of heavy atoms. SIHE allows a size unbiased comparative binding characterization of compounds. It can find use in hit and lead selection and also in monitoring optimization in drug discovery programs. The physical background of decreasing maximal pK(H) with molecular size is discussed, and its consequences to drug discovery are analyzed. It is concluded that the feasibility of simultaneous optimization of affinity and enthalpy diminishes with increasing molecular size. Consequently, binding thermodynamics considerations are to be applied primarily in hit prioritization and hit-to-lead optimization.

How are fragments optimized? A retrospective analysis of 145 fragment optimizations.

Fragment optimizations in nearly 150 fragment-based drug discovery programs reported in the literature during the past fifteen years were investigated. By analyzing physicochemical properties and ligand efficiency indices we found that biochemical detection methods yield hits with superior ligand efficiency and lipophilicity indices than do X-ray and NMR. These advantageous properties are partially preserved in the optimization since higher affinity starting points allow optimizations better balanced between affinity and physicochemical property improvements. Size independent ligand efficiency (SILE) and lipophilic indices (primarily LELP) were found to be appropriate metrics to monitor optimizations. Small and medium enterprises (SME) produce optimized compounds with better properties than do big pharma companies and universities. It appears that the use of target structural information is a major reason behind this finding. Structure-based optimization was also found to dominate successful fragment optimizations that result in clinical candidates. These observations provide optimization guidelines for fragment-based drug discovery programs.

Thermodynamics of Fragment Binding

The ligand binding pockets of proteins have preponderance of hydrophobic amino acids and are typically within the apolar interior of the protein; nevertheless, they are able to bind low complexity, polar, water-soluble fragments. In order to understand this phenomenon, we analyzed high resolution X-ray data of protein-ligand complexes from the Protein Data Bank and found that fragments bind to proteins with two near optimal geometry H-bonds on average. The linear extent of the fragment binding site was found not to be larger than 10 Å, and the H-bonding region was found to be restricted to about 5 Å on average. The number of conserved H-bonds in proteins cocrystallized with multiple different fragments is also near to 2. These fragment binding sites that are able to form limited number of strong H-bonds in a hydrophobic environment are identified as hot spots. An estimate of the free-energy gain of H-bond formation versus apolar desolvation supports that fragment sized compounds need H-bonds to achieve detectable binding. This suggests that fragment binding is mostly enthalpic that is in line with their observed binding thermodynamics documented in Isothermal Titration Calorimetry (ITC) data sets and gives a thermodynamic rationale for fragment based approaches. The binding of larger compounds tends to more rely on apolar desolvation with a corresponding increase of the entropy content of their binding free-energy. These findings explain the reported size-dependence of maximal available affinity and ligand efficiency both behaving differently in the small molecule region featured by strong H-bond formation and in the larger molecule region featured by apolar desolvation.

Design Principles for Fragment Libraries: Maximizing the Value of Learnings from Pharma Fragment-Based Drug Discovery (FBDD) Programs for Use in Academia

Fragment-based drug discovery (FBDD) is well suited for discovering both drug leads and chemical probes of protein function; it can cover broad swaths of chemical space and allows the use of creative chemistry. FBDD is widely implemented for lead discovery in industry but is sometimes used less systematically in academia. Design principles and implementation approaches for fragment libraries are continually evolving, and the lack of up-to-date guidance may prevent more effective application of FBDD in academia. This Perspective explores many of the theoretical, practical, and strategic considerations that occur within FBDD programs, including the optimal size, complexity, physicochemical profile, and shape profile of fragments in FBDD libraries, as well as compound storage, evaluation, and screening technologies. This compilation of industry experience in FBDD will hopefully be useful for those pursuing FBDD in academia.

Crystal and electronic structure of two polymorphic modifications of famotidine. An experimental and theoretical study

Abstract The antiulcer agent (i.e. a histamine H2 receptor antagonist), famotidine is known to crystallize in two polymorphic modifications. A comparison of the two forms based on single crystal X-ray data and on quantum chemical calculations is presented. Form A is constructed from molecules with higher internal energy. The crystal field together with the intermolecular hydrogen bond network causes important deformation of the molecular charge densities with respect to that of the free molecules. This charge density deformation is more pronounced in form A as it is manifested by an increased dipole moment and by enhanced electrostatic and polarisation interaction energies. Thus the excess internal energy of form A with respect to form B is largely counterbalanced by the increased intermolecular interaction energy in the former modification.

Methods for determining the reliability of semiempirical electrostatic potentials and potential derived charges

Abstract The reliability of the ZDO semiempirical molecular electrostatic potential (i.e. the potential determined without deorthogonalization) was investigated with regard to its ability to reproduce the negative potential well above aromatic systems and to generate potential derived atomic charges. The negative well above benzene and 5-hydroxytryptophan derivatives has indeed been observed; this is an important observation as this well plays an important role in many recognition processes, such as the clustering of aromatic rings within enzymes. The well may also be observed in classical molecular electrostatic potentials calculated using atomic charges derived from the semiempirical molecular electrostatic potential. The potential derived charges were assessed using an energy-based criterion which involves a comparison of the electrostatic interaction energy between two small molecules. The electrostatic interaction energy was determined for both a full quantum mechanical system, using the Morokuma-Kitaura energy decomposition analysis and for a quantum-classical hybrid system, using a simplified decomposition procedure. The semiempirical potential derived charges were found to be of similar quality to the ab initio potential derived charges and superior to ab initio Mulliken charges.

Towards improved force fields: III. Polarization through modified atomic charges

A method of modeling polarization by representing an atomic‐centered induced dipole as a set of induced charges on the atom and its immediate neighbors is presented. The method is based on earlier work on deriving atomic charges from a distributed multipole analysis (P. J. Winn et al., J Phys Chem A 1997, 101, 5437; G. G. Ferenczy, J Comput Chem 1991, 12, 913). The method has been applied to the water dimer, water trimers, formaldehyde–water complexes, methanol complexes, and DNA basepairs. It was found that the induced charges described the various cooperative and anticooperative hydrogen bonding systems well, both qualitatively and quantitatively (as compared with Hartree–Fock calculations). Importantly, it has been shown that, when an induced charge (or induced dipole) model is used for larger molecules, a correction term is required for the underlying electrostatics. ©1999 John Wiley & Sons, Inc. J Comput Chem 20: 704–712, 1999

Imidazo[1,2-c]quinazolines with lipid peroxidation inhibitory effect

Abstract A series of imidazo[1,2- c ]quinazolines of different lipophilic character was prepared. According to their antioxidant (cyclic voltammetry) properties they all should be potent inhibitors of lipid peroxidation. Under the given circumstances (NADPH-induced lipid peroxidation in rat brain microsomes and Fe 2+ -induced lipid peroxidation in rat brain homogenate), however, their lipid peroxidation inhibitory activity was strongly dependent on their lipophilicity.