Transformer-Decoder GPT Models for Generating Virtual Screening Libraries of HMG-Coenzyme A Reductase Inhibitors: Effects of Temperature, Prompt Length, and Transfer-Learning Strategies.

Attention-based decoder models were used to generate libraries of novel inhibitors for the HMG-Coenzyme A reductase (HMGCR) enzyme. These deep neural network models were pretrained on previously synthesized drug-like molecules from the ZINC15 database to learn the syntax of SMILES strings and then fine-tuned with a set of ∼1000 molecules that inhibit HMGCR. The number of layers used for pretraining and fine-tuning was varied to find the optimal balance for robust library generation.

Role of Exact Exchange and Empirical Dispersion in Density Functional Theory-Based Three-Body Noncovalent Interactions.

Total and three-body interaction energies are calculated for a benchmark set of three-body systems using a range of different types of density functional theory (DFT) methods, with the results compared to CCSD(T)/CBS results from the benchmark reference [ , , 28621-28637]. Inclusion of Hartree-Fock exchange, via either a global or range-separated hybrid approach or inclusion of empirical dispersion corrections, increases accuracy for total and three-body interactions. Basis set convergence testing shows that the aug-cc-pVTZ basis set is well converged with little to no change seen when using quadruple-ζ basis sets.

Pairwise Additivity and Three-Body Contributions for Density Functional Theory-Based Protein-Ligand Interaction Energies.

The prediction of protein-ligand binding energies is crucial in computer-assisted drug design. This property can be calculated in a straightforward fashion as the difference in the energies between a binding site-ligand complex and the separated binding site and ligand. Often, though, there is value in knowing how different amino acid residues in the protein binding site interact with the ligand.

Investigating Paracetamol's Role as a Potential Treatment for Parkinson's Disease: Ab Initio Analysis of Dopamine, l-DOPA, Paracetamol, and NAPQI Interactions with Enzymes Involved in Dopamine Metabolism.

Recently, it was found that paracetamol can extend the therapeutic window of l-DOPA treatment for Parkinson's disease [Golding (2019) BJPharm, 4(2), Article 619]. It has been posited that the effect could be due to paracetamol and its metabolite, NAPQI, inhibiting pain signals in the spinal column. In this work, we examine the possibility that the therapeutic effect of the paracetamol for the Parkinson's disease patient may be due to an inhibition of the enzymes that metabolize dopamine and/or l-DOPA, thus effectively extending the lifetime of the l-DOPA treatment.

Synthesis and analysis of novel catecholic ligands as inhibitors of catechol-O-methyltransferase.

l-DOPA, a dopamine precursor, is commonly used as a treatment for patients with conditions such as Parkinson's disease. This therapeutic l-DOPA, as well as the dopamine derived from l-DOPA, can be deactivated via metabolism by catechol-O-methyltransferase (COMT). Targeted inhibition of COMT prolongs the effectiveness of l-DOPA and dopamine, resulting in a net increase in pharmacological efficiency of the treatment strategy.

Propargylglycine-based antimicrobial compounds are targets of TolC-dependent efflux systems in Escherichia coli.

A library of novel l-propargylglycine-based compounds were designed and synthesized with the goal of inhibiting the growth of Gram-negative bacteria by targeting LpxC, a highly conserved Gram-negative enzyme which performs an essential step in the lipid A biosynthetic pathway. These compounds were designed with and without a nucleoside and had varying tail structures, which modulate their lipophilicity. The synthetic scheme was improved compared to previous methods: a methyl ester intermediate was converted to a hydroxamic acid, which obviated the need for a THP protecting group and improved the yields and purity of the final compounds.

Design, Modeling and Synthesis of 1,2,3-Triazole-Linked Nucleoside-Amino Acid Conjugates as Potential Antibacterial Agents.

Copper-catalyzed azide-alkyne cycloadditions (CuAAC or click chemistry) are convenient methods to easily couple various pharmacophores or bioactive molecules. A new series of 1,2,3-triazole-linked nucleoside-amino acid conjugates have been designed and synthesized in 57-76% yields using CuAAC. The azido group was introduced on the 5'-position of uridine or the acyclic analogue using the tosyl-azide exchange method and alkylated serine or proparylglycine was the alkyne.

Examination of tyrosine/adenine stacking interactions in protein complexes.

The π-stacking interactions between tyrosine amino acid side chains and adenine-bearing ligands are examined. Crystalline protein structures from the protein data bank (PDB) exhibiting face-to-face tyrosine/adenine arrangements were used to construct 20 unique 4-methylphenol/N9-methyladenine (p-cresol/9MeA) model systems. Full geometry optimization of the 20 crystal structures with the M06-2X density functional theory method identified 11 unique low-energy conformations.

Aromatic interactions in the binding of ligands to HMGCoA reductase.

3-Hydroxy-3-methyglutaryl-coenzyme A (HMGCoA) reductase is the enzyme that catalyzes the rate-determining step in cholesterol synthesis; it is also the target for statin drugs, which are competitive inhibitors of the enzyme. We examine potentially important enzyme-ligand interactions currently not incorporated into statin drug design: weak, induction/dispersion interactions between ligands and residue tyrosine 479 in the HMGCoA reductase active site. HMGCoA is a large molecule with a long coenzyme A "tail", and in order to study the interactions of interest, it was necessary to find the smallest possible portion of the HMGCoA molecule that would serve as a reasonable model for the entire molecule.

The importance of aromatic-type interactions in serotonin synthesis: protein-ligand interactions in tryptophan hydroxylase and aromatic amino acid decarboxylase.

The biosynthesis of serotonin requires aromatic substrates to be bound in the active sites of the enzymes tryptophan hydroxylase and aromatic amino acid decarboxylase. These aromatic substrates are held in place partially by dispersion and induction interactions with the enzymes' aromatic amino acid residues. Mutations that decrease substrate binding can result in a decrease in serotonin production and thus can lead to depression and related disorders.

Interaction energies between tetrahydrobiopterin analogues and aromatic residues in tyrosine hydroxylase and phenylalanine hydroxylase.

The phenylalanine residues 300 and 309 in the enzyme tyrosine hydroxylase are known to aid in the positioning and binding of tetrahydrobiopterin (BH4) to the enzyme active site. The residues phenylalanine 254 and tyrosine 325 similarly aid in binding BH4 in phenylalanine hydroxylase. BH4 is a cofactor necessary for enzyme function, and mutations in these residues have been shown to cause a decrease in enzyme function.

Non-Born-Oppenheimer molecular structure and one-particle densities for H2D+.

We show that the nonadiabatic (non-Born-Oppenheimer) ground state of a three-nuclei system can be effectively calculated with the use of an explicitly correlated Gaussian basis set with floating centers. Sample calculations performed for the H2D+ system with various basis set sizes show good convergence with respect to both the total energy and the expectation values of the internuclear distances (molecular geometry), the distances between the nuclei and the electrons, and between the electrons. We also provide a derivation of the formulas for one-particle density calculations and some density plots showing the spatial distribution of the H2D+ nuclear and electronic densities.

Non-Born-Oppenheimer isotope effects on the polarizabilities of H2.

We present non-Born-Oppenheimer quantum-mechanical calculations of the behavior of isolated molecules of the H2 isotopomer series in static electric fields. Some conceptual aspects of such calculations are discussed. The values for polarizabilities and hyperpolarizabilities of the H2 isotopomers which we present are the first ever fully nonadiabatic calculated values of these properties.

Nonadiabatic Calculations of the Dipole Moments of LiH and LiD.

We present very high-accuracy fully nonadiabatic calculated values for the dipole moments for the ground states of LiH and LiD. These results were calculated via numerical differentiation of the energy obtained at different electric field strengths. The values for the energy were obtained from variational optimization with analytical gradients of the wave function expanded in a basis of explicitly correlated floating s-type Gaussian functions.