Amiram Goldblum

Liposome drugs' loading efficiency: a working model based on loading conditions and drug's physicochemical properties.

Remote loading of liposomes by transmembrane gradients is one of the best approaches for achieving the high enough drug level per liposome required for the liposomal drug to be therapeutically efficacious. This breakthrough, which enabled the approval and clinical use of nanoliposomal drugs such as Doxil, has not been paralleled by an in-depth understanding that allows predicting loading efficiency of drugs. Here we describe how applying data-mining algorithms on a data bank based on Barenholzs laboratorys 15 years of liposome research experience on remote loading of 9 different drugs enabled us to build a model that relates drug physicochemical properties and loading conditions to loading efficiency. This model enables choosing candidate molecules for remote loading and optimizing loading conditions according to logical considerations. The model should also help in designing pro-drugs suitable for remote loading. Our approach is expected to improve and accelerate development of liposomal formulations for clinical applications.

Understanding drug‐likeness

‘Drug‐likeness’, a qualitative property of chemicals assigned by experts committee vote, is widely integrated into the early stages of lead and drug discovery. Its conceptual evolution paralleled work related to Pfizers ‘rule of five’ and lead‐likeness, and is placed within this framework. The discrimination between ‘drugs’ (represented by a collection of pharmaceutically relevant small molecules, some of which are marketed drugs) and ‘nondrugs’ (typically, chemical reagents) is possible using a wide variety of statistical tools and chemical descriptor systems. Here we summarize 18 papers focused on drug‐likeness, and provide a comprehensive overview of progress in the field. Tools that estimate drug‐likeness are valuable in the early stages of lead discovery, and can be used to filter out compounds with undesirable properties from screening libraries and to prioritize hits from primary screens. As the goal is, most often, to develop orally available drugs, it is also useful to optimize drug‐like pharmacokinetic properties. We examine tools that evaluate drug‐likeness and some of their shortcomings, challenges facing these tools, and address the following issues: What is the definition of drug‐likeness and how can it be utilized to reduce attrition rate in drug discovery? How difficult is it to distinguish drugs from nondrugs? Are nondrug datasets reliable? Can we estimate oral drug‐likeness? We discuss a drug‐like filter and recent advances in the prediction of oral drug‐likeness. The heuristic aspect of drug‐likeness is also addressed.

Quantitative structure-property relationship modeling of remote liposome loading of drugs.

Remote loading of liposomes by trans-membrane gradients is used to achieve therapeutically efficacious intra-liposome concentrations of drugs. We have developed Quantitative Structure Property Relationship (QSPR) models of remote liposome loading for a data set including 60 drugs studied in 366 loading experiments internally or elsewhere. Both experimental conditions and computed chemical descriptors were employed as independent variables to predict the initial drug/lipid ratio (D/L) required to achieve high loading efficiency. Both binary (to distinguish high vs. low initial D/L) and continuous (to predict real D/L values) models were generated using advanced machine learning approaches and 5-fold external validation. The external prediction accuracy for binary models was as high as 91-96%; for continuous models the mean coefficient R(2) for regression between predicted versus observed values was 0.76-0.79. We conclude that QSPR models can be used to identify candidate drugs expected to have high remote loading capacity while simultaneously optimizing the design of formulation experiments.

A stochastic algorithm for global optimization and for best populations: A test case of side chains in proteins

The problem of global optimization is pivotal in a variety of scientific fields. Here, we present a robust stochastic search method that is able to find the global minimum for a given cost function, as well as, in most cases, any number of best solutions for very large combinatorial “explosive” systems. The algorithm iteratively eliminates variable values that contribute consistently to the highest end of a cost functions spectrum of values for the full system. Values that have not been eliminated are retained for a full, exhaustive search, allowing the creation of an ordered population of best solutions, which includes the global minimum. We demonstrate the ability of the algorithm to explore the conformational space of side chains in eight proteins, with 54 to 263 residues, to reproduce a population of their low energy conformations. The 1,000 lowest energy solutions are identical in the stochastic (with two different seed numbers) and full, exhaustive searches for six of eight proteins. The others retain the lowest 141 and 213 (of 1,000) conformations, depending on the seed number, and the maximal difference between stochastic and exhaustive is only about 0.15 Kcal/mol. The energy gap between the lowest and highest of the 1,000 low-energy conformers in eight proteins is between 0.55 and 3.64 Kcal/mol. This algorithm offers real opportunities for solving problems of high complexity in structural biology and in other fields of science and technology.

High quality binding modes in docking ligands to proteins

Multiple near‐optimal conformations of protein‐ligand complexes provide a better chance for accurate representation of biomolecular interactions, compared with a single structure. We present ISE‐dock – a docking program which is based on the iterative stochastic elimination (ISE) algorithm. ISE eliminates values that consistently lead to the worst results, thus optimizing the search for docking poses. It constructs large sets of such poses with no additional computational cost compared with single poses. ISE‐dock is validated using 81 protein‐ligand complexes from the PDB and its performance was compared with those of Glide, GOLD, and AutoDock. ISE‐dock has a better chance than the other three to find more than 60% top single poses under RMSD = 2.0 Å and more than 80% under RMSD = 3.0 Å from experimental. ISE alone produced at least one 3.0 Å or better solutions among the top 20 poses in the entire test set. In 98% of the examined molecules, ISE produced solutions that are closer than 2.0 Å from experimental. Paired t‐tests (PTT) were used throughout to assess the significance of comparisons between the performances of the different programs. ISE‐dock provides more than 100‐fold docking solutions in a similar time frame as LGA in AutoDock. We demonstrate the usefulness of the large near optimal populations of ligand poses by showing a correlation between the docking results and experiments that support multiple binding modes in p38 MAP kinase (Pargellis et al., Nat Struct Biol 2002;9:268‐272] and in Human Transthyretin (Hamilton, Benson, Cell Mol Life Sci 2001;58:1491‐1521). Proteins 2008.

Refined structure of bovine carboxypeptidase A at 1.25 A resolution.

The crystal structure of the bovine zinc metalloproteinase carboxypeptidase A (CPA) has been refined to 1.25 A resolution based on room-temperature X-ray synchrotron data. The significantly improved structure of CPA at this resolution (anisotropic temperature factors, R factor = 10.4%, R(free) = 14.5%) allowed the modelling of conformational disorders of side chains, improved the description of the protein solvent network (375 water molecules) and provided a more accurate picture of the interactions between the active-site zinc and its ligands. The calculation of standard uncertainties in individual atom positions of the refined model of CPA allowed the deduction of the protonation state of some key residues in the active site and confirmed that Glu72 and Glu270 are negatively charged in the resting state of the enzyme at pH 7.5. These results were further validated by theoretical calculations that showed significant reduction of the pK(a) of these side chains relative to solution values. The distance between the zinc-bound solvent molecule and the metal ion is strongly suggestive of a neutral water molecule and not a hydroxide ion in the resting state of the enzyme. These findings could support both the general acid/general base mechanism, as well as the anhydride mechanism suggested for CPA.

Acylphosphonic acids and methyl hydrogen acylphosphonates: physical and chemical properties and theoretical calculations

Acylphosphonic acids (5) and methyl hydrogen acylphosphonates (3) were synthesized by di- and mono-demethylation of dimethyl acylphosphonates (1). Spectroscopic data (i.r., 31P and 1H n.mr.) are reported for these types of compounds for the first time. Examination of their hydrolytic stability under acidic and basic conditions revealed that except for methyl hydrogen acylphosphonates (3) that are unstable under highly alkaline pH conditions, the C–P bond in these types of compounds is stable in most cases. Nucleophilic reagents, e.g. amines, borohydride, or hydroxylamine react with the carbonyl group of ionized acylphosphonates with the preservation of the C–P bond, to yield α-imino-, α-hydroxy-, or α-oxyimino-alkylphosphononate anions, respectively. Semi-empirical quantum mechanical (MNDO/H) calculations were performed on benzoylphosphonic acid (5c) and on the esters and anions derived from it, as representatives of their classes, in order to assess bond lengths and preferred conformations, and to estimate charges on the carbonyl and phosphoryl groups. Calculations show that for both neutral and ionized (anions) compounds free rotation around the C–P bond is expected due to the low energy barriers.

A novel energy-based stochastic method for positioning polar protons in protein structures from X-rays

A novel automated method for the optimal placement of polar hydrogens in a protein structure is presented. The algorithm adds initially, to a protein data bank file of the protein, nonrotatable hydrogens such as peptide backbone hydrogens according to geometric considerations. Then, water protons and polar side chain protons of lysine, serine, threonine, tyrosine, aspartic acid, glutamic acid, and the C and N termini of a protein are added according to energy considerations. A unique stochastic approach has been developed to overcome a combinatorial explosion in the search for the lowest energy structure. First, the system is divided into ensembles. Each ensemble is treated separately: N conformations are sampled at random, their energies computed, whereas common components of high‐energy combinations are gathered on one hand, and low‐energy combinations on the other. Components that yield only high‐energy conformations and do not contribute to any low energies are excluded. This is reiterated while the total amount of combinations is decreased along the iterative process. When the total number of combinations is lower than a user defined threshold, all remaining combinations are evaluated by exhaustive search. Energy evaluations use nonbonding energy expressions alone. The program was tested on five high‐resolution crystal structures: bovine pancreatic trypsin inhibitor (Brookhaven Protein Data Bank file 5PTI), RNase‐A (5RSA), trypsin (1NTP), and carbon monoxymyoglobin (2MB5), for which neutron diffraction structures are available, as well as phosphate binding protein (1IXH) for which very high resolution X‐ray crystallography was used. The low RMS values prove the efficiency of this algorithm as a tool for positioning protons in proteins. It may be used for other biological structures. Proteins 2000;38:273–287.

Modulation of the affinity of aspartic proteases by the mutated residues in active site models

The active sites of 3 types of aspartic proteases are modeled, based on crystallographic coordinates of endothiapepsin and of a model of HIV‐1 protease. The enthalpies of deprotonation from neutral to mono‐anion and to dianion are calculated with semiempirical minimal neglect of differential overlap, hydrogen bonding corrected (MNDO/H). This quantum mechanical study of models for the active sites of pepsins, human renin and retroviral aspartic proteases demonstrates that the replacements ofThr‐218 from pepsins by Ala in human renin and of both Ser‐35 and Thr‐218 by alanines in retroviral proteases increases the proton affinity and modulates the charge distribution of those active sites compared to the pepsins.

α-Oxyiminophosphonates: chemical and physical properties. Reactions, theoretical calculations, and X-ray crystal structures of (E) and (Z)-dimethyl α-hydroxyiminobenzylphosphonates

Dialkyl α-oxyiminoalkylphosphonates, obtained by treatment of dialkyl acylphosphonates with hydroxylamine or methoxylamine, are mostly mixtures of E and Z isomers. Structural assignments of the oximes were based on X-ray crystallographic analysis of representative compounds: (E)- and (Z)-dimethyl α-hydroxyiminobenzylphosphonates [(E)- and (Z)-(1a)]. The 31P n.m.r. chemical shifts of the E isomers always appear at lower field than those of the corresponding Z isomers. Thermal fragmentation of (1a) affords benzonitrile and dimethyl hydrogen phosphate, with (Z)-(1a) reacting faster than the E isomer. α-Oxyiminophosphonates undergo E⇌Z isomerization catalysed either by acid or by base under certain conditions, the E isomer being the thermodynamically more stable one. An E+Z mixture of dialkyl α-oxyiminophosphonates can be mono-de-alkylated by non-basic nucleophiles (e.g. NaI) to afford monoalkyl oxyiminophosphonates of unchanged isomeric composition. The geometrical isomers of α-hydroxyiminophosphonates differ in their behaviour under basic conditions. While treatment of (E)-(1a) with NaOH in boiling methanol leads, by mono-de-alkylation, to sodium methyl α-hydroxyiminobenzylphosphonate [(E)-(2a)], under the same conditions (Z)-(1a) undergoes fragmentation, by C–P bond cleavage, to benzonitrile and dimethyl hydrogen phosphate. Control experiments established that the fragmentation of (Z)-α-hydroxyiminophosphonates involves an intramolecular attack on the phosphorus atom by the ionized Z oriented oxime oxygen. Similar differences in behaviour are noted between the isomers of the monoanions of α-hydroxyiminophosphonates. MNDO/H Calculations demonstrate the feasibility of forming internal hydrogen bonds in Z isomers, and their possible contribution to conformational preferences. Single-crystal X-ray diffraction studies of (E)- and (Z)-(1a), and (E)-(2b), clearly identified the geometric isomers and correlated them with the 1H and 31P n.m.r. resonances.