Identification of a Rule to Predict Response to Sarilumab in Patients with Rheumatoid Arthritis Using Machine Learning and Clinical Trial Data.

INTRODUCTION: In rheumatoid arthritis, time spent using ineffective medications may lead to irreversible disease progression. Despite availability of targeted treatments, only a minority of patients achieve sustained remission, and little evidence exists to direct the choice of biologic disease-modifying antirheumatic drugs in individual patients. Machine learning was used to identify a rule to predict the response to sarilumab and discriminate between responses to sarilumab versus adalimumab, with a focus on clinically feasible blood biomarkers. METHODS: The decision tree model GUIDE was trained using a data subset from the sarilumab trial with the most biomarker data, MOBILITY, to identify a rule to predict disease activity after sarilumab 200 mg. The training set comprised 18 categorical and 24 continuous baseline variables; some data were omitted from training and used for validation by the algorithm (cross-validation). The rule was tested using full datasets from four trials (MOBILITY, MONARCH, TARGET, and ASCERTAIN), focusing on the recommended sarilumab dose of 200 mg. RESULTS: In the training set, the presence of anti-cyclic citrullinated peptide antibodies, combined with C-reactive protein > 12.3 mg/l, was identified as the "rule" that predicts American College of Rheumatology 20% response (ACR20) to sarilumab. In testing, the rule reliably predicted response to sarilumab in MOBILITY, MONARCH, and ASCERTAIN for many efficacy parameters (e.g., ACR70 and the 28-joint disease activity score using CRP [DAS28-CRP] remission). The rule applied less to TARGET, which recruited individuals refractory to tumor necrosis factor inhibitors. The potential clinical benefit of the rule was highlighted in a clinical scenario based on MONARCH data, which found that increased ACR70 rates could be achieved by treating either rule-positive patients with sarilumab or rule-negative patients with adalimumab. CONCLUSIONS: Well-established and clinically feasible blood biomarkers can guide individual treatment choice. Real-world validation of the rule identified in this post hoc analysis is merited. CLINICAL TRIAL REGISTRATION: NCT01061736, NCT02332590, NCT01709578, NCT01768572.

In Silico Head-to-Head Comparison of Insulin Glargine 300 U/mL and Insulin Degludec 100 U/mL in Type 1 Diabetes.

Background: Second-generation long-acting insulin glargine 300 U/mL (Gla-300) and degludec 100 U/mL (Deg-100) provide novel basal insulin therapies for the treatment of type 1 diabetes (T1D). Both offer a flatter pharmacokinetic (PK) profile than the previous generation of long-acting insulins, thus improving glycemic control while reducing hypoglycemic events. This work describes an in silico head-to-head comparison of the two basal insulins on 24-h glucose profiles and was used to guide the design of a clinical trial. Materials and Methods: The Universities of Virginia (UVA)/Padova T1D simulator describes the intra-/interday variability of glucose-insulin dynamics and thus provides a robust bench-test for assessing glucose control for basal insulin therapies. A PK model describing subcutaneous absorption of Deg-100, in addition to the one already available for Gla-300, has been developed based on T1D clinical data and incorporated into the simulator. One hundred in silico T1D subjects received a basal insulin dose (Gla-300 or Deg-100) for 12 weeks (8 weeks uptitration, 4 weeks stable dosing) by morning or evening administration in a basal/bolus regimen. The virtual patients were uptitrated to their individual doses with two different titration rules. Results: The last 2-week simulated continuous glucose monitoring data were used to calculate various outcome metrics for both basal insulin treatments, with primary outcome being the percent time in glucose target (70-140 mg/dL). The simulations show no statistically significant difference for Gla-300 versus Deg-100 in the main endpoints. Conclusions: This work suggests comparable glucose control using either Gla-300 or Deg-100 and was used to guide the design of a clinical trial intended to compare second-generation long-acting insulin analogues.

Modeling Subcutaneous Absorption of Long-Acting Insulin Glargine in Type 1 Diabetes.

OBJECTIVE: Subcutaneous (sc) administration of long-acting insulin analogs is often employed in multiple daily injection (MDI) therapy of type 1 diabetes (T1D) to cover patient's basal insulin needs. Among these, insulin glargine 100 U/mL (Gla-100) and 300 U/mL (Gla-300) are formulations indicated for once daily sc administration in MDI therapy of T1D. A few semi-mechanistic models of sc absorption of insulin glargine have been proposed in the literature, but were not quantitatively assessed on a large dataset. The aim of this paper is to propose a model of sc absorption of insulin glargine able to describe the data and provide precise model parameters estimates with a clear physiological interpretation. METHODS: Three candidate models were identified on a total of 47 and 77 insulin profiles of T1D subjects receiving a single or repeated sc administration of Gla-100 or Gla-300, respectively. Model comparison and selection were performed on the basis of their ability to describe the data and numerical identifiability. RESULTS: The most parsimonious model is linear two-compartment and accounts for the insulin distribution between the two compartments after sc administration through parameter k. Between the two formulations, we report a lower fraction of insulin in the first versus second compartment (k = 86% versus 94% in Gla-100 versus Gla-300, p < 0.05), a lower dissolution rate from the first to the second compartment ([Formula: see text] versus 0.0008 min-1 in Gla-100 versus Gla-300, p << 0.001), and a similar rate of insulin absorption from the second compartment to plasma ([Formula: see text] versus 0.0016 min-1 in Gla-100 versus Gla-300, p = NS), in accordance with the mechanisms of insulin glargine protraction. CONCLUSIONS: The proposed model is able to both accurately describe plasma insulin data after sc administration and precisely estimate physiologically plausible parameters. SIGNIFICANCE: The model can be incorporated in simulation platforms potentially usable for optimizing basal insulin treatment strategies.

Incorporating Long-Acting Insulin Glargine Into the UVA/Padova Type 1 Diabetes Simulator for In Silico Testing of MDI Therapies.

OBJECTIVE: Glargine 100 U/mL (Gla-100) and 300 U/mL (Gla-300) are long-acting insulin analogs providing basal insulin supply in multiple daily injection (MDI) therapy of type 1 diabetes (T1D). Both insulins require extensive testing to arrive at the optimal dosing regimen, e.g., timing and amount. Here we aim at a simulation tool for evaluating benefits/risks of different dosing schemes and up-titration rules for both Gla-100 and Gla-300 before clinical testing. METHODS: A new pharmacokinetic (PK) model of both Gla-100 and Gla-300 was incorporated into the FDA-accepted University of Virginia/Padova T1D simulator: Specifically, a joint parameter distribution, built from PK parameter estimates, was used to generate individual PK parametrizations for each in silico subject. A virtual trial comparing Gla-100 vs. Gla-300 was performed and assessed against a clinical study to validate the glargine simulator. RESULTS: Like in vivo, in silico both insulins performed similarly with respect to glucose control: percent time of glucose between [80-140] mg/dL with Gla-100 vs. Gla-300 (primary endpoint) were 41.5 ± 1.1% vs. 39.0 ± 1.2% (P = 0.11) in silico, 31.0 ± 1.6% vs. 31.8 ± 1.5% (P = 0.73) in vivo. CONCLUSIONS: The glargine simulator reproduced the main findings of the clinical trial, proving its validity for testing MDI therapies. SIGNIFICANCE: In silico testing of MDI therapies can help designing clinical trials. Due to the more standardized settings in silico (e.g., standardized meals and strict adherence to titration rule), any potential treatment effect is reaching statistical significance in simulation vs. clinical trial.

Long-acting Insulin in Diabetes Therapy: In Silico Clinical Trials with the UVA/Padova Type 1 Diabetes Simulator.

The University of Virginia /Padova Type 1 Diabetes (TID) simulator has been widely used for testing artificial pancreas controllers, and, recently, novel insulin formulations and glucose sensors. However, a module describing the pharmacokinetics of the new long-acting insulin analogues is not available. The aim of this contribution is to reproduce multiple daily insulin injection (MDI) therapy, with insulin glargine 100 U/mL (Gla-100) as basal insulin, using the TID simulator. This was achieved by developing a model of Gla-100 and by incorporating it into the simulator. The methodology described here can be extended to other insulins, allowing an extensive in silico testing of different long-acting insulin analogues under various settings before starting human trials.

Improving Efficacy of Inhaled Technosphere Insulin (Afrezza) by Postmeal Dosing: In-silico Clinical Trial with the University of Virginia/Padova Type 1 Diabetes Simulator.

BACKGROUND: Technosphere(®) insulin (TI), an inhaled human insulin with a fast onset of action, provides a novel option for the control of prandial glucose. We used the University of Virginia (UVA)/Padova simulator to explore in-silico the potential benefit of different dosing regimens on postprandial glucose (PPG) control to support the design of further clinical trials. Tested dosing regimens included at-meal or postmeal dosing, or dosing before and after a meal (split dosing). METHODS: Various dosing regimens of TI were compared among one another and to insulin lispro in 100 virtual type-1 patients. Individual doses were identified for each regimen following different titration rules. The resulting postprandial glucose profiles were analyzed to quantify efficacy and the risk for hypoglycemic events. RESULTS: This approach allowed us to assess the benefit/risk for each TI dosing regimen and to compare results with simulations of insulin lispro. We identified a new titration rule for TI that could significantly improve the efficacy of treatment with TI. CONCLUSION: In-silico clinical trials comparing the treatment effect of different dosing regimens with TI and of insulin lispro suggest that postmeal dosing or split dosing of TI, in combination with an appropriate titration rule, can achieve a superior postprandial glucose control while providing a lower risk for hypoglycemic events than conventional treatment with subcutaneously administered rapid-acting insulin products.

MARS: computing three-dimensional alignments for multiple ligands using pairwise similarities.

The three-dimensional (3D) superimposition of molecules of one biological target reflecting their relative bioactive orientation is key for several ligand-based drug design studies (e.g., QSAR studies, pharmacophore modeling). However, with the lack of sufficient ligand-protein complex structures, an experimental alignment is difficult or often impossible to obtain. Several computational 3D alignment tools have been developed by academic or commercial groups to address this challenge. Here, we present a new approach, MARS (Multiple Alignments by ROCS-based Similarity), that is based on the pairwise alignment of all molecules within the data set using the tool ROCS (Rapid Overlay of Chemical Structures). Each pairwise alignment is scored, and the results are captured in a score matrix. The ideal superimposition of the compounds in the set is then identified by the analysis of the score matrix building stepwise a superimposition of all molecules. The algorithm exploits similarities among all molecules in the data set to compute an optimal 3D alignment. This alignment tool presented here can be used for several applications, including pharmacophore model generation, 3D QSAR modeling, 3D clustering, identification of structural outliers, and addition of compounds to an already existing alignment. Case studies are shown, validating the 3D alignments for six different data sets.

Development of in silico filters to predict activation of the pregnane X receptor (PXR) by structurally diverse drug-like molecules.

The pregnane X receptor (PXR), a member of the nuclear hormone superfamily, regulates the expression of several enzymes and transporters involved in metabolically relevant processes. The significant induction of CYP450 enzymes by PXR, in particular CYP3A4, might significantly alter the metabolism of prescribed drugs. In order to early identify molecules in drug discovery with a potential to activate PXR as antitarget, we developed fast and reliable in silico filters by ligand-based QSAR techniques. Two classification models were established on a diverse dataset of 434 drug-like molecules. A second augmented set allowed focusing on interesting regions in chemical space. These classifiers are based on decision trees combined with a genetic algorithm based variable selection to arrive at predictive models. The classifier for the first dataset on 29 descriptors showed good performance on a test set with a correct classification of both 100% for PXR activators and non-activators plus 87% for activators and 83% for non-activators in an external dataset. The second classifier then correctly predicts 97% activators and 91% non-activators in a test set and 94% for activators and 64% non-activators in an external set of 50 molecules, which still qualifies for application as a filter focusing on PXR activators. Finally a quantitative model for PXR activation for a subset of these molecules was derived using a regression-tree approach combined with GA variable selection. This final model shows a predictive r(2) of 0.774 for the test set and 0.452 for an external set of 33 molecules. Thus, the combination of these filters consistently provide guidelines for lowering PXR activation in novel candidate molecules.

Identification and Application of Antitarget Activity Hotspots to Guide Compound Optimization.

The optimization of a lead structure to a development candidate often requires removal of undesirable antitarget activities. To this end, we have developed an approach to extract antitarget activity hotspots from larger databases and to transfer this knowledge onto novel chemical series. These antitarget activity hotspots will be captured as pairs of informative molecules, which are chemically closely related, but differ significantly in biological activity. We illustrate the application of antitarget activity hotspots as informative compound pairs for the optimization of side effects in lead structures for relevant antitargets in pharmaceutical research. The use for prospective design requires establishing a structural link between known antitarget hotspot pairs and a new lead structure: we employ 3D-based similarity comparison for this task. The entire workflow serves as idea generator in early optimization. The feasibility of this approach is demonstrated in several optimization problems related to hERG inhibition, and CYP3A4 inhibition. Several structural examples demonstrate the ability of the 3D-shape searching to identify related scaffolds and the usefulness of the antitarget hotspot information to guide optimization by modulating the undesirable antitarget activity. Such a concept based on the analysis of local similarities and the transfer to 3D-related series is especially promising in those cases, where the construction of antitarget QSAR models fails to detect local SAR trends for guiding the next optimization cycle.

Sequence-derived three-dimensional pharmacophore models for G-protein-coupled receptors and their application in virtual screening.

G-protein-coupled receptors (GPCRs) comprise a large protein family of significant past and current interest of pharmaceutical research. X-ray crystallography and molecular modeling combined with site-directed mutagenesis studies suggest that most family A GPCRs share a small-molecule binding site located in the outer part of the seven-transmembrane (7TM) bundle. Here we describe an automated method to derive sequence-derived three-dimensional (3D) pharmacophore models capturing the key elements for addressing this binding site by a small-molecule ligand. We have generated structure-based pharmacophore models from 10 homology models and 3 X-ray structures of receptor-ligand complexes. These 13 pharmacophores have been dissected into 35 different single-feature pharmacophore elements, each associated with a sequence motif or chemoprint, describing its molecular interaction partner(s) in the receptor. Subsequently, the protein sequences of 270 GPCRs have been searched for the presence of chemoprints and the appropriate single-feature pharmacophores have been assembled into three- to seven-feature 3D-pharmacophore models for each human family A GPCR. These models can be applied for virtual screening and for the design of subfamily directed libraries. A case study demonstrates the successful application of this approach for the identification of potent agonists for the complement component 3a receptor 1 (C3AR1) by virtual screening.