Viral Kinetics Model of SARS-CoV-2 Infection Informs Drug Discovery, Clinical Dose, and Regimen Selection.

Quantitative systems pharmacology (QSP) has been an important tool to project safety and efficacy of novel or repurposed therapies for the SARS-CoV-2 virus. Here, we present a QSP modeling framework to predict response to antiviral therapeutics with three mechanisms of action (MoA): cell entry inhibitors, anti-replicatives, and neutralizing biologics. We parameterized three distinct model structures describing virus-host interaction by fitting to published viral kinetics data of untreated COVID-19 patients. The models were used to test theoretical behaviors and map therapeutic design criteria of the different MoAs, identifying the most rapid and robust antiviral activity from neutralizing biologic and anti-replicative MoAs. We found good agreement between model predictions and clinical viral load reduction observed with anti-replicative nirmatrelvir/ritonavir (Paxlovid®) and neutralizing biologics bamlanivimab and casirivimab/imdevimab (REGEN-COV®), building confidence in the modeling framework to inform a dose selection. Finally, the model was applied to predict antiviral response with ensovibep, a novel DARPin therapeutic designed as a neutralizing biologic. We developed a new in silico measure of antiviral activity, area under the curve (AUC) of free spike protein concentration, as a metric with larger dynamic range than viral load reduction. By benchmarking to bamlanivimab predictions, we justified dose levels of 75, 225, and 600 mg ensovibep to be administered intravenously in a Phase 2 clinical investigation. Upon trial completion, we found model predictions to be in good agreement with the observed patient data. These results demonstrate the utility of this modeling framework to guide the development of novel antiviral therapeutics.

Model-based dose selection to inform translational clinical oncology development of WNT974, a first-in-class Porcupine inhibitor.

WNT974 is a potent, selective, and orally bioavailable first-in-class inhibitor of Porcupine, a membrane-bound O-acyltransferase required for Wnt secretion, currently under clinical development in oncology. A phase I clinical trial is being conducted in patients with advanced solid tumors. During the dose-escalation part, various dosing regimens, including once or twice daily continuous and intermittent dosing at a dose range of 5-45 mg WNT974 were studied, however, the protocol-defined maximum tolerated dose (MTD) was not established based on dose-limiting toxicity. To assist in the selection of the recommended dose for expansion (RDE), a model-based approach was utilized. It integrated population pharmacokinetic (PK) modeling and exposure-response analyses of a target-inhibition biomarker, skin AXIN2 mRNA expression, and the occurrence of the adverse event, dysgeusia. The target exposure range of WNT974 that would provide a balance between target inhibition and tolerability was estimated based on exposure-response analyses. The dose that was predicted to yield an exposure within the target exposure range was selected as RDE. This model-based approach integrated PK, biomarker, and safety data to determine the RDE and represented an alternative as opposed to the conventional MTD approach for selecting an optimal biological dose. The strategy can be broadly applied to select doses in early oncology trials and inform translational clinical oncology drug development.

Efficient use of unlabeled data for protein sequence classification: a comparative study.

BACKGROUND: Recent studies in computational primary protein sequence analysis have leveraged the power of unlabeled data. For example, predictive models based on string kernels trained on sequences known to belong to particular folds or superfamilies, the so-called labeled data set, can attain significantly improved accuracy if this data is supplemented with protein sequences that lack any class tags-the unlabeled data. In this study, we present a principled and biologically motivated computational framework that more effectively exploits the unlabeled data by only using the sequence regions that are more likely to be biologically relevant for better prediction accuracy. As overly-represented sequences in large uncurated databases may bias the estimation of computational models that rely on unlabeled data, we also propose a method to remove this bias and improve performance of the resulting classifiers. RESULTS: Combined with state-of-the-art string kernels, our proposed computational framework achieves very accurate semi-supervised protein remote fold and homology detection on three large unlabeled databases. It outperforms current state-of-the-art methods and exhibits significant reduction in running time. CONCLUSION: The unlabeled sequences used under the semi-supervised setting resemble the unpolished gemstones; when used as-is, they may carry unnecessary features and hence compromise the classification accuracy but once cut and polished, they improve the accuracy of the classifiers considerably.

Tisagenlecleucel Model-Based Cellular Kinetic Analysis of Chimeric Antigen Receptor-T Cells.

Tisagenlecleucel is a chimeric antigen receptor-T cell therapy that facilitates the killing of CD19+ B cells. A model was developed for the kinetics of tisagenlecleucel and the impact of therapies for treating cytokine release syndrome (tocilizumab and corticosteroids) on expansion. Data from two phase II studies in pediatric and young adult relapsed/refractory B cell acute lymphoblastic leukemia were pooled to evaluate this model and evaluate extrinsic and intrinsic factors that may impact the extent of tisagenlecleucel expansion. The doubling time, initial decline half-life, and terminal half-life for tisagenlecleucel were 0.78, 4.3, and 220 days, respectively. No impact of tocilizumab or corticosteroids on the expansion rate was observed. This work represents the first mixed-effect model-based analysis of chimeric antigen receptor-T cell therapy and may be clinically impactful as future studies examine prophylactic interventions in patients at risk of higher grade cytokine release syndrome and the effects of these interventions on chimeric antigen receptor-T cell expansion.

Population Pharmacokinetics of Ceritinib in Adult Patients With Tumors Characterized by Genetic Abnormalities in Anaplastic Lymphoma Kinase.

Ceritinib is a second-generation selective and potent oral anaplastic lymphoma kinase (ALK) inhibitor approved for ALK-positive advanced non-small cell lung cancer previously treated with crizotinib. Population pharmacokinetic (PK) analysis was performed to describe the PK of ceritinib and was used to evaluate the covariate effects on systemic exposure at its label dose (750 mg orally once daily). Ceritinib concentration-time data from 4 clinical studies were described by a 1-compartment model with delayed first-order absorption and time-dependent elimination. The apparent clearance at steady state (CL/Fss ) was determined to increase with body weight and albumin but decrease with an increase in alanine aminotransferase. Japanese ethnicity appeared to significantly influence the apparent fractional turnover rate of the inhibited metabolic enzyme (kout ). No dose adjustment was necessary in patients with lower body weight or with preexisting mild hepatic impairment. The ceritinib steady-state exposure (AUCss ) at 750 mg increased by 8% (90% prediction interval [PI], 2-16) in non-Japanese Asians and 31% (90%PI, 17-44) in Japanese patients compared with that in white patients. Other covariates including sex, age, baseline Eastern Cooperative Oncology Group performance status, baseline total bilirubin, baseline estimated glomerular filtration rate, prior crizotinib treatment, and concomitant use of proton pump inhibitors had no statistically significant effect on ceritinib PK parameters. In conclusion, the nonlinear PK of ceritinib was described using a population-based approach in patients with ALK-positive tumors. None of the covariates assessed in this study were considered clinically relevant and therefore do not warrant dose adjustment.

Population pharmacokinetics of sonidegib (LDE225), an oral inhibitor of hedgehog pathway signaling, in healthy subjects and in patients with advanced solid tumors.

PURPOSE: Sonidegib (Odomzo) selectively inhibits smoothened and suppresses the growth of hedgehog pathway-dependent tumors. A population pharmacokinetic (PK) analysis of sonidegib in healthy subjects and patients with advanced solid tumors was conducted to characterize PK, determine variability, and estimate covariate effects. METHODS: PK data from five phase 1 or 2 studies (N = 436) in the dose range from 100 to 3000 mg were analyzed using NONMEM. A two-compartment base model with first-order absorption, lag time, linear elimination, and bioavailability that decreased with dose was updated to describe the PK of sonidegib. Covariate analyses were performed and were incorporated into the population PK full model. RESULTS: The base and full models were robust with a good fit to the study data. Population-predicted geometric means (inter-individual variability, CV%) of apparent oral clearance, apparent volume of distribution at steady state, accumulation ratio, and elimination half-life were 9.5 L/h (71.4 %), 9163 L (74.9 %), 21 (131 %) and 29.6 days (109 %). Clinically relevant covariate effects were: A high-fat meal increased sonidegib bioavailability fivefold, healthy volunteers had threefold higher clearance, sonidegib bioavailability decreased with increasing dose levels, and PPI coadministration reduced sonidegib bioavailability by 30 %. Sonidegib PK was not significantly impacted by baseline age, weight, total bilirubin, alanine aminotransferase, albumin, creatinine clearance, gender, and ethnicity (Western countries versus Japanese). CONCLUSION: No dose adjustment is needed for mild hepatic impairment, mild and moderate renal impairment, age, weight, gender, or ethnicity. This population PK model adequately characterizes sonidegib PK characteristics and can be used for various simulations and applications.

Protein homology detection with biologically inspired features and interpretable statistical models.

Computational classification of proteins using methods such as string kernels and Fisher-SVM has demonstrated great success. However, the resulting models do not offer an immediate interpretation of the underlying biological mechanisms. In this work, we propose a biologically motivated feature set combined with a sparse classifier, based on a small subset of positions and residues in protein sequences, for protein superfamily detection and show the performance of our models is comparable to that of the state-of-the-art methods on a benchmark dataset. The set of sparse critical features discovered by the models is consistent with the confirmed biological findings.

Fast and accurate multi-class protein fold recognition with spatial sample kernels.

Establishing structural or functional relationship between sequences, for instance to infer the structural class of an unannotated protein, is a key task in biological sequence analysis. Recent computational methods such as profile and neighborhood mismatch kernels have shown very promising results for protein sequence classification, at the cost of high computational complexity. In this study we address the multi-class sequence classification problems using a class of string-based kernels, the sparse spatial sample kernels (SSSK), that are both biologically motivated and efficient to compute. The proposed methods can work with very large databases of protein sequences and show substantial improvements in computing time over the existing methods. Application of the SSSK to the multi-class protein prediction problems (fold recognition and remote homology detection) yields significantly better performance than existing state-of-the-art algorithms.