Prediction of Human Clearance Using In Silico Models with Reduced Bias.

Predicting human clearance with high accuracy from in silico-derived parameters alone is highly desirable, as it is fast, saves in vitro resources, and is animal-sparing. We derived random forest (RF) models from 1340 compounds with human intravenous pharmacokinetic (PK) data, the largest data set publicly available today. To assess the general applicability of the RF models, we systematically removed structural-therapeutic class analogues and other compounds with structural similarity from the training sets. For a quasi-prospective test set of 343 compounds, we show that RF models devoid of structurally similar compounds in the training set predict human clearance with a geometric mean fold error (GMFE) of 3.3. While the observed GMFE illustrates how difficult it is to generate a useful model that is broadly applicable, we posit that our RF models yield a more realistic assessment of how well human clearance can be predicted prospectively. We deployed the conformal prediction formalism to assess the model applicability and to determine the prediction confidence intervals for each prediction. We observed that clearance can be predicted better for renally cleared compounds than for other clearance mechanisms. We show that applying a classification model for predicting renal clearance identifies a subset of compounds for which clearance can be predicted with higher accuracy, yielding a GMFE of 2.3. In addition, our in silico RF human clearance models compared well to models derived from scaling human hepatocytes or preclinical in vivo data.

The allosteric inhibitor ABL001 enables dual targeting of BCR-ABL1.

Chronic myeloid leukaemia (CML) is driven by the activity of the BCR-ABL1 fusion oncoprotein. ABL1 kinase inhibitors have improved the clinical outcomes for patients with CML, with over 80% of patients treated with imatinib surviving for more than 10 years. Second-generation ABL1 kinase inhibitors induce more potent molecular responses in both previously untreated and imatinib-resistant patients with CML. Studies in patients with chronic-phase CML have shown that around 50% of patients who achieve and maintain undetectable BCR-ABL1 transcript levels for at least 2 years remain disease-free after the withdrawal of treatment. Here we characterize ABL001 (asciminib), a potent and selective allosteric ABL1 inhibitor that is undergoing clinical development testing in patients with CML and Philadelphia chromosome-positive (Ph+) acute lymphoblastic leukaemia. In contrast to catalytic-site ABL1 kinase inhibitors, ABL001 binds to the myristoyl pocket of ABL1 and induces the formation of an inactive kinase conformation. ABL001 and second-generation catalytic inhibitors have similar cellular potencies but distinct patterns of resistance mutations, with genetic barcoding studies revealing pre-existing clonal populations with no shared resistance between ABL001 and the catalytic inhibitor nilotinib. Consistent with this profile, acquired resistance was observed with single-agent therapy in mice; however, the combination of ABL001 and nilotinib led to complete disease control and eradicated CML xenograft tumours without recurrence after the cessation of treatment.

An Accurate In Vitro Prediction of Human VDss Based on the Øie-Tozer Equation and Primary Physicochemical Descriptors. 3. Analysis and Assessment of Predictivity on a Large Dataset.

We present a model for volume of distribution at steady state (VDss) prediction, via fraction unbound in tissues, from the Øie-equation as an extension of our and other authors' previous work. It is based on easily determined or computed physicochemical descriptors such as logD7.4 and fi (7.4) (cationic fraction ionized at pH 7.4) in addition to fraction unbound in plasma (fup). We had collected, as part of other work, an extensive dataset of VDss and fup values and used the descriptors above, gathered from the literature, for a preliminary assessment of the robustness of the method applied to 191 different compounds belonging to different charge classes and scaffolds. After this step, we addressed the use of easily computed physicochemical descriptors and experimentally derived fup on the same data set and compare the results between the two approaches and against the Øie-Tozer equation using in vivo data. This approach positions itself between fully computational models and scaling methods based on in vivo animal models or in vitro Kp (tissue:plasma) data utilizing model tissues. We consider it a useful and orthogonal complement to the two very diverse approaches mentioned yet requiring minimal in vitro experimental work. It offers a relatively inexpensive, rapid, intuitive, and simple way to predict VDss in humans, at a relatively early stage of the drug discovery. SIGNIFICANCE STATEMENT: This method allows the prediction of volume of distribution at steady state for small molecules in humans without the use of animal PK data because it utilizes only in vitro data. It is therefore amenable to use at early stages, simple, intuitive, animal-sparing, and quite accurate, and it may serve scaling efforts well. Furthermore, utilizing the same dataset, we show that the performance of a model using computed pKa and logD7.4, still using experimental fraction unbound in plasma, compares well with the model using experimentally derived values.

Trend Analysis of a Database of Intravenous Pharmacokinetic Parameters in Humans for 1352 Drug Compounds.

We report a trend analysis of human intravenous pharmacokinetic data on a data set of 1352 drugs. The aim in building this data set and its detailed analysis was to provide, as in the previous case published in 2008, an extended, robust, and accurate resource that could be applied by drug metabolism, clinical pharmacology, and medicinal chemistry scientists to a variety of scaling approaches. All in vivo data were obtained or derived from original references, either through the literature or regulatory agency reports, exclusively from studies utilizing intravenous administration. Plasma protein binding data were collected from other available sources to supplement these pharmacokinetic data. These parameters were analyzed concurrently with a range of physicochemical properties, and resultant trends and patterns within the data are presented. In addition, the date of first disclosure of each molecule was reported and the potential "temporal" impact on data trends was analyzed. The findings reported here are consistent with earlier described trends between pharmacokinetic behavior and physicochemical properties. Furthermore, the availability of a large data set of pharmacokinetic data in humans will be important to further pursue analyses of physicochemical properties, trends, and modeling efforts and should propel our deeper understanding (especially in terms of clearance) of the absorption, distribution, metabolism, and excretion behavior of drug compounds.

Benefit of Retraining pKa Models Studied Using Internally Measured Data.

The ionization state of drugs influences many pharmaceutical properties such as their solubility, permeability, and biological activity. It is therefore important to understand the structure property relationship for the acid-base dissociation constant pKa during the lead optimization process to make better-informed design decisions. Computational approaches, such as implemented in MoKa, can help with this; however, they often predict with too large error especially for proprietary compounds. In this contribution, we look at how retraining helps to greatly improve prediction error. Using a longitudinal study with data measured over 15 years in a drug discovery environment, we assess the impact of model training on prediction accuracy and look at model degradation over time. Using the MoKa software, we will demonstrate that regular retraining is required to address changes in chemical space leading to model degradation over six to nine months.

In silico prediction of total human plasma clearance.

The prediction of the total human plasma clearance of novel chemical entities continues to be of paramount importance in drug design and optimization, because it impacts both dose size and dose regimen. Although many in vivo and in vitro methods have been proposed, a well-constructed, well-validated, and less resource-intensive computational tool would still be very useful in an iterative compound design cycle. A new completely in silico linear PLS (partial least-squares) model to predict the human plasma clearance was built on the basis of a large data set of 754 compounds using physicochemical descriptors and structural fragments, the latter able to better represent biotransformation processes. The model has been validated using the "ELASTICO" approach (Enhanced Leave Analog-Structural, Therapeutic, Ionization Class Out) based on ten therapeutic/structural analog classes. The model yields a geometric mean fold error (GMFE) of 2.1 and a percentage of compounds predicted within 2- and 3-fold error of 59% and 80%, respectively, showing an improved performance when compared with previous published works in predicting clearance of neutral compounds, and a very good performance with ionized molecules at pH 7.5, able to compare favorably with fairly accurate in vivo methods.

In Silico Models of Human PK Parameters. Prediction of Volume of Distribution Using an Extensive Data Set and a Reduced Number of Parameters.

A novel, descriptor-parsimonious in silico model to predict human VDss (volume of distribution at steady-state) has been derived and thoroughly tested in a quasi-prospective regimen using an independent test set of 213 compounds. The model performs on par with a former benchmark model that relied on far more descriptors. As a result, the new random forest model relying on only six descriptors allows for interpretations that help chemists to design compounds with desired human VDss values. A comparison of in silico predictions of VDss with models using in vitro derived descriptors or in vivo scaling methods supports the strength of the in-silico approach, considering its resource- and animal-sparing nature. The strong performance of the in silico VDss models on structurally novel compounds supports the high degree of confidence that can be placed in using in silico human VDss predictions for compound design and human dose predictions.

Interaction of DDSDEEN peptide with N-CAM protein. Possible mechanism enhancing neuronal differentiation.

DDSDEEN chromatin peptide, after dansylation, was studied for its ability to bind N-CAM protein. The binding causes a quenching of the Dns-peptide fluorescence emission. Dose- and time-dependent binding of Dns-peptide with N-CAM has been shown. Fluorescence quenching is completely lost if the Dns-peptide is subjected to carboxypeptidase digestion. Moreover the undansylated peptide pEDDSDEEN competes with the DnsDDSDEEN peptide for the binding with the N-CAM protein. The Dns-peptide-N-CAM bond has been related to the peptide biological activity probably involved in the promotion of neuronal differentiation. An attempt to recognize a possible N-CAM binding site for Dns-peptide was performed by alignment of N-CAM from various sources with some sequences that have been previously reported as binding sites for the pEDDSDEEN and DDSDEEN peptides. Interestingly, the alignment of N-CAM from various sources with the peptides WHPREGWAL and WFPRWAGQA recognizes on rat and human N-CAM a unique sequence that could be the specific binding site for chromatin peptide: WHSKWYDAK. This sequence is present in fibronectin type-III domain of N-CAM. In addition molecular modeling studies indicate the N-CAM sequence WHSKWYDAK as, probably, the main active site for DnsDDSDEEN (or pEDDSDEEN) peptide ligand. Accordingly the binding experiments show a high affinity between WHSKWYDAK and DnsDDSDEEN peptides.

Design and synthesis of trans 2-(furan-2-yl)vinyl heteroaromatic iodides with antitumour activity.

A new molecular modelling approach based on physico-chemical and pharmacokinetic properties, called Volsurf plus, was used to design new heterocyclic compounds with antiproliferative activity. The synthesis and in vitro antitumour tests on a breast carcinoma cell line (MCF7) confirmed VOLSURF predicted activity values.

Discovery of Asciminib (ABL001), an Allosteric Inhibitor of the Tyrosine Kinase Activity of BCR-ABL1.

Chronic myelogenous leukemia (CML) arises from the constitutive activity of the BCR-ABL1 oncoprotein. Tyrosine kinase inhibitors (TKIs) that target the ATP-binding site have transformed CML into a chronic manageable disease. However, some patients develop drug resistance due to ATP-site mutations impeding drug binding. We describe the discovery of asciminib (ABL001), the first allosteric BCR-ABL1 inhibitor to reach the clinic. Asciminib binds to the myristate pocket of BCR-ABL1 and maintains activity against TKI-resistant ATP-site mutations. Although resistance can emerge due to myristate-site mutations, these are sensitive to ATP-competitive inhibitors so that combinations of asciminib with ATP-competitive TKIs suppress the emergence of resistance. Fragment-based screening using NMR and X-ray yielded ligands for the myristate pocket. An NMR-based conformational assay guided the transformation of these inactive ligands into ABL1 inhibitors. Further structure-based optimization for potency, physicochemical, pharmacokinetic, and drug-like properties, culminated in asciminib, which is currently undergoing clinical studies in CML patients.

Systematic Evaluation of Wajima Superposition (Steady-State Concentration to Mean Residence Time) in the Estimation of Human Intravenous Pharmacokinetic Profile.

We present a systematic evaluation of the Wajima superpositioning method to estimate the human intravenous (i.v.) pharmacokinetic (PK) profile based on a set of 54 marketed drugs with diverse structure and range of physicochemical properties. We illustrate the use of average of "best methods" for the prediction of clearance (CL) and volume of distribution at steady state (VDss) as described in our earlier work (Lombardo F, Waters NJ, Argikar UA, et al. J Clin Pharmacol. 2013;53(2):178-191; Lombardo F, Waters NJ, Argikar UA, et al. J Clin Pharmacol. 2013;53(2):167-177). These methods provided much more accurate prediction of human PK parameters, yielding 88% and 70% of the prediction within 2-fold error for VDss and CL, respectively. The prediction of human i.v. profile using Wajima superpositioning of rat, dog, and monkey time-concentration profiles was tested against the observed human i.v. PK using fold error statistics. The results showed that 63% of the compounds yielded a geometric mean of fold error below 2-fold, and an additional 19% yielded a geometric mean of fold error between 2- and 3-fold, leaving only 18% of the compounds with a relatively poor prediction. Our results showed that good superposition was observed in any case, demonstrating the predictive value of the Wajima approach, and that the cause of poor prediction of human i.v. profile was mainly due to the poorly predicted CL value, while VDss prediction had a minor impact on the accuracy of human i.v. profile prediction.

Pharmacophore, drug metabolism, and pharmacokinetics models on non-peptide AT1, AT2, and AT1/AT2 angiotensin II receptor antagonists.

About 20 non-peptide angiotensin II receptor antagonists are in various stages of clinical development. Different modeling approaches were used to predict the pharmacophoric requirements for AT(1) (angiotensin II receptor subtype 1) affinity. However, to our knowledge, none was used to predict both the selectivity toward AT(1) and AT(2) (angiotensin II receptor subtype 2) receptor subtypes. In this paper, partial least squares discriminant analysis is applied to derive the chemical features guiding AT(1) and AT(2) selectivity or mixed AT(1)/AT(2) receptor binding. The method can be used to modulate AT(1) versus AT(2) selectivity. Concerns that unopposed stimulation of the AT(2) receptor might produce adverse effects initiated a search for new balanced antagonists. Moreover, it can serve as a fast filtering procedure in database searches. Finally, some relevant pharmacokinetics and metabolic properties of the database of 53 compounds are calculated using the VolSurf and MetaSite software to allow the simultaneous characterization of pharmacodynamic and pharmacokinetics properties of the chemical space of angiotensin II receptor antagonists.

Englerin A Agonizes the TRPC4/C5 Cation Channels to Inhibit Tumor Cell Line Proliferation.

Englerin A is a structurally unique natural product reported to selectively inhibit growth of renal cell carcinoma cell lines. A large scale phenotypic cell profiling experiment (CLiP) of englerin A on ¬over 500 well characterized cancer cell lines showed that englerin A inhibits growth of a subset of tumor cell lines from many lineages, not just renal cell carcinomas. Expression of the TRPC4 cation channel was the cell line feature that best correlated with sensitivity to englerin A, suggesting the hypothesis that TRPC4 is the efficacy target for englerin A. Genetic experiments demonstrate that TRPC4 expression is both necessary and sufficient for englerin A induced growth inhibition. Englerin A induces calcium influx and membrane depolarization in cells expressing high levels of TRPC4 or its close ortholog TRPC5. Electrophysiology experiments confirmed that englerin A is a TRPC4 agonist. Both the englerin A induced current and the englerin A induced growth inhibition can be blocked by the TRPC4/C5 inhibitor ML204. These experiments confirm that activation of TRPC4/C5 channels inhibits tumor cell line proliferation and confirms the TRPC4 target hypothesis generated by the cell line profiling. In selectivity assays englerin A weakly inhibits TRPA1, TRPV3/V4, and TRPM8 which suggests that englerin A may bind a common feature of TRP ion channels. In vivo experiments show that englerin A is lethal in rodents near doses needed to activate the TRPC4 channel. This toxicity suggests that englerin A itself is probably unsuitable for further drug development. However, since englerin A can be synthesized in the laboratory, it may be a useful chemical starting point to identify novel modulators of other TRP family channels.

Clearance mechanism assignment and total clearance prediction in human based upon in silico models.

We introduce a two-tier model based on an exhaustive data set, where discriminant models based on principal component analysis (PCA) and partial least squares (PLS) are used separately and in conjunction, and we show that PCA is highly discriminant approaching 95% accuracy in the assignment of the primary clearance mechanism. Furthermore, the PLS model achieved a quantitative predictive performance comparable to methods based on scaling of animal data while not requiring the use of either in vivo or in vitro data, thus sparing the use of animal. This is likely the highest performance that can be expected from a computational approach, and further improvements may be difficult to reach. We further offer the medicinal scientist a PCA model to guide in vitro and/or in vivo studies to help limit the use of resources via very rapid computations.

Comprehensive assessment of human pharmacokinetic prediction based on in vivo animal pharmacokinetic data, part 1: volume of distribution at steady state.

The authors present a comprehensive analysis on the estimation of volume of distribution at steady state (VD(ss) ) in human based on rat, dog, and monkey data on nearly 400 compounds for which there are also associated human data. This data set, to the authors- knowledge, is the largest publicly available, has been carefully compiled from literature reports, and was expanded with some in-house determinations such as plasma protein binding data. This work offers a good statistical basis for the evaluation of applicable prediction methods, their accuracy, and some methods-dependent diagnostic tools. The authors also grouped the compounds according to their charge classes and show the applicability of each method considered to each class, offering further insight into the probability of a successful prediction. Furthermore, they found that the use of fraction unbound in plasma, to obtain unbound volume of distribution, is generally detrimental to accuracy of several methods, and they discuss possible reasons. Overall, the approach using dog and monkey data in the íie-Tozer equation offers the highest probability of success, with an intrinsic diagnostic tool based on aberrant values (<0 or >1) for the calculated fraction unbound in tissue. Alternatively, methods based on dog data (single-species scaling) and rat and dog data (íie-Tozer equation with 2 species or multiple regression methods) may be considered reasonable approaches while not requiring data in nonhuman primates.

Comprehensive assessment of human pharmacokinetic prediction based on in vivo animal pharmacokinetic data, part 2: clearance.

A comprehensive analysis on the prediction of human clearance based on intravenous pharmacokinetic data from rat, dog, and monkey for approximately 400 compounds was undertaken. This data set has been carefully compiled from literature reports and expanded with some in-house determinations for plasma protein binding and rat clearance. To the authors- knowledge, this is the largest publicly available data set. The present examination offers a comparison of 37 different methods for prediction of human clearance across compounds of diverse physicochemical properties. Furthermore, this work demonstrates the application of each prediction method to each charge class of the compounds, thus presenting an additional dimension to prediction of human pharmacokinetics. In general, the observations suggest that methods employing monkey clearance values and a method incorporating differences in plasma protein binding between rat and human yield the best overall predictions as suggested by approximately 60% compounds within 2-fold geometric mean-fold error. Other single-species scaling or proportionality methods incorporating the fraction unbound in the corresponding preclinical species for prediction of free clearance in human were generally unsuccessful.

Evolution of the physicochemical properties of marketed drugs: can history foretell the future?

A set of diverse bioactive molecules, relevant from a medicinal chemistry viewpoint, was assembled and used to navigate the physicochemical property space of new and old, or traditional drugs against a larger set of 12,000 diverse bioactive small molecules. Most drugs on the market only occupy a fraction of the property space of the bioactive molecules, whereas new molecular entities (NMEs) approved since 2002 are moving away from this traditional drug space. In this new territory, semi-empirical rules derived from knowledge accumulated from historic, older molecules are not necessarily valid and different liabilities become more prominent.

In silico prediction of volume of distribution in human using linear and nonlinear models on a 669 compound data set.

The prediction of human pharmacokinetics early in the drug discovery cycle has become of paramount importance, aiding candidate selection and benefit-risk assessment. We present herein computational models to predict human volume of distribution at steady state (VD(ss)) entirely from in silico structural descriptors. Using both linear and nonlinear statistical techniques, partial least-squares (PLS), and random forest (RF) modeling, a data set of human VD(ss) values for 669 drug compounds recently published ( Drug Metab. Disp. 2008 , 36 , 1385 - 1405 ) was explored. Descriptors covering 2D and 3D molecular topology, electronics, and physical properties were calculated using MOE and Volsurf+. Model evaluation was accomplished using a leave-class-out approach on nine therapeutic or structural classes. The models were assessed using an external test set of 29 additional compounds. Our analysis generated models, both via a single method or consensus which were able to predict human VD(ss) within geometric mean 2-fold error, a predictive accuracy considered good even for more resource-intensive approaches such as those requiring data generated from studies in multiple animal species.