Open-source QSAR models for pKa prediction using multiple machine learning approaches.

BACKGROUND: The logarithmic acid dissociation constant pKa reflects the ionization of a chemical, which affects lipophilicity, solubility, protein binding, and ability to pass through the plasma membrane. Thus, pKa affects chemical absorption, distribution, metabolism, excretion, and toxicity properties. Multiple proprietary software packages exist for the prediction of pKa, but to the best of our knowledge no free and open-source programs exist for this purpose. Using a freely available data set and three machine learning approaches, we developed open-source models for pKa prediction. METHODS: The experimental strongest acidic and strongest basic pKa values in water for 7912 chemicals were obtained from DataWarrior, a freely available software package. Chemical structures were curated and standardized for quantitative structure-activity relationship (QSAR) modeling using KNIME, and a subset comprising 79% of the initial set was used for modeling. To evaluate different approaches to modeling, several datasets were constructed based on different processing of chemical structures with acidic and/or basic pKas. Continuous molecular descriptors, binary fingerprints, and fragment counts were generated using PaDEL, and pKa prediction models were created using three machine learning methods, (1) support vector machines (SVM) combined with k-nearest neighbors (kNN), (2) extreme gradient boosting (XGB) and (3) deep neural networks (DNN). RESULTS: The three methods delivered comparable performances on the training and test sets with a root-mean-squared error (RMSE) around 1.5 and a coefficient of determination (R2) around 0.80. Two commercial pKa predictors from ACD/Labs and ChemAxon were used to benchmark the three best models developed in this work, and performance of our models compared favorably to the commercial products. CONCLUSIONS: This work provides multiple QSAR models to predict the strongest acidic and strongest basic pKas of chemicals, built using publicly available data, and provided as free and open-source software on GitHub.

Drug-induced Liver Fibrosis: Testing Nevirapine in a Viral-like Liver Setting Using Histopathology, MALDI IMS, and Gene Expression.

Nevirapine (NVP) is associated with hepatotoxicity in 1-5% of patients. In rodent studies, NVP has been shown to cause hepatic enzyme induction, centrilobular hypertrophy, and skin rash in various rat strains but not liver toxicity. In an effort to understand whether NVP is metabolized differently in a transiently inflamed liver and whether a heightened immune response alters NVP-induced hepatic responses, female brown Norway rats were dosed with either vehicle or NVP alone (75 mg/kg/day for 15 days) or galactosamine alone (single intraperitoneal [ip] injection on day 7 to mimic viral hepatitis) or a combination of NVP (75/100/150 mg/kg/day for 15 days) and galactosamine (single 750 mg/kg ip on day 7). Livers were collected at necropsy for histopathology, matrix-assisted laser desorption/ionization imaging mass spectrometry and gene expression. Eight days after galactosamine, hepatic fibrosis was noted in rats dosed with the combination of NVP and galactosamine. No fibrosis occurred with NVP alone or galactosamine alone. Gene expression data suggested a viral-like response initiated by galactosamine via RNA sensors leading to apoptosis, toll-like receptor, and dendritic cell responses. These were exacerbated by NVP-induced growth factor, retinol, apoptosis, and periostin effects. This finding supports clinical reports warning against exacerbation of fibrosis by NVP in patients with hepatitis C.

An initial characterization of N-terminal-proatrial natriuretic peptide in serum of Sprague Dawley rats.

In the clinical setting, natriuretic peptides (NPs) have proven to be reliable noninvasive markers for diagnostic, prognostic, and therapeutic monitoring of heart failure. Given their proven utility in humans, NPs are potential candidates for translational biomarkers during drug development to detect drug-induced hemodynamic stress resulting in cardiac hypertrophy in preclinical species. We evaluated the intra- and interassay precision and the stability of serum N-terminal-proatrial natriuretic peptide (NT-proANP) using a commercially available enzyme-linked immunoassay (EIA). We then measured NT-proANP concentrations in 532 serum samples from 337 male Crl:CD(SD) rats with or without pressure-induced cardiac hypertrophy. Additionally, we established a reference range using samples from control animals across multiple studies. The data demonstrate that the NT-proANP EIA is a robust and reproducible assay for the measurement of NT-proANP. The noninvasive translational utility, minimal sample volume requirement, and the lack of existing hypertrophic biomarkers in the male rat make NT-proANP an excellent candidate for further interrogation as a biomarker of cardiac hypertrophy in preclinical toxicology investigations.

Biarylaniline phenethanolamines as potent and selective beta3 adrenergic receptor agonists.

The synthesis of a series of phenethanolamine aniline agonists that contain an aniline ring on the right-hand side of the molecule substituted at the meta position with a benzoic acid or a pyridyl carboxylate is described. Several of the analogues (e.g., 34, 36-38, 40, and 44) have high beta(3) adrenergic receptor (AR) potency and selectivity against beta(1) and beta(2) ARs in Chinese hamster ovary (CHO) cells expressing beta ARs. The dog pharmacokinetic profile of some of these analogues showed >25% oral bioavailability and po half-lives of at least 1.5 h. Among the compounds described herein, the 3,3'-biarylaniline carboxylate derivatives 36, 38 and the phenylpyridyl derivative 44 demonstrated outstanding in vitro properties and reasonable dog pharmacokinetic profiles. These three analogues also showed dose dependent beta(3) AR mediated responses in mice. The ease of synthesis and superior dog pharmacokinetics of compound 38 relative to that of 44 in combination with its in vitro profile led us to choose this compound as a development candidate for the treatment of type 2 diabetes.

Flow cytometric assessment of peroxisome proliferation from frozen liver of fibrate-treated monkeys.

Multiple methods currently exist for the assessment of peroxisome proliferation, including gene expression, enzyme activity, immunolabeling coupled with image analysis, and electron microscopy. This study describes a novel flow cytometric method to efficiently quantify peroxisome proliferation in cells from frozen livers. Frozen livers from cynomolgus monkeys treated with ciprofibrate at doses of 0, 3, 30, 150, and 400 mg/kg/day for 15 days were mechanically disaggregated using an automated dispersion method. The resulting cell suspensions were labeled using an allophycocyanin (APC)-conjugated antibody directed against peroxisomal membrane protein 70 (PMP70). Statistically significant increases in mean fluorescence intensity were observed from animals dosed at 30, 150, and 400 mg/kg/day compared to control. Parallel comparisons using electron microscopy and immunofluorescence microscopy suggest that flow cytometry may be an alternative to electron microscopy in determinations of peroxisome proliferation. Flow cytometric analysis of freshly isolated hepatocytes and frozen liver from rats treated with fenofibrate at 200 mg/kg/day for 10 days showed the flow cytometric method could detect peroxisome proliferation in both species. The research described here demonstrates the feasibility of applying flow cytometry for the detection of peroxisome proliferation.

Gene expression profiling of the PPAR-alpha agonist ciprofibrate in the cynomolgus monkey liver.

Fibrates, such as ciprofibrate, fenofibrate, and clofibrate, are peroxisome proliferator-activated receptor-alpha (PPARalpha) agonists that have been in clinical use for many decades for treatment of dyslipidemia. When mice and rats are given PPARalpha agonists, these drugs cause hepatic peroxisome proliferation, hypertrophy, hyperplasia, and eventually hepatocarcinogenesis. Importantly, primates are relatively refractory to these effects; however, the mechanisms for the species differences are not clearly understood. Cynomolgus monkeys were exposed to ciprofibrate at various dose levels for either 4 or 15 days, and the liver transcriptional profiles were examined using Affymetrix human GeneChips. Strong upregulation of many genes relating to fatty acid metabolism and mitochondrial oxidative phosphorylation was observed; this reflects the known pharmacology and activity of the fibrates. In addition, (1) many genes related to ribosome and proteasome biosynthesis were upregulated, (2) a large number of genes downregulated were in the complement and coagulation cascades, (3) a number of key regulatory genes, including members of the JUN, MYC, and NFkappaB families were downregulated, which appears to be in contrast to the rodent, where JUN and MYC are reported to upregulated after PPARalpha agonist treatment, (4) no transcriptional signal for DNA damage or oxidative stress was observed, and (5) transcriptional signals consistent with an anti-proliferative and a pro-apoptotic effect were seen. We also compared the primate data to literature reports of hepatic transcriptional profiling in PPARalpha-treated rodents, which showed that the magnitude of induction in beta-oxidation pathways was substantially greater in the rodent than the primate.

Fibrates induce hepatic peroxisome and mitochondrial proliferation without overt evidence of cellular proliferation and oxidative stress in cynomolgus monkeys.

There is little primate risk factor data in the literature evaluating the relationship between proposed mechanisms of PPAR agonist-induced hepatocarcinogenesis at clinically relevant therapeutic exposures. These studies were conducted to characterize the hepatic effects of fenofibrate and ciprofibrate in the cynomolgus monkey. Male cynomolgus monkeys were given fenofibrate (250, 1250 or 2500 mg/kg/day) or ciprofibrate (3, 30, 150 or 400 mg/kg/day) for up to 15 days. The highest doses used were approximately 4 times (fenofibrate) and 9.4 times (ciprofibrate) the human therapeutic exposure for these agents based on AUC (area under the curve). For both compounds, there was a treatment-related increase in liver weight and periportal hepatocellular hypertrophy, which was related to increases in peroxisomes (up to 2.8 times controls) and mitochondria (up to 2.5 times controls). An increase in smooth endoplasmic reticulum probably contributed to the hypertrophy. There was no indication of cell proliferation as determined by the number of mitotic figures and this was confirmed by evaluating cell proliferation by immunohistochemical staining for the Ki-67 antigen. Consistent with the findings by light microscopy, there was no treatment-related effect on the level of mRNA for proteins known to be involved in the control of hepatocyte cell division or apoptosis (e.g. P21, Cyclin D1, PCNA, CDKN1A). Furthermore, there was minimal indication of oxidative stress. Thus, there was no evidence of lipofuscin accumulation, and there was no remarkable increase in the mRNA levels for most proteins known to respond to oxidative stress (e.g. catalase, glutathione peroxidase). A mild induction in the mRNA levels of cellular beta-oxidation and detoxification enzymes (e.g. acyl CoA oxidase, thioredoxin reductase) was observed. Collectively, the data from these studies suggest that the primate responds to PPARalpha agonists in a manner that is different from the rodent suggesting that the primate may be refractory to PPAR-induced hepatocarcinogenesis.

Genetic toxicology of remifentanil, an opiate analgesic.

Compounds that interact with opioid receptors are commonly used as analgesics. Opioid agonists vary in their potency and pharmacokinetic properties as well as in their affinity for distinct opioid receptors. The fentanyl opiate analogues are an important group of analgesics that interact with the mu opioid receptor. Remifentanil (GI87084) is a particularly interesting member of this group of opioids because its action is especially short in duration. This report examines the genetic toxicology of remifentanil. Remifentanil was not genotoxic in an Ames test, an in vitro chromosome aberration assay in Chinese hamster ovary cells, an in vivo micronucleus assay in rat erythrocytes, or an in vivo/in vitro unscheduled DNA synthesis assay in rat hepatocytes. In the in vitro L5178Y tk(+/-) mouse lymphoma assay, remifentanil produced a genotoxic response at dose levels >or=308 microg/mL only in the presence of rat liver S9 metabolic activation; primarily tiny and small mutant colonies were produced. This pattern of activity in a battery of genetic toxicology assays is not unique to remifentanil, but has also been observed for other pharmaceuticals, including the opioid fentanyl. A weight-of-evidence analysis, taking into consideration genotoxic mechanisms, in vivo results, and the conditions of clinical use, suggests remifentanil does not pose a genotoxic risk to patients.

Comparison of the computer programs DEREK and TOPKAT to predict bacterial mutagenicity. Deductive Estimate of Risk from Existing Knowledge. Toxicity Prediction by Komputer Assisted Technology.

The performance of two computer programs, DEREK and TOPKAT, was examined with regard to predicting the outcome of the Ames bacterial mutagenicity assay. The results of over 400 Ames tests conducted at Glaxo Wellcome (now GlaxoSmithKline) during the last 15 years on a wide variety of chemical classes were compared with the mutagenicity predictions of both computer programs. DEREK was considered concordant with the Ames assay if (i) the Ames assay was negative (not mutagenic) and no structural alerts for mutagenicity were identified or (ii) the Ames assay was positive (mutagenic) and at least one structural alert was identified. Conversely, the DEREK output was considered discordant if (i) the Ames assay was negative and any structural alert was identified or (ii) the Ames assay was positive and no structural alert was identified. The overall concordance of the DEREK program with the Ames results was 65% and the overall discordance was 35%, based on over 400 compounds. About 23% of the test molecules were outside the permissible limits of the optimum prediction space of TOPKAT. Another 4% of the compounds were either not processable or had indeterminate mutagenicity predictions; these molecules were excluded from the TOPKAT analysis. If the TOPKAT probability was (i) > or =0.7 the molecule was predicted to be mutagenic, (ii) < or =0.3 the compound was predicted to be non-mutagenic and (iii) between 0.3 and 0.7 the prediction was considered indeterminate. From over 300 acceptable predictions, the overall TOPKAT concordance was 73% and the overall discordance was 27%. While the overall concordance of the TOPKAT program was higher than DEREK, TOPKAT fared more poorly than DEREK in the critical Ames-positive category, where 60% of the compounds were incorrectly predicted by TOPKAT as negative but were mutagenic in the Ames test. For DEREK, 54% of the Ames-positive molecules had no structural alerts and were predicted to be non-mutagenic. Alternative methods of analyzing the output of the programs to increase the accuracy with Ames-positive compounds are discussed.

Improving structure-linked access to publicly available chemical toxicity information.

Publicly available toxicity databases serve as the central resource in efforts to develop algorithms for assessing potential chemical toxicity. File standardization and linkage of chemical structures with chemical toxicity information are essential first steps in providing broad access to existing toxicity information, for deriving useful structure-activity relationship (SAR) models, performing analog searches, and estimating the potential toxicity of new chemicals. This review will focus on current efforts to improve structure-linked access to publicly available sources of toxicity information, outlining current web-based resources as well as two new database initiatives for standardizing and consolidating public chemical toxicity information.