AAPS Journal最新文献

Accurate assessment of tissue distribution for oligonucleotide therapeutic (ONT) drug candidates is essential for understanding pharmacokinetic behavior and predicting therapeutic efficacy. ONTs present a unique challenge with their rapid systemic clearance coupled with prolonged tissue retention, making comprehensive tissue concentration evaluation critical for successful drug development. The IQ Consortium Tissue Concentration Working Group surveyed member companies about their current tissue concentration assessment methods to understand industry practices and identify areas for improvement. Most companies reported that ONTs still represent a relatively small portion of their pre-candidate selection portfolios, reflecting the evolving nature of this therapeutic modality. siRNAs dominated development efforts across surveyed organizations, followed by antisense oligonucleotides, indicating clear therapeutic class preferences within the industry. Assessment strategies varied considerably across organizations, highlighting different approaches to resource allocation and risk management. While some companies routinely evaluate tissue concentrations for all ONT programs regardless of indication or target, others take a more selective, program-dependent approach based on compound characteristics and therapeutic objectives. Despite this strategic variability, there was universal reliance on LC-MS for quantification, often supplemented with qPCR/RT-qPCR and hybridization assays for comprehensive analytical coverage. All surveyed companies integrate tissue concentration data into translational pharmacokinetic modeling efforts, yet few have adopted physiologically-based pharmacokinetic (PBPK) models as standard practice. Companies recognize the value of improving ONT tissue distribution assessment through standardized methodology tailored to specific oligonucleotide classes.

COVID-19 disease outcomes can vary considerably among infected patients. Most studies have focused on patients with severe COVID-19. However, investigations of asymptomatic infection can provide insights into patient-specific immunological features that protect patients from COVID-19 symptoms. Recent studies have shown an association between common human leukocyte antigen (HLA) alleles and asymptomatic COVID-19 infections. Here we utilize machine learning in conjunction with explainable AI (XAI) to identify alleles in five HLA loci that can be either protective or put the patient at risk for symptomatic COVID-19. Data from the public online HLA-COVID database (1946 samples) was used for training and validating multiple ML classification models to identify the top performing model. The model was then further processed with XAI via SHAP (SHapley Additive exPlanations) to identify the protective and high-risk HLA alleles. This study provides a proof-of-concept study for utilizing machine learning to provide valuable insights for COVID-19 patients. These findings can be translated into clinical algorithms to help physicians personalize COVID-19 treatments and achieve better clinical outcomes.

Immunogenicity testing for anti-drug antibodies (ADAs) is crucial in therapeutic protein development, yet current quasi-quantitative assays struggle to accurately measure ADAs when the antibodies have different binding strength (affinities) or due to heterogeneity of ADAs and residual drug interference. While traditional QC-based assay development is limited by the lack of representative ADA reference standards, we propose Model-Informed Assay Development (MIAD) as a transformative solution. MIAD mathematically simulates complex analyte-reagent interactions to identify optimal conditions for signal-generating analyte-reagent complex (ARC) formation, enabling scientifically sound assay optimization independent of positive controls. Our findings demonstrate that optimal sample dilution and reagent concentrations can overcome drug interference and improved detection of antibodies (ADAs) with different binding strengths. This work applies MIAD to address critical ADA assay challenges: drug tolerance and affinity-dependent detectability. We tested MIAD's prediction in three real world case studies and found strong agreement. Our findings show that optimized sample dilutions and reagent concentrations effectively overcome drug interference and affinity differences, enhancing ADA detectability and recovery. MIAD also helps understanding whether a hook-shaped curve is due to a prozone effect or drug interference, guiding the development of unbiased assays crucial for accurate S/N-based magnitude estimation.

Pimozide and PSSI-51 are under study for their potential glucose-lowering effects in type 2 diabetes, through their abilities to inhibit succinyl-CoA:3-ketoacid CoA transferase, the rate-limiting enzyme of ketone oxidation. To understand their pharmacokinetics, they were administered to Sprague Dawley male and female rats after standard and high-fat diets. Initially five rats each were given 10 mg/kg of each agent orally, and using serial blood withdrawals from jugular vein cannulas, the blood samples were assayed for drug and basic pharmacokinetic data estimated using compartmental analysis. A separate group of male and female rats were given the same single dose after either 10 (pimozide) or 13 (PSSI-51) weeks of feeding with a standard or high-fat diet, followed by two blood samples after each dose from the saphenous vein. Bayesian forecasting in conjunction with the mean and variance of pharmacokinetic parameters and assay coefficient of variation, was used to estimate the pharmacokinetic parameters in these rats. The two drugs differed in their optimal pharmacokinetic model (pimozide one compartment, PSSI-51 two compartment). Both drugs possessed a high volume of distribution (Vd/F), but the oral clearance (CL/F) of PSSI-51 was much higher than that of pimozide, in line with earlier observations using rat microsomal experiments. The high-fat diet significantly reduced the oral CL and Vd of PMZ in both male and female rats, whereas no such effect was observed for PSSI-51.

Glycosylation is a critical quality attribute of certain therapeutic proteins, influencing efficacy, safety, and pharmacokinetics. This study analyzed glycan characterization data and drug substance release specifications from 209 Biologics License Applications (BLAs) approved by the U.S. Food and Drug Administration (FDA) through May 2025. Ten predominant Fc N-glycans were identified across IgG antibodies expressed by CHO, NS0, and Sp2/0 cell lines, with six glycans common to all systems. Five low-abundance afucosylated glycans (< 10%) were tightly controlled within drug substance release specifications for antibodies with Fc effector functions, although acceptance criteria varied across products. Glycan profiles were strongly dependent on the expression system: CHO-derived antibodies predominantly contained human-compatible glycan structures, whereas NS0 and Sp2/0 antibodies introduced non-human epitopes. Fc fusion proteins exhibited higher branching and sialylation compared with conventional IgG antibodies. Notably, analysis of product labels revealed that Fc effector functions were described exclusively as in vitro observations or proposed mechanisms, with limited clinical validation. Overall, these findings establish a comprehensive benchmark for glycan profiles of FDA-approved therapeutic antibodies and underscore the need for harmonized control strategies and stronger correlation between in vitro Fc assays and clinical outcomes.

To evaluate the strengths and limitations of retrieval-augmented generative (RAG) artificial intelligence (AI) for natural language querying of biologics immunogenicity data. The package inserts for drugs approved with biologics license applications (BLA) were retrieved from DailyMed ( https://dailymed.nlm.nih.gov/dailymed/ ). The RAG system integrated natural language processing, retrieval, and large language model (LLM) components. ChatGPT, Gemini, DeepSeek, and Llama were queried with five clinical pharmacology-focused questions on factors influencing anti-drug antibody (ADA) incidence and tolerability, including effects of target protein, administration route, and citrate excipients. Outputs were assessed for relevance, faithfulness, and domain-specific accuracy. The dataset included 663 biologics, of which 206 (31.1%) were monoclonal antibodies. The RAG system retrieved relevant contexts for all queries, but several contexts contained inaccuracies related to the presence of non-antibody protein drugs. All four LLMs generated coherent summaries and identified determinants of ADA incidence, such as drug type, assay methods, and concomitant therapy. All models found that injection-site pain occurred with some protein therapeutics containing citrate excipients, and that evidence for a direct causal role of citrate was mixed. Comparative evaluation showed that LLM outputs were generally relevant and faithful to the source text, with variation in the level of detail and comprehensiveness across models. Domain-specific evaluations indicated that responses accurately identified trends in immunogenicity and highlighted the knowledge gaps. While RAG-based systems can retrieve and synthesize immunogenicity assessments from multiple source documents, significant limitations were noted in this use case. The effectiveness of the retriever can limit RAG performance and warrant refinement.

An assessment of parallelism is critical for biomarker assays to confirm whether the assay recognizes the endogenous analyte similarly to the calibrator, the suitability of a surrogate calibrator matrix and the potential need for a minimal required dilution. While the importance of parallelism has been raised in numerous publications there remains a lack of detail on how to conduct and interpret parallelism experiments, as well as some confusion between parallelism, dilution linearity, and spike recovery experiments. This best practice paper provides a detailed discussion of the reasons for conducting parallelism, as well as recommendations for when to conduct parallelism experiments, the number of samples needed, the selection of appropriate surrogate matrices, the interpretation of parallelism data, including graphical and statistical methods, and parallelism results reporting. It emphasizes the need for continuous evaluation of parallelism throughout the assay life cycle to ensure reliable measurement of the desired analyte within the context of use. Finally, a number of short case studies are provided to illustrate the application and interpretation of parallelism.

Innate immune response modulating impurities (IIRMI) with adjuvant potential have emerged as important factors in the immunogenicity risk assessment of protein, peptide, and oligonucleotide therapeutics, particularly for follow-on products where minimal or no clinical studies are available. To assess the impact of differences in impurities on specific cell types, we developed a new IIRMI assay termed multiplexed immunophenotyping for innate activation assessment (MIIAA) that employs spectral flow cytometry to capture single-cell responses to drug products and potential impurities. This technique introduces a new live fluorescent cell barcoding platform that enables sample multiplexing for homogeneous staining with a single fluorescent antibody cocktail composed of identity and activation markers that are acquired simultaneously with a five laser Cytek Aurora. Samples are digitally reassigned to their original testing conditions by positive and negative gating of barcode dyes. Cellular subsets are identified by dimensionality reduction of surface markers with UMAP then gated using cell-specific markers. Here we use trace levels of TLR3, 7/8 and 9 agonists (Poly(I:C), R848, and CpG ODN) to characterize specific responses in B cells, monocytes, cDC and pDC. Importantly, MIIAA captures single-cell responses to nucleic acid impurities in the presence of therapeutic oligonucleotides or monoclonal antibodies with high sensitivity. Taken together, MIIAA offers a powerful immunophenotyping tool to characterize single-cell responses to drug products and potential immunomodulatory impurities that may find utility in drug pipelines to characterize the impact of therapeutics on specific immune cells and to interrogate immunogenic or immunomodulatory risk in comparisons between reference and follow-on products.

Obesity significantly alters drug disposition and contributes to large inter-individual variability in pharmacokinetics (PK). The virtual-twin concept is increasingly used to support model-informed precision dosing in specific populations. In this study, physiologically-based pharmacokinetic models linked with virtual twins (VT-PBPK) have been developed and applied to predict the PK of midazolam and digoxin in patients with obesity (n = 15) and severe obesity (n = 22). The first step of the individualization included basic demographic data with lean liver volume. In the second step, individual serum creatinine, albumin, and hepatic CYP3A4/5, UGT1A4 and P-gp abundance quantified from liver biopsies in the same individuals, were integrated within models. Substrate specific improvements were presented via the stepwise individualization. The final (Step 2) VT-PBPK models predicted midazolam AUC_0-inf,iv within 2-fold for 86% of the individuals (geometric mean fold error, GMFE = 1.5; 95% confidence interval (CI95) = 1.36-1.78), with 36% within the 0.8 to 1.25-fold of the observed values. For digoxin, 97% of C_max and AUC_0-24 values were predicted within 2-fold of the observed data (GMFE = 1.25; CI95 = 1.19-1.33), with 59% of predicted values within the 0.8-1.25-fold range. In the case of digoxin, the prediction accuracy was higher for patients with severe obesity (60% of C_max and AUC_0-24 values within the 1.25-fold range); no clear trends were evident for midazolam. This is the first study that applied the VT-PBPK modelling approach in patients with obesity. It highlights the potential of this approach to predict the PK of other CYP3A and P-gp substrates to support individual dose optimization in this population.

In 2024, 3-Chloromethcathinone (3-CMC) accounted for over 63% of all New Psychoactive Substances seized in Europe, yet its human pharmacology remains poorly understood. This observational, uncontrolled, naturalistic study involved 16 regular psychostimulant users to evaluate and compare 3-CMC metabolism, and distribution in urine and oral fluid (OF) following oral and intranasal administration. Two groups, each consisting of 8 participants (6 males, 2 females) self-administered 3-CMC in two separate sessions: 100-150 mg orally and 60-80 mg intranasally. Urine was collected in two pooled intervals (0-2 h and 2-5 h). Samples were analyzed via four untargeted HPLC-HRMS/MS methods in full MS and ddMS² to characterize the unknown metabolites supported by Compound Discoverer™ software with an established workflow. The data were grouped into four groups concerning the route of administration and the time intervals and the average area were statistically compared with a one-way ANOVA. The parent drug was detected in all the samples at different levels. In total, nine metabolites were observed, of those 4 were phase I and 5 phase II metabolites. Considering the route of administration, distinct metabolic patterns emerged: three metabolites, including two N-acetylated forms and a carboxylated metabolite, were found only after oral intake, suggesting N-acetylation occurs primarily via this route. In contrast, β-OH-3-CMC accumulated more after intranasal use. Furthermore, 3-CMC N- glucuronidation was hypothesized for the first time. These findings indicate that the administration route significantly influences 3-CMC metabolism, highlighting the need for tailored forensic and toxicological assessments.