Pharmacokinetic/Pharmacodynamic Translation and Model‐Informed Drug Development for Oligonucleotide Therapeutics

Abstract

Small interfering RNAs (siRNAs) and antisense oligonucleotides (ASOs) have emerged as clinically validated therapeutic modalities, with approvals and late‐stage development programs spanning rare genetic, neurologic, cardiovascular, metabolic, and infectious diseases. Despite these advances, oligonucleotide development presents unique challenges compared with small molecules and biologics, including rapid plasma distributive clearance, nuclease‐mediated degradation, limited extrahepatic distribution, and prolonged pharmacodynamic effects driven by tissue retention and intracellular mechanisms such as RNA‐induced silencing complex loading or RNase H‐mediated activity. Consequently, tissue disposition and intracellular pharmacology most often govern therapeutic response more directly than plasma exposures alone, complicating conventional approaches to dose selection, efficacy prediction, and safety assessment. Model‐informed drug development (MIDD) offers a quantitative framework to address these challenges through integration of preclinical, translational, and clinical data into empirical, mechanistic, and systems‐level models. This review summarizes current and emerging MIDD applications in oligonucleotide therapeutics, with primary emphasis on siRNAs and complementary insights from ASOs. Approaches discussed include empirical and semi‐mechanistic pharmacokinetic/pharmacodynamic (PK/PD) models, physiologically based pharmacokinetic (PBPK) frameworks describing tissue‐selective biodistribution and intracellular disposition, and quantitative systems pharmacology (QSP) models linking molecular target modulation with downstream biologic and clinical responses. Collectively, these approaches have supported cross‐species translation; human dose selection; clinical trial optimization; and mechanistic understanding of oligonucleotide absorption, distribution, metabolism, excretion, and pharmacology. Finally, we discuss future opportunities and remaining challenges for MIDD in oligonucleotide therapeutics, including enabling extrahepatic delivery, characterizing interindividual variability, and integrating systems‐level and data‐driven approaches to improve translational predictability and accelerate development of next‐generation oligonucleotide medicines.

Introduction

RNA‐ and nucleic acid‐based therapeutics, including small interfering RNAs (siRNAs) and antisense oligonucleotides (ASOs), have emerged as a transformative class of medicines capable of modulating gene expression with high specificity. ¹ By directly targeting RNA, these modalities enable therapeutic intervention against disease drivers previously considered “undruggable,” thereby expanding precision medicine approaches across both rare and prevalent diseases. ² Regulatory approvals spanning diverse indications underscore the growing clinical maturity of oligonucleotide therapeutics (Table 1), with numerous additional compounds currently in clinical development. ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ Recent FDA approvals further highlight their rapid clinical expansion, including plozasiran and fitusiran in 2025, as well as the expanded indication for olezarsen, which was initially approved for familial chylomicronemia syndrome in 2024 and subsequently approved for severe hypertriglyceridemia in 2026. Their sequence‐driven specificity, modular design, prolonged pharmacodynamic activity, and increasingly scalable manufacturing have further broadened the applicability of oligonucleotides across diverse therapeutic areas including metabolic, infectious, and neurodegenerative diseases. ⁹ , ¹⁰

Table 1.

US FDA‐Approved siRNA and ASO Therapeutics as of 2026

Drug (Trade Name)	Year	Platform	Route	Target Tissue	Indication(s)
Small Interfering RNA (siRNA)
Plozasiran (Redemplo)	2025	GalNAc conjugate	SC	Liver	Familial chylomicronemia syndrome
Givosiran (Givlaari)	2019	GalNAc conjugate	SC	Liver	Acute hepatic porphyria
Lumasiran (Oxlumo)	2020	GalNAc conjugate	SC	Liver	Primary hyperoxaluria type 1
Inclisiran (Leqvio)	2021	GalNAc conjugate	SC	Liver	Primary hypercholesterolemia/mixed dyslipidemia
Vutrisiran (Amvuttra)	2022/2025	GalNAc conjugate	SC	Liver	hATTR polyneuropathy/ATTR cardiomyopathy
Nedosiran (Rivfloza)	2023	GalNAc conjugate	SC	Liver	Primary hyperoxaluria type 1
Fitusiran (Qfitlia)	2025	GalNAc conjugate	SC	Liver	Hemophilia A or B
Patisiran (Onpattro)	2018	Lipid nanoparticle (LNP)	IV	Liver	hATTR amyloidosis with polyneuropathy
Antisense Oligonucleotides (ASOs)
Inotersen (Tegsedi)	2018	Phosphorothioate ASO	SC	Liver	hATTR polyneuropathy
Eplontersen (Wainua)	2023	GalNAc‐conjugated ASO	SC	Liver	hATTR polyneuropathy
Mipomersen (Kynamro)	2013	Phosphorothioate ASO	SC	Liver	Homozygous familial hypercholesterolemia
Olezarsen (Tryngolza)	2024/2026	GalNAc‐conjugated ASO	SC	Liver	Familial chylomicronemia syndrome / severe hypertriglyceridemia
Donidalorsen (Dawnzera)	2025	GalNAc‐conjugated ASO	SC	Liver	Hereditary angioedema prophylaxis
Eteplirsen (Exondys 51)	2016	PMO‐ASO	IV	Skeletal muscle	Duchenne muscular dystrophy (exon 51 skipping)
Casimersen (Amondys 45)	2021	PMO‐ASO	IV	Skeletal muscle	Duchenne muscular dystrophy (exon 45 skipping)
Viltolarsen (Viltepso)	2020	PMO‐ASO	IV	Skeletal muscle	Duchenne muscular dystrophy (exon 53 skipping)
Golodirsen (Vyondys 53)	2019	PMO‐ASO	IV	Skeletal muscle	Duchenne muscular dystrophy (exon 53 skipping)
Nusinersen (Spinraza)	2016	Phosphorothioate ASO	Intrathecal	CNS	Spinal muscular atrophy
Tofersen (Qalsody)	2023	Phosphorothioate ASO	Intrathecal	CNS	SOD1‐associated amyotrophic lateral sclerosis
Fomivirsen (Vitravene)	1998	Phosphorothioate ASO	Intravitreal	Eye	CMV retinitis

CMV, cytomegalovirus; CNS, central nervous system; GalNAc, N‐acetylgalactosamine; hATTR, hereditary transthyretin; IV, intravenous; LNP, lipid nanoparticle; PMO‐ASO, phosphorodiamidate morpholino antisense oligonucleotide; SC, subcutaneous; SOD1, superoxide dismutase.

Although siRNAs and ASOs differ in structure and their mechanism(s) of action, both ultimately reduce target gene expression through sequence‐directed RNA targeting pathways (Figure 1). Their pharmacologic activity is governed by a coordinated cascade encompassing tissue uptake, cellular internalization, intracellular trafficking, subcellular localization, and target RNA degradation. siRNAs are typically double‐stranded RNAs of ∼21–23 nucleotides in length that mediate RNA interference through incorporation of the antisense strand into the RNA‐induced silencing complex (RISC), ¹¹ enabling sequence‐specific cleavage of target messenger RNA (mRNA). ¹² In contrast, ASOs are generally single‐stranded oligonucleotides that primarily act through RNase H‐mediated degradation of target RNA or steric modulation of splicing and translation. ¹³ Consequently, oligonucleotide pharmacology is governed not only by systemic disposition, but also by intracellular processes including receptor‐ or adsorptive‐mediated cellular uptake, endosomal trafficking and escape, target engagement, and downstream mRNA and protein turnover. ¹⁴

Figure 1.

Schematic overview of the distinct and shared pathways governing hepatocyte uptake, intracellular disposition, and pharmacologic activity of GalNAc‐conjugated small interfering RNAs (GalNAc‐siRNAs), GalNAc‐conjugated antisense oligonucleotides (GalNAc‐ASOs), and unconjugated phosphorothioate antisense oligonucleotides (PS‐ASOs). GalNAc‐conjugated oligonucleotides undergo ASGPR‐mediated endocytosis, whereas unconjugated PS‐ASOs are internalized primarily through adsorptive endocytosis, likely mediated through multiple low‐affinity surface proteins. Following endosomal (or endolysosomal) trafficking and escape, GalNAc‐siRNAs are loaded into the RNA‐induced silencing complex (RISC) in the cytoplasm, resulting in catalytic target mRNA cleavage, while ASOs recruit RNase H1 to mediate degradation of complementary target mRNA within the nucleus and/or cytoplasm. Despite these distinct uptake and intracellular trafficking pathways, all three modalities ultimately reduce target protein expression through sequence‐specific degradation of target mRNA. Ago2, Argonaute‐2; ASGPR, asialoglycoprotein receptor; ASO, antisense oligonucleotide; GalNAc, N‐acetylgalactosamine; mRNA, messenger ribonucleic acid; PS, phosphorothioate; RISC, RNA‐induced silencing complex; RNase H1, ribonuclease H1; siRNA, small interfering RNA.

Historically, clinical translation of oligonucleotide therapeutics was limited by several unfavorable physicochemical and biologic properties. Unmodified oligonucleotides undergo rapid degradation by endo‐ and exonucleases and may be subject to efficient renal clearance. ¹⁵ Their relatively large size and highly anionic phosphate backbone restrict passive cellular uptake, ¹⁰ while endolysosomal processing following endocytosis severely limits cytosolic availability, with only a small fraction of internalized molecules escaping into the cytoplasm. ¹⁶ In addition, exogenous, unmodified nucleic acids may activate innate immune pathways, creating potential immunogenicity and tolerability concerns. ¹⁷ , ¹⁸

Over the past two decades, substantial advances in oligonucleotide chemistry and delivery technologies have addressed many of these limitations and enabled broader clinical success. Chemical modifications to the sugar‐phosphate backbone, including 2′‐O‐methyl and 2′‐fluoro substitutions, improve nuclease resistance, metabolic stability, and tolerability. ¹⁹ , ²⁰ , ²¹ Parallel advances in delivery technologies, particularly lipid nanoparticles (LNPs) and ligand conjugation strategies, have substantially improved in vivo disposition and pharmacologic activity. ²¹ , ²² Among these, N‐acetylgalactosamine (GalNAc) conjugation has emerged as the most widely adopted strategy for hepatocyte‐selective delivery. ²³ , ²⁴ , ²⁵ By leveraging high‐affinity binding to the asialoglycoprotein receptor (ASGPR), abundantly expressed on hepatocytes, GalNAc‐conjugated siRNAs and ASOs undergo highly efficient receptor‐mediated uptake following subcutaneous (SC) administration, in contrast to unconjugated phosphorothioate ASOs, which are internalized predominantly through adsorptive endocytosis ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ (Figure 1).

Following receptor‐mediated or adsorptive endocytosis, oligonucleotides undergo complex intracellular trafficking, with the vast majority remaining sequestered within endosomal and lysosomal compartments and only ∼1% successfully escaping into the cytoplasm. ²⁹ Endosomal escape is now recognized as a major rate‐limiting step governing pharmacologic activity. ³⁰ , ³¹ Following escape, siRNAs undergo Argonaute‐2 (Ago2)‐mediated loading into the RISC, where the passenger (sense) strand is discarded and the guide (antisense) strand directs catalytic cleavage of complementary target mRNA. In contrast, ASOs distribute to both cytoplasmic and nuclear compartments, ³² where they recruit RNase H1 to degrade target RNA or modulate RNA processing. Together, these intracellular processes produce efficient but capacity‐limited tissue distribution, prolonged intracellular retention, and pharmacodynamic effects that frequently persist long after plasma concentrations have declined.

Despite these advances, significant challenges and opportunities remain in oligonucleotide clinical pharmacology and development. Their pharmacokinetic/pharmacodynamic (PK/PD) relationships are inherently non‐traditional, reflecting a multistep pharmacologic cascade in which plasma exposure, tissue delivery, intracellular trafficking, target engagement, and downstream protein turnover occur over markedly different timescales. Consequently, conventional plasma‐based exposure metrics are often insufficient to describe pharmacologically relevant concentrations at tissue sites of action. ¹⁰ , ³³ In addition, most systemically administered oligonucleotides remain predominantly liver‐directed, limiting broader tissue targeting from an efficacy viewpoint. Variabilities in uptake, intracellular processing, and target biology may further complicate exposure–response characterization, particularly in special populations such as pediatrics or patients with organ impairment. ³⁰ These features create unique considerations for dose selection, efficacy prediction, and safety assessment within traditional clinical pharmacology paradigms.

Model‐informed drug development (MIDD) provides a quantitative framework to address these complexities through integration of in vitro, preclinical, translational, and clinical data using empirical, mechanistic, and systems‐level modeling approaches. ³⁴ For oligonucleotide therapeutics, MIDD approaches are uniquely positioned to characterize multiscale processes governing drug disposition and response, including tissue uptake, intracellular trafficking, target engagement, and downstream biologic effects. Empirical and semi‐mechanistic PK/PD models, physiologically based pharmacokinetic (PBPK) frameworks, and quantitative systems pharmacology (QSP) approaches have progressively been applied to characterize oligonucleotide PK/PD relationships, support translational predictions, optimize first‐in‐human (FIH) dose selection, and inform clinical trial design. ³⁵ Recent regulatory guidance ³⁶ and consortium‐led industry recommendations ³⁷ have largely focused on foundational clinical pharmacology considerations including PK/PD characterization, organ impairment, immunogenicity risk, and drug–drug interactions, while the role of quantitative modeling, though appreciated, remains comparatively underdeveloped within current regulatory paradigms. This represents an important opportunity for broader integration of MIDD approaches to improve translational predictability and support more efficient oligonucleotide development.

In this review, we provide an overview of MIDD applications and opportunities in oligonucleotide therapeutics. We first discuss key challenges in oligonucleotide research and development, including their unique multiscale PK/PD relationships, tissue‐selective delivery, bioanalytical limitations, exposure–safety relationships, interspecies translatability, and growing interest in new approach methodologies (NAMs), while offering concurrent perspectives on how MIDD can help address these challenges. We then review evolving MIDD strategies across preclinical and clinical development, highlighting applications of empirical, semi‐mechanistic, PBPK, and QSP approaches in translational and clinical settings. Finally, we discuss emerging opportunities and remaining challenges for MIDD, including extrahepatic targeting, interindividual variability, NAMs, and integration of systems‐level and data‐driven approaches to advance the safe, effective, and efficient development of next‐generation oligonucleotide therapeutics.

Distinct Clinical Pharmacology and Translational Challenges of Oligonucleotide Therapeutics

Multiscale Pharmacology and Unique PK/PD Relationships

One of the defining challenges in oligonucleotide therapeutics development is the fundamentally unconventional and multiscale nature of their PK/PD relationships. Unlike small molecules or monoclonal antibodies, where systemic exposure often serves as a useful surrogate driver of pharmacologic response, oligonucleotides, particularly GalNAc‐conjugated siRNAs, exhibit pronounced temporal disconnects between plasma pharmacokinetics and tissue pharmacodynamics. ³³ Following administration, plasma concentrations decline rapidly and may become undetectable within hours, while substantial uptake and retention occur within target tissues such as the liver through ASGPR‐mediated endocytosis. ²⁷ , ²⁸ However, tissue exposure alone incompletely reflects pharmacologically active drug, since only a very small fraction of internalized oligonucleotide ultimately escapes endosomal sequestration to engage intracellular target machinery. ¹⁶ Consequently, pharmacodynamic response is governed by a cascade of sequential and often rate‐limiting processes spanning tissue uptake, intracellular trafficking, RISC loading or RNase H engagement, target mRNA degradation, and downstream protein turnover. ¹¹ , ¹² , ¹³ , ²⁶ , ³⁸ These relationships, and their connection to diverse MIDD approaches, are summarized in Figure 2.

Figure 2.

Unique PK/PD characteristics and resulting temporal profiles of oligonucleotide therapeutics, together with MIDD approaches that may be applied to characterize their PK/PD properties, optimize dose and regimen, predict efficacy and safety, enable interspecies translation, and guide clinical study design. AI/ML, artificial intelligence/machine learning; BLQ, below limit of quantification; IVIVC, in vitro–in vivo correlations; K‐PD, kinetic‐pharmacodynamic; MBMA, model‐based meta‐analysis; MIDD, model‐informed drug development; MoA, mechanism of action; mRNA, messenger RNA; NCA, noncompartmental analysis; PBPK, physiologically based pharmacokinetics; PBPK‐PD, physiologically based pharmacokinetics‐pharmacodynamics; PK‐PD, pharmacokinetics‐pharmacodynamics; QSP, quantitative systems pharmacology; RISC, RNA‐induced silencing complex; SC, subcutaneous; siRNA, small interfering RNA. The conceptual organization and visual framework of this figure were inspired by Zhou et al ¹³³ and have been substantially adapted to illustrate the unique PK/PD characteristics and MIDD applications for oligonucleotide therapeutics.

For siRNA therapeutics, the distinction between plasma PK, total tissue exposure, and pharmacologically active intracellular species is particularly important. Following hepatocellular uptake, most internalized siRNA remains confined within endosomal and lysosomal compartments, whereas only a small fraction (∼1%) undergoes cytosolic release and RISC loading. ¹⁶ Despite representing only a minor proportion of total hepatic siRNA, RISC‐loaded siRNA serves as the principal driver of sustained catalytic mRNA degradation and downstream pharmacologic activity. ²⁶ , ³⁸ As conceptually illustrated in Figures 2 and 3, clinically relevant biomarkers and efficacy endpoints therefore reflect the integrated effects of cascading, temporally linked biologic processes, including hepatocyte uptake, intracellular sequestration, endolysosomal escape, RISC engagement, mRNA degradation, and target protein turnover. ²⁴ , ²⁵ , ³⁹ Each step may independently influence the magnitude and duration of pharmacodynamic response, resulting in delayed and prolonged biologic effects that can persist long after plasma drug concentrations become undetectable. It is now well understood that slow, rate‐limiting release of a small fraction of siRNA from acidic endolysosomal compartments, where oligonucleotides may remain stabilized and degrade gradually, enables continuous RISC loading and sustained target gene silencing over extended periods at typical clinical doses. ¹⁰ , ³⁰ This phenomenon provides an important mechanistic basis for the infrequent (e.g., quarterly or longer) dosing regimens achievable with many GalNAc‐siRNA therapeutics.

Figure 3.

Simplified kinetic representation of the GalNAc‐siRNA pharmacologic cascade in hepatocytes. Following ASGPR‐mediated uptake, chemically modified siRNAs undergo prolonged residence with slow release from within acidic endolysosomal compartments before cytosolic RISC loading, target mRNA knockdown, and subsequent reduction in protein expression, with each transition governed by distinct rate constants and timescales. The bold and dashed arrows emphasize mechanistically important features of the oligonucleotide PK/PD cascade, namely the frequently rate‐limiting intracellular disposition step and additional delay that depends on target protein turnover. ASGPR, asialoglycoprotein receptor; GalNAc, N‐acetylgalactosamine; mRNA, messenger ribonucleic acid; RISC, RNA‐induced silencing complex; siRNA, small interfering RNA.

An additional layer of complexity arises from the indirect nature of many oligonucleotide‐mediated pharmacodynamic responses. ⁴⁰ siRNAs and many ASOs reduce synthesis of disease‐associated proteins through degradation or modulation of target mRNA rather than directly eliminating existing protein pools. Consequently, the onset and duration of response are frequently governed not only by intracellular oligonucleotide persistence, but also by intrinsic turnover kinetics of target mRNA and protein. These dynamics are commonly described using indirect‐response or turnover‐based PK/PD models in which oligonucleotides inhibit protein synthesis (input) rather than enhance elimination (output). ⁴¹ , ⁴² , ⁴³ For targets with slow biological turnover rates, pharmacodynamic effects may therefore be substantially delayed relative even to tissue drug exposure.

These features challenge conventional clinical pharmacology paradigms in which plasma PK metrics or profiles serve as the principal correlate for efficacy or toxicity. ⁴⁴ This also represents a key translational challenge, as plasma PK alone provides limited insight into clinically relevant pharmacology. Traditional noncompartmental analyses and empirical (plasma) exposure–response approaches are often insufficient to mechanistically characterize oligonucleotide activity or predict clinical response. Instead, quantitative frameworks capable of integrating plasma kinetics, tissue disposition, intracellular pharmacology, target engagement, and downstream biomarker dynamics are required to support translational interpretation of PK/PD relationships. As summarized in Figure 2, a broad range of MIDD methodologies, including mechanistic PK/PD, population PK/PD, PBPK, QSP, (tissue) exposure–response modeling, clinical trial simulation, and emerging artificial intelligence/machine learning (AI/ML)‐enabled approaches, can be applied to bridge the disconnect between systemic exposure and intracellular pharmacology, thereby supporting translational dose selection, efficacy prediction, and optimization of oligonucleotide clinical development strategies.

Delivery Biology and Routes, Tissue Distribution, and Therapeutic Index

Delivery is both the central enabler and a principal constraint of oligonucleotide therapeutics. ⁴⁵ The clinical success of GalNAc‐conjugated siRNA and ASO platforms reflects the ability to exploit highly efficient hepatocyte uptake through ASGPR‐mediated endocytosis, producing a relatively predictable liver‐directed disposition profile characterized by rapid plasma clearance, substantial hepatic uptake, prolonged tissue retention, nuclease‐mediated endolysosomal metabolism, and more limited renal and hepatobiliary elimination. ²⁸ , ³⁹ , ⁴⁶ , ⁴⁷ For example, nonclinical studies with inclisiran demonstrated highest tissue distribution in liver and kidney, with liver tissue half‐lives ranging from ∼280 to 2000 h in rats and nonhuman primates, respectively despite rapid plasma clearance. ⁴⁸ These findings further illustrate why tissue PK and PD response, rather than systemic exposure alone, are central to dose selection and safety assessment for liver‐targeted oligonucleotides.

Among currently approved oligonucleotides, GalNAc conjugation has emerged as the dominant platform for hepatocyte‐directed delivery following SC administration. GalNAc‐conjugated oligonucleotides leverage high‐affinity binding to ASGPR, which is abundantly and selectively expressed on hepatocytes (∼10⁶ receptors/cell), enabling efficient receptor‐mediated uptake and intracellular accumulation. ²⁴ , ²⁷ , ³⁰ , ³¹ Rapid ASGPR internalization and recycling further support sustained hepatic uptake even at relatively low systemic concentrations, enabling highly potent liver‐directed therapies with infrequent dosing schedules and favorable therapeutic indices. The resulting disposition profile is characterized by preferential liver uptake, appreciable kidney exposure, and limited distribution within all other tissues. ⁴⁷

Mechanistic modeling and preclinical studies have additionally demonstrated important route‐ and dose‐dependent nonlinearities in GalNAc‐siRNA disposition. ³¹ , ⁴⁴ Subcutaneous administration was shown to achieve more efficient and sustained liver targeting than intravenous (IV) dosing due to absorption‐rate limited kinetics that minimize transient ASGPR saturation, whereas high IV doses reduced hepatic uptake efficiency and increased redistribution toward kidney. ³¹ , ⁴⁹ These observations indicate that conventional development paradigms favoring high or frequent IV loading doses to rapidly achieve systemic exposure may be counterproductive for GalNAc‐conjugated oligonucleotides. ⁵⁰ Importantly, such nonlinear uptake processes also complicate estimation of absolute SC bioavailability using conventional plasma AUC‐based noncompartmental analyses, since transient receptor saturation following IV administration (at doses equivalent to pharmacologically relevant SC doses) can artificially inflate plasma exposure (AUC) through reduced hepatic uptake. Given the strong PK‐ and patient convenience‐driven rationale supporting SC administration, coupled with the burden of executing dedicated human bioavailability PK studies involving IV microdosing, the practical importance of precisely quantifying absolute bioavailability may therefore be reduced for many GalNAc‐siRNA programs, particularly when SC dosing is intended throughout clinical development. Collectively, such quantitative insights highlight how mechanistic translational frameworks can reveal clinically relevant disposition principles not readily apparent from conventional plasma PK analyses alone ³¹ and directly inform actionable clinical pharmacology strategies.

An alternative and historically transformative delivery strategy for siRNA therapeutics is LNP‐mediated delivery. ⁵¹ Patisiran, the first approved siRNA therapeutic, utilizes an intravenously administered LNP formulation composed of ionizable lipids, helper phospholipids, cholesterol, and polyethylene glycol (PEG) lipids to encapsulate and systemically deliver siRNA to hepatocytes. ⁵² Unlike GalNAc conjugates, which rely on direct receptor‐mediated uptake of chemically conjugated oligonucleotide, LNPs function as multicomponent carriers that protect siRNA from nuclease degradation, prolong systemic stability, and facilitate intracellular delivery through endosomal uptake and ionizable lipid‐mediated endosomal escape. ⁵³ Following IV administration, patisiran LNPs adsorb apolipoprotein E (ApoE), enabling indirect hepatocyte targeting through low‐density lipoprotein (LDL) receptor‐mediated uptake. ⁵⁴ Importantly, LNP systems introduce additional pharmacologic complexity because both nanoparticle disposition and intracellular siRNA release kinetics contribute to efficacy and safety. ⁵⁵ , ⁵⁶ , ⁵⁷ Consequently, MIDD approaches for LNP‐delivered oligonucleotides may require simultaneous characterization of carrier biodistribution, intracellular release, active siRNA formation, and downstream pharmacology.

Despite the clinical success of liver‐targeted delivery, selective tissue distribution also represents a major limitation of current oligonucleotide therapeutics. GalNAc‐conjugated oligonucleotides generally demonstrate predominant liver uptake with markedly lower exposure in most other tissues. ²¹ In preclinical studies with the GalNAc‐siRNA SLN360, all evaluated organs exhibited <1% of peak liver concentrations following SC administration. ⁵⁸ While this profile provides a favorable therapeutic index for hepatocyte‐derived targets, it limits broader application to diseases requiring efficient delivery to skeletal muscle, lung, central nervous system (CNS), heart, or solid tumors. Achieving efficient oligonucleotide delivery into extrahepatic target tissues or cell types therefore represents another major hurdle when pursuing drug targets outside the liver.

Tissue‐selective distribution also strongly influences oligonucleotide safety profiles. Across GalNAc‐siRNA programs, repeat‐dose toxicology studies have generally demonstrated recognizable platform‐related microscopic findings involving liver and kidney at supratherapeutic doses or exposures, including hepatocellular and renal tubular vacuolation, mild liver enzyme elevations, and intracellular basophilic granules that are often reversible and considered nonadverse. ⁵⁸ These findings highlight that toxicologic interpretation for oligonucleotides requires integration of tissue accumulation, organ‐ or even cell‐level exposure margins, pharmacology, reversibility, and microscopic adaptation rather than reliance solely on plasma toxicokinetics. Therefore, reliance on plasma‐based exposure (AUC or C_max) metrics for comparison of NOAEL (no‐observed‐adverse‐effect level) between preclinical species for tissue‐targeted oligonucleotides may offer limited predictive value for desired and untoward effects. Toxicity may also arise from sequence‐dependent hybridization‐mediated off‐target effects rather than tissue burden alone. Janas et al demonstrated that hepatotoxicity associated with certain GalNAc‐siRNA constructs was largely driven by seed‐region off‐target silencing and could be mitigated through rational chemical modification strategies. ⁵⁹ Together, these observations reinforce that oligonucleotide safety is governed by both exposure‐driven tissue distribution and sequence‐dependent pharmacology, necessitating integrated quantitative assessment of tissue PK, intracellular active species, and transcriptomic off‐target liability.

Delivery beyond the liver remains an active and rapidly evolving frontier. Intrathecal ASO administration has already demonstrated clinical feasibility for CNS disorders, while newer conjugation strategies including C16‐siRNAs and antibody‐oligonucleotide conjugates (AOCs) are expanding the potential for broader tissue distribution and prolonged extrahepatic gene silencing in preclinical models. ⁶⁰ , ⁶¹ , ⁶² Pulmonary delivery represents another promising strategy, particularly for local respiratory diseases. ⁶³ , ⁶⁴ However, unlike hepatocyte‐directed delivery, efficient pulmonary oligonucleotide delivery must overcome additional barriers including mucus entrapment, mucociliary clearance, alveolar macrophage uptake, and surfactant interactions. Recent inhaled lipid‐, polymer‐, and conjugate‐based siRNA systems have demonstrated improved pulmonary uptake and local gene silencing in preclinical models. ⁶¹ , ⁶⁵

Additional extrahepatic targeting strategies are also being explored, including transferrin receptor 1 (TfR1)‐targeted oligonucleotides for skeletal and cardiac muscle, ⁶⁶ GLP‐1 receptor‐directed conjugates for pancreatic delivery, ⁶⁷ integrin‐ and folate‐receptor‐targeted systems for tumors, ⁶⁸ and antibody‐ or peptide‐mediated ⁶⁹ approaches designed to enhance tissue‐selective uptake and intracellular delivery. Such emerging extrahepatic delivery approaches have been reviewed extensively by Seth et al and others. ²⁰ , ⁷⁰

Recent advances show promise in expanding the application of siRNA therapeutics in oncology toward historically difficult‐to‐drug oncogenic drivers such as KRAS. A series of recent preclinical studies demonstrated how integrated innovations in targeted delivery and RNAi engineering can overcome longstanding barriers to KRAS‐directed therapy. ⁷¹ , ⁷² , ⁷³ An EGFR‐directed, tumor‐targeted RNAi platform (EFTX‐G12V) enabled selective delivery of allele‐specific siRNA to EGFR‐expressing tumors. ⁷² Building upon this strategy, inverted chimeric RNAi molecules simultaneously targeting KRAS and MYC achieved synergistic antitumor activity through dual oncogene suppression. ⁷¹ These studies represent an innovative RNAi‐based precision oncology strategy for historically difficult‐to‐drug oncogenic drivers and illustrate how advances in targeted delivery and RNAi engineering are extending siRNA therapeutics beyond traditional liver‐directed applications toward previously intractable oncology targets. ⁷³ As these emerging platforms progress toward clinical evaluation, quantitative characterization of their disposition, tumor biodistribution, intracellular pharmacology, and safety profile will be essential to support rational clinical development and future MIDD applications.

In summary, delivery should be viewed not merely as a formulation challenge, but as a fundamentally quantitative clinical pharmacology problem. Therapeutic efficacy and safety are jointly governed by the fraction of administered dose reaching target tissues, cellular uptake efficiency, intracellular release kinetics, active species formation, tissue residence time, and off‐target distribution. Consequently, MIDD approaches, including mechanistic PK/PD, PBPK, and QSP frameworks, are critical for linking delivery platform properties with tissue exposure, intracellular pharmacology, toxicology, and ultimately clinical dose selection and translational optimization of next‐generation oligonucleotide therapeutics.

Bioanalytical and Translational Hurdles

A major challenge in oligonucleotide therapeutics development is the ability to accurately quantify, interpret, and translate pharmacologically relevant exposure across plasma, tissues, intracellular compartments, and downstream biomarkers. Clinically relevant pharmacology is ultimately driven by intracellularly active species that are difficult, or often impossible, to directly measure in humans. Consequently, substantial uncertainty may arise when attempting to connect systemic exposure, tissue pharmacology, and clinical response using conventional plasma‐based bioanalytical and translational approaches.

Several intrinsic physicochemical properties of oligonucleotides contribute to these challenges. Oligonucleotides are relatively large, highly charged molecules with extensive protein and tissue binding, heterogeneous intracellular trafficking, and complex metabolism resulting from progressive exonuclease‐ and endonuclease‐mediated chain shortening. ¹⁰ , ³⁷ In addition, oligonucleotides frequently incorporate diverse chemical modifications‐including phosphorothioate backbones, 2′‐O‐methyl substitutions, locked nucleic acids, and hydrophobic conjugates—that influence stability, tissue distribution, protein binding, assay recovery, and metabolite formation. ¹⁹ , ²¹ , ⁷⁴ , ⁷⁵ Consequently, multiple analyte species may coexist simultaneously in vivo, including full‐length parent oligonucleotide, truncated metabolites, conjugated and unconjugated forms, intracellular active species, and oligonucleotides bound to delivery carriers or intracellular protein complexes. ⁷⁶ Determining which molecular species are pharmacologically relevant therefore becomes a critical and nontrivial challenge for quantitative pharmacology.

Current oligonucleotide bioanalysis relies primarily on three major analytical platforms: liquid chromatography‐mass spectrometry (LC‐MS), ligand‐binding assays (LBAs), and polymerase chain reaction (PCR)‐based approaches. ³⁷ LC‐MS approaches are widely utilized because they provide improved structural specificity and enable characterization of parent oligonucleotides and metabolites. ⁷⁷ , ⁷⁸ , ⁷⁹ Depending on platform configuration and matrix complexity, lower limits of quantification (LLOQ) for LC‐MS assays commonly range from low picogram‐per‐mL to low nanogram‐per‐mL concentrations in plasma, although sensitivity may become substantially reduced in tissues or intracellular compartments due to matrix effects and extraction recovery limitations. LBAs provide high analytical sensitivity, often achieving sub‐ng/mL or low pg/mL detection limits, while also enabling relatively high throughput. ⁸⁰ , ⁸¹ However, LBAs frequently detect multiple related oligonucleotide species simultaneously, including chain‐shortened metabolites, potentially confounding interpretation of exposure–response relationships. Furthermore, LBAs may inadequately distinguish conjugated from unconjugated oligonucleotide species or differentiate free versus protein‐bound analyte.

PCR‐based methods have become particularly important for siRNA because they enable quantification of intracellular active species and downstream RNA knockdown. Reverse transcription quantitative PCR (RT‐qPCR), particularly stem‐loop RT‐qPCR approaches, are utilized to quantify intracellular and RISC‐loaded antisense siRNA strands that more directly reflect pharmacologically active exposure. ⁸² Recent work by Chen et al developed a highly sensitive stem‐loop RT‐qPCR platform capable of quantifying both cytoplasmic and RISC‐loaded siRNA with quantification ranges spanning 0.0002–20 femtomole (fmol) for unmodified siRNA and 0.02–20 fmol for chemically modified siRNA, while also enabling discrimination of full‐length antisense strands from 3′ metabolites through specialized overlapping primer design. ⁸³ At any time following dosing, RISC‐loaded siRNA often represents only ∼1%–2% of total liver siRNA concentrations despite serving as the principal driver of pharmacologic activity in vivo. ²⁴ , ⁸⁴ , ⁸⁵ Specialized stem‐loop RT‐qPCR methods, immunoprecipitation‐based RISC isolation techniques, and hybridization‐based assays have therefore become important for mechanistic characterization of siRNA PK/PD in the laboratory. ⁸⁶ , ⁸⁷ Nevertheless, direct clinical measurement of intracellular active species remains technically challenging and generally infeasible in humans due to limited accessibility of repeated biopsy‐based tissue sampling. This general inability to directly measure intracellular active species represents one of the most important translational gaps in clinical development programs.

Consequently, many key determinants governing oligonucleotide PK/PD relationships remain incompletely measurable or entirely inaccessible in clinical settings. Plasma concentrations may decline below quantifiable limits within hours despite prolonged tissue retention and sustained pharmacodynamic effects. At the same time, clinically relevant intracellular processes including endosomal escape, cytosolic release, RISC loading, RNase H engagement, and intracellular trafficking—are extremely difficult to quantify directly in vivo. Liver biopsy sampling, while theoretically informative, is invasive and rarely feasible in routine clinical development. ³¹ , ⁸⁸ Furthermore, many oligonucleotide targets are intracellular, membrane‐bound, or non‐secreted proteins, limiting availability of circulating biomarkers that directly reflect tissue target engagement. ³⁸ , ⁸⁹ Consequently, many mechanistic drivers of efficacy and toxicity cannot be empirically observed in humans and instead must be inferred indirectly through translational modeling approaches.

These measurement limitations also create challenges for preclinical‐to‐clinical translation. Relationships among plasma PK, tissue concentrations, intracellular active species, target knockdown, and downstream biomarkers may differ substantially across species due to differences in receptor expression, intracellular trafficking efficiency, target turnover, tissue accessibility, and delivery platform behavior. Conventional empirical exposure–response analyses may therefore prove insufficient to mechanistically interpret oligonucleotide pharmacology or reliably predict human efficacy and safety.

MIDD approaches are particularly valuable in this setting because they provide quantitative frameworks capable of integrating sparse and heterogeneous data across biological scales. ⁹⁰ Mechanistic PK/PD and PBPK models can infer otherwise unobservable intracellular processes by jointly integrating plasma PK, tissue disposition, biomarker kinetics, and downstream pharmacodynamic responses. These approaches further enable characterization of nonlinear uptake mechanisms, tissue‐selective distribution, and translational prediction of bioactive exposures across species and disease states. As oligonucleotide therapeutics continue evolving toward complex delivery systems and extrahepatic targets, overcoming these bioanalytical and translational limitations will remain central to enabling efficient and mechanistically informed clinical development.

Interspecies Translation, NAMs, and Emerging Opportunities

Application of PK/PD principles to dose selection and preclinical‐to‐clinical translation for oligonucleotides presents unique challenges compared with small molecules and protein therapeutics. First, pharmacologically relevant siRNA concentrations at tissue sites of action often differ substantially from transiently measurable concentrations in accessible biofluids such as plasma. ³³ , ⁹⁰ Second, receptor‐mediated uptake and intracellular disposition processes can produce species‐, route‐, and dose‐dependent nonlinearities that are difficult to reliably translate using traditional body weight‐based allometric scaling alone. ⁹¹ Third, the durability of gene silencing is governed predominantly by intracellular rate‐limiting processes including endosomal retention, cytosolic release, RISC loading, and target turnover kinetics. ³³

Accordingly, development of predictive translational PK/PD frameworks for oligonucleotide therapeutics has historically been challenging. Conventional “bench‐to‐bedside” paradigms are further complicated by the limited clinical relevance of plasma PK, restricted feasibility of repeated tissue biopsies in humans, and growing practical and ethical constraints surrounding extensive animal testing. Despite these challenges, recent mechanistic and systems‐based modeling approaches have progressively demonstrated the feasibility of quantitatively linking preclinical and clinical oligonucleotide pharmacology. ³¹ , ⁸⁴

Emerging evidence suggests that tissue‐level and intracellular pharmacology for GalNAc‐siRNA therapeutics may exhibit encouraging translational scalability across species when evaluated using mechanistically relevant exposure metrics rather than plasma alone. ⁴⁷ , ⁸⁵ , ⁹² Mechanistic translational modeling has successfully predicted liver exposure, RISC dynamics, and target gene silencing from rodents and nonhuman primates to humans using combined allometric and physiologically based scaling approaches. ⁸⁴ Notably, after accounting for interspecies differences in liver PK, estimates of “true” in vivo potency, defined using RISC‐bound siRNA exposure metrics, remained within approximately two‐fold across species. ⁸⁴ These findings support the growing feasibility of predictive and mechanistically informed translational frameworks for oligonucleotide therapeutics.

Recent studies have also highlighted opportunities to incorporate human‐relevant in vitro systems into translational workflows. Basiri et al demonstrated that hepatocyte‐based stability systems for GalNAc‐siRNA therapeutics could reasonably predict in vivo pharmacodynamic durability, supporting the potential utility of in vitro–in vivo correlation (IVIVC) approaches for oligonucleotide development. ⁹³ More broadly, advances in hepatocyte cultures, organoids, microphysiological systems, and organ‐on‐chip technologies have generated increasing interest in NAMs capable of complementing or partially replacing traditional animal studies. ⁹⁴ , ⁹⁵ , ⁹⁶ , ⁹⁷

These scientific advances coincide with broader regulatory and ethical efforts to reduce reliance on animal testing. The principles of reduction, refinement, and replacement (“3Rs”) have become influential across pharmaceutical development, particularly for advanced therapeutic modalities. ⁹⁸ In 2022, the FDA Modernization Act 2.0 clarified that nonanimal approaches may support drug development programs in place of certain animal studies. ⁹⁹ , ¹⁰⁰ Subsequent regulatory initiatives and activity on NAMs have further encouraged integration of computational modeling, human‐relevant in vitro systems, and mechanistic translational approaches into development workflows. ⁹⁴

These developments are particularly relevant for oligonucleotide therapeutics, where tissue‐driven pharmacology and intracellular mechanisms often make mechanistic translational modeling more informative than empirical plasma PK comparisons alone. However, despite growing enthusiasm surrounding NAM‐enabled paradigms, widespread replacement of animal studies for oligonucleotide therapeutics will likely remain gradual. Tissue‐selective distribution, intracellular trafficking, immune activation, and whole‐body toxicology remain central determinants of efficacy and safety, while robust IVIVC frameworks linking emerging in vitro systems to human tissue pharmacology are currently lacking (or unvalidated in a quantitative sense). Species differences in receptor expression, nuclease activity, intracellular processing, sequence homology, and immune activation pathways may also substantially influence both pharmacology and safety interpretation across species. ³³ , ⁸⁴ , ⁸⁹ , ¹⁰¹ , ¹⁰²

These advances reinforce the growing importance of MIDD approaches, including mechanistic PK/PD, PBPK, and QSP frameworks, for translational prediction, dose selection, and development of more efficient and human‐relevant paradigms for oligonucleotides. As regulatory frameworks continue evolving toward greater incorporation of NAMs and reduced reliance on animal testing, oligonucleotide therapeutics may represent a particularly impactful area in which integrated experimental and computational approaches can simultaneously improve mechanistic understanding, translational efficiency, and clinical decision‐making. ³⁴ , ⁹⁹

MIDD Strategies for Oligonucleotide Therapeutics

The issues presented in the preceding section highlight several fundamental translational challenges for siRNA therapeutics. Notably, the pharmacologically relevant analyte evolves throughout intracellular disposition, from circulating siRNA to total tissue and ultimately RISC‐loaded siRNA, such that plasma PK alone provides an incomplete correlate of intracellular gene knockdown. Because these active intracellular species are highly challenging to measure clinically, computational model‐based approaches are required to infer tissue pharmacology and bridge species differences in delivery, intracellular processing, and target biology.

As shown previously in Figure 2, a broad spectrum of MIDD approaches have emerged to address the unique translational and quantitative pharmacology challenges associated with oligonucleotide therapeutics. These approaches span empirical dose–response time (DRT) or kinetic‐pharmacodynamic (K‐PD) models, population and exposure–response analyses, mechanistic PK/PD frameworks, PBPK and minimal PBPK (mPBPK) models, and QSP approaches. Representative structures for the quantitative modeling frameworks discussed throughout this section—from empirical K‐PD models to mechanism‐based PK/PD and PBPK/PD frameworks—are presented in Figure 4. Selection of an appropriate modeling framework depends on the specific development question, available data, stage of development, and desired level of mechanistic inference.

Figure 4.

Three major modeling paradigms applied to oligonucleotide therapeutics. (a) Kinetic‐pharmacodynamic (K‐PD) model in which an inferred biophase drives target protein suppression without explicitly modeling drug pharmacokinetics. (b) Mechanistic PK/PD model describing plasma disposition, receptor‐mediated uptake, intracellular trafficking, RNA‐induced silencing complex (RISC) loading, target mRNA degradation, and downstream protein suppression. (c) Whole‐body or minimal physiologically based pharmacokinetic‐pharmacodynamic (PBPK‐PD) model integrating organ distribution, ASGPR‐mediated hepatic uptake and receptor dynamics, intracellular pharmacology, and target engagement to support mechanistic interspecies translation and prediction of human PK/PD. Representative model structures adapted and redrawn from Jansson‐Löfmark and Gennemark (a), ¹⁰⁵ Ayyar and Song (b), ³¹ and Ayyar et al (c). ⁸⁴ Readers are referred to the original publications for complete model descriptions and mathematical formulations. ASGPR, asialoglycoprotein receptor; PBPK, physiologically based pharmacokinetics; PD, pharmacodynamics; PK, pharmacokinetics; RISC, RNA‐induced silencing complex.

Importantly, these approaches should be viewed as complementary tools existing along a continuum of increasing biological and physiological complexity. Empirical frameworks may be particularly useful in sparse‐data settings or for interpolation within studied dose ranges, whereas mechanistic and physiologically based models support characterization of tissue disposition, intracellular pharmacology, nonlinear receptor‐mediated uptake, and translational prediction across species and patient populations. Accordingly, the field has progressively evolved beyond descriptive plasma PK analyses toward integrated multiscale frameworks linking systemic exposure, tissue pharmacology, target engagement, biomarker dynamics, and clinical response.

The following sections highlight representative applications of MIDD in oligonucleotide therapeutics, emphasizing mechanistic PK/PD characterization, dosing route considerations, regimen optimization, special population assessment, and mechanistically informed clinical development strategies.

Empirical and Semi‐Mechanistic Characterization of Oligonucleotide PK/PD

Conventional plasma PK‐driven PK/PD approaches are often insufficient for oligonucleotide therapeutics, particularly for GalNAc‐conjugated siRNA platforms where rapid receptor‐mediated hepatic uptake and prolonged intracellular retention create marked temporal discordance between plasma PK and tissue PD. Accordingly, empirical approaches capable of inferring local tissue kinetics from pharmacodynamic response profiles have emerged as useful tools for characterizing oligonucleotide PK/PD relationships in sparse‐data settings.

Among the most widely explored empirical or semi‐mechanistic approaches are DRT or K‐PD models (Figure 4a). ¹⁰³ , ¹⁰⁴ These frameworks are based on the premise that longitudinal pharmacodynamic data inherently contain information regarding drug kinetics at the “effect site” or biophase, even in the absence of plasma kinetics. Applications of DRT modeling to siRNA and ASO therapeutics were demonstrated by Jansson‐Löfmark and Gennemark, who developed empirical models linking dose administration to downstream target suppression through an inferred tissue compartment driving pharmacodynamic response. ¹⁰⁵ Across multiple clinical siRNA and ASO datasets, estimated biophase half‐lives for siRNA therapeutics substantially exceeded both plasma PK and target protein turnover half‐lives, supporting the concept that intracellular persistence at the tissue site of action predominantly governs pharmacodynamic durability. For example, patisiran exhibited a plasma terminal half‐life of approximately 1.6–2.5 days, whereas the estimated biophase half‐life was approximately 11 days. In contrast, estimated ASO biophase half‐lives were often more similar to terminal plasma half‐lives, consistent with modality‐specific differences in tissue disposition and intracellular pharmacology. ¹⁰⁶

However, these models remain largely empirical and retrospective in nature. Their utility is strongest for descriptive characterization and interpolation within studied dose ranges or regimens, particularly in late‐stage or sparse‐data clinical settings where tissue PK measurements are unavailable. Extrapolation across dose levels, routes of administration, or novel regimens may be limited because these frameworks generally do not explicitly account for nonlinear receptor‐mediated uptake, intracellular trafficking, tissue heterogeneity, or metabolite kinetics. Consequently, mechanistic translational models may be preferable when prospective prediction or extrapolation is required.

It must be noted that considerations may differ across oligonucleotide classes and delivery platforms. For certain unconjugated phosphorothioate ASOs, the terminal plasma disposition phase may approach equilibrium with tissue concentrations and therefore remain mechanistically informative of tissue pharmacology and pharmacodynamic durability. ¹⁰⁷ , ¹⁰⁸ In such cases, conventional plasma PK‐driven population PK/PD approaches may retain greater translational relevance than for GalNAc‐conjugated siRNAs or ASOs. ¹⁰⁹ , ¹¹⁰

Empirical (pharmacostatistical) approaches have been applied to relate target knockdown with clinical endpoint responses. For example, during development of the siRNA therapeutic fitusiran, an empirical mathematical relationship was constructed linking antithrombin (AT) lowering with thrombin generation (TG), where TG = 16.1 + 0.0065 × AT lowering² adequately characterized the nonlinear increase in thrombin generation observed with increasing antithrombin suppression. ¹¹¹ Although descriptive in nature, such analyses represent early quantitative efforts to bridge oligonucleotide‐mediated target modulation with downstream clinical response.

Mechanistic Translational (PB)PK/PD Modeling for Human Dose Selection and Early Development

Mechanistic models provide quantitative insight into factors controlling the nonlinear disposition and prolonged pharmacodynamic durability characteristics of oligonucleotide therapeutics and offer rational translation from preclinical species to humans. Unlike empirical PK/PD approaches, these frameworks explicitly integrate biologically relevant processes governing oligonucleotide disposition and activity, including receptor‐mediated uptake, intracellular trafficking, active species formation, target turnover, and downstream biomarker response, as illustrated by the representative mechanism‐based and PBPK/PD model structures shown in Figures 4b,c. Consequently, they are particularly well suited to characterize how siRNA properties, target biology, and physiologic systems collectively influence the magnitude and duration of gene silencing. ³³ , ¹¹² , ¹¹³

A mechanistic mPBPK/PD framework developed by Ayyar et al (Figure 4c) represented one of the first integrated translational models to quantitatively characterize the multiscale disposition and pharmacodynamics of GalNAc‐siRNA therapeutics across species. ¹⁰ , ⁸⁴ The model integrated plasma PK, liver biodistribution, ASGPR‐mediated uptake and receptor dynamics, intracellular siRNA trafficking, RISC loading, and downstream target mRNA and protein turnover using hybrid mechanistic and allometric scaling principles across rodents, nonhuman primates, and humans for fitusiran. Importantly, the framework demonstrated encouraging cross‐species scalability of liver PK and reasonable consistency in estimated in vivo potency parameters across nonclinical species and humans, supporting the concept that tissue‐level and intracellular pharmacology of GalNAc‐siRNA therapeutics may be quantitatively translatable across species when evaluated using mechanistically relevant RISC‐bound siRNA exposure metrics rather than plasma PK alone.

More broadly, this work established a foundational mechanistic and translational PK/PD framework for GalNAc‐siRNA therapeutics and helped catalyze subsequent model development efforts across the field. ³³ , ¹¹⁴ Subsequent studies have expanded or adapted the original framework to evaluate nonlinear receptor‐mediated disposition and metabolite kinetics, ³¹ whole‐body distribution, ¹¹⁵ , ¹¹⁶ intracellular pharmacology, ¹¹⁷ organ impairment effects, ¹¹⁸ and translational PK/PD for emerging siRNA therapeutics. ¹¹⁹ , ¹²⁰ , ¹²¹ For example, Tian et al adapted and successfully applied the foundational model to quantitatively evaluate interspecies translation of a new apolipoprotein C‐III (APOC3) silencing GalNAc‐siRNA, RBD5044. ¹²⁰ More recently, Fan et al developed a computational translational platform for GalNAc‐siRNA therapeutics that explicitly incorporated and expanded upon the mechanistic structure originally described by Ayyar et al. ¹²¹ The framework integrated preclinical and clinical datasets across multiple approved GalNAc‐siRNAs and was subsequently applied prospectively to SAL0132, a GalNAc‐siRNA targeting lipoprotein(a), to support translational PK/PD prediction and early clinical development decisions. Collectively, these studies illustrate how mechanistic frameworks are evolving from compound‐specific models toward more reusable and platform‐level quantitative development tools.

Building on these concepts, mechanistic translational PK/PD models are being operationalized prospectively to support FIH dose projection and early clinical development decisions. Figure 5 illustrates a translational modeling workflow in which preclinical measurements (plasma PK, tissue biodistribution, and pharmacodynamics of target mRNA and/or protein knockdown) and mechanistic model‐based scaling principles are iteratively integrated to project pharmacologically active human dose ranges and guide early clinical study design. Importantly, this paradigm shifts translational decision‐making away from reliance on plasma exposure alone toward mechanistically informed prediction of biologically active intracellular exposure and tissue pharmacology.

Figure 5.

Mechanistic translational PK/PD modeling workflow to inform first‐in‐human dose projection and Phase 1 study design for GalNAc‐siRNA therapeutics. Experimental data are indicated by white boxes, and model‐based translation or prediction steps are indicated by grey shaded boxes. ASGPR, asialoglycoprotein receptor; EC₅₀, half‐maximal effective concentration; E_max, maximum effect; FIH, first‐in‐human; GalNAc, N‐acetylgalactosamine; IV, intravenous; mRNA, messenger ribonucleic acid; NHP, nonhuman primate; PBPK‐PD, physiologically based pharmacokinetic‐pharmacodynamic; PD, pharmacodynamics; PK, pharmacokinetics; RISC, RNA‐induced silencing complex; SC, subcutaneous; siRNA, small interfering RNA.

As an example, a translational mPBPK/PD framework developed for the PNPLA3‐targeted GalNAc‐siRNA JNJ‐75220795 successfully characterized systemic, kidney, and liver disposition together with durable PNPLA3 mRNA silencing across rodents and nonhuman primates. ¹²² Translational simulations informed a projected pharmacologically active human dose range of 75–400 mg SC, which was subsequently explored prospectively in Phase 1 clinical studies in subjects with metabolic dysfunction‐associated fatty liver disease (MAFLD). ¹¹⁴ , ¹²³ Time‐course and magnitude of observed clinical pharmacodynamic responses (hepatic fat fraction via MRI‐PDFF) demonstrated encouraging alignment with model‐informed projections of target (mutant I148M PNPLA3) protein modulation, supporting feasibility of PBPK/PD model frameworks for prospective dose selection and early clinical decision‐making in oligonucleotide therapeutics. ¹²² , ¹²³

These studies highlight the continuing evolution of MIDD for novel modalities such as oligonucleotides from retrospective PK/PD characterization toward predictive, prospectively applied translational modeling frameworks. More broadly, these approaches enable quantitative integration of tissue pharmacology, intracellular disposition, and interspecies scaling principles to support rational clinical dose selection and development of next‐generation RNA‐based therapeutics.

Mechanistic Modeling Beyond Dose Selection

Following establishment of mechanistic translational PK/PD frameworks for FIH dose selection, modeling efforts for oligonucleotide therapeutics have progressively expanded toward more complex disposition, clinical pharmacology, and biomarker‐informed questions. These include characterization of nonlinear receptor‐mediated uptake, route‐dependent delivery efficiency, intracellular active metabolite disposition, special populations, and a priori prediction in scenarios where empirical observations remain limited or experimentally inaccessible.

A major recent focus has been understanding nonlinear ASGPR‐mediated hepatic uptake and its implications for GalNAc‐siRNA disposition and delivery efficiency. Ayyar and Song developed a mechanistic model, illustrated in simplified form in Figure 4b, incorporating liver and kidney biodistribution, active metabolite formation, and competitive receptor‐mediated hepatic uptake to characterize disposition of GalNAc‐siRNA therapeutics across species using rich nonclinical and clinical givosiran datasets. ³¹ , ⁷⁶ , ¹²⁴ The analyses enabled quantitative estimation of in vivo ASGPR binding and internalization kinetics alongside resolution of body weight‐based cross‐species scaling exponents for SC absorption (−0.24) and intracellular disposition (−0.27). Mechanistically, parameter estimates suggested that the durability of siRNA pharmacology emerged not from plasma persistence (t_1/2 ∼2.5 h), but from slow intracellular sequestration and dissociation kinetics within acidic endolysosomal compartments (t_1/2 ∼200 h), substantially exceeding RISC turnover (∼17 h) and target mRNA turnover (∼2.3 h) timescales. The model additionally enabled successful a priori prediction of human plasma PK for both givosiran and its major active metabolite, AS(N‐1)3′ givosiran.

Relatedly, Sandra et al adopted simplified mechanistic concepts within a population PK framework to characterize plasma and liver PK of JNJ‐73763989, a GalNAc‐conjugated siRNA combination product, in recombinant adeno‐associated virus hepatitis B virus‐infected mice. ⁴⁹ The model provided quantitative understanding of hepatic uptake efficiency as a function of dose and route of administration and was subsequently leveraged translationally to estimate human liver exposure associated with clinically observed antiviral responses of the GalNAc‐conjugated siRNA triggers JNJ‐73763976 and JNJ‐73763924 targeting HBV transcripts. ¹²⁵ These analyses further illustrated the growing utility of mechanistically informed liver PK frameworks for optimizing late‐stage clinical development strategies in oligonucleotide therapeutics.

Mechanistic systems models have also been adapted to evaluate and anticipate oligonucleotide pharmacology in special patient populations where dedicated clinical data may remain sparse. Using Ayyar et al's established mPBPK‐PD framework, ⁸⁴ Lumen et al recalibrated and deployed the model to assess the impact of renal and hepatic impairment on plasma PK and downstream pharmacodynamic biomarkers for inclisiran and vutrisiran. ¹¹⁸ The analyses suggested that disease‐associated physiologic changes may alter systemic exposure while exerting comparatively smaller effects on pharmacodynamic response due to preserved hepatic target engagement, particularly in patients with mild‐to‐moderate hepatic impairment. Such applications highlight the growing utility of mechanistic models for quantitatively informing clinical pharmacology questions in scenarios where direct empirical evaluation may be difficult, resource intensive, or clinically impractical.

Perhaps one of the most exciting emerging roles of mechanistic modeling is its transition from describing existing oligonucleotide molecules to prospectively informing the discovery and design of improved next‐generation candidates. Once sufficiently calibrated and qualified, mechanistic frameworks enable quantitative testing of scientific hypotheses before molecules are synthesized or advanced into development, complementing empirical experimentation and emerging human‐relevant NAMs. Rather than relying solely on iterative screening and expensive in vivo animal testing, these models can evaluate target suitability and interrogate integrated early‐stage questions such as: Which biologic targets are most likely to yield durable therapeutic responses? Which processes among target protein turnover, cell lifespan, receptor abundance and tissue uptake, intracellular trafficking, or endosomal escape are major rate‐limiting determinants of pharmacologic activity? How do specific chemical modifications simultaneously influence nuclease resistance, tissue and intracellular disposition, and productive engagement with RISC or RNase H1? Would optimization of receptor affinity, conjugate chemistry, or tissue‐selective delivery meaningfully improve productive intracellular exposure of the oligonucleotide? By quantitatively integrating target biology with tissue and intracellular drug disposition and pharmacology, these frameworks provide a rational basis for resolving such multidimensional design trade‐offs, optimizing delivery strategies, and refining candidate selection prior to clinical translation. Emerging computational tools and insights from whole‐body PBPK ¹¹⁶ and intracellular PK/PD modeling, ¹¹⁷ discussed in the following subsection, increasingly illustrate this paradigm.

These studies and examples illustrate the evolving role of mechanistic modeling in oligonucleotide development beyond retrospective PK/PD characterization and FIH dose projection. Increasingly, these frameworks are serving as quantitative decision‐making tools across the development continuum, informing target and delivery platform optimization, nonlinear in vivo disposition, tissue‐ and even cell‐selective delivery, intracellular pharmacology, special populations, and predictive human PK/PD.

Toward Systems Pharmacology and Integrated Multiscale Modeling

As oligonucleotide therapeutics expand beyond liver‐directed applications and into more complex disease settings, MIDD frameworks are correspondingly evolving from relatively minimal translational PK/PD models toward integrated multiscale and systems pharmacology approaches. These emerging frameworks seek to mechanistically connect whole‐body biodistribution, intracellular disposition, pathway biology, and downstream therapeutic outcomes across multiple biologic scales.

Although mechanistic PBPK frameworks were initially established for GalNAc‐siRNA therapeutics, the underlying modeling paradigm is increasingly being extended across diverse oligonucleotide modalities. Recent applications include LNP‐based mRNA therapeutics ¹²⁶ and CNS‐targeted ASOs ¹²⁷ following intrathecal administration, while emerging modalities such as AOCs present exciting opportunities to integrate established systems PK and PBPK principles from both oligonucleotide therapeutics ⁸⁴ and antibody‐drug conjugates (ADCs) ¹²⁸ , ¹²⁹ into unified translational modeling frameworks. While each modality requires representation of distinct biologic processes governing tissue targeting, intracellular trafficking, and bioactive drug release, the underlying mechanistic architecture often remains fundamentally similar, integrating tissue biodistribution, cellular uptake, intracellular processing, and target biology to predict pharmacologically active exposure and response.

Whole‐Body PBPK Modeling

Recent efforts have progressively extended liver‐centric translational models toward whole‐body physiologically based frameworks capable of quantitatively describing oligonucleotide biodistribution across tissues, species, and therapeutic modalities. For example, Derbalah et al developed a cross‐species mechanistic whole‐body PBPK model integrating plasma, liver, kidney, spleen, muscle, adipose, and other tissue compartments to characterize pharmacokinetics of both siRNA and ASO across conjugated and unconjugated modalities. ¹¹⁶ The analyses identified tissue endocytosis kinetics and fraction unbound as important determinants of oligonucleotide disposition and demonstrated encouraging prediction of observed clinical plasma PK profiles within approximately two‐fold across compounds and species. Importantly, such frameworks move beyond liver‐focused characterization toward a more generalized systems understanding of oligonucleotide biodistribution and may become even more important as the field advances toward extrahepatic delivery platforms, novel conjugates, and more diverse chemistries where quantitative understanding of tissue‐selective exposure, intracellular delivery, and off‐target distribution will likely become central determinants of efficacy and safety.

Cellular‐Level PK/PD Modeling

In parallel, more sophisticated intracellular PK/PD models have emerged to mechanistically interrogate subcellular determinants of oligonucleotide pharmacology. Chen et al developed a mechanistic intracellular model integrating siRNA uptake, intracellular trafficking, Ago2‐mediated RISC loading, mRNA degradation, protein turnover, and cell proliferation dynamics to evaluate determinants of pharmacodynamic durability and target silencing efficiency. ¹¹⁷ The analyses suggested that cellular system‐specific properties, including target turnover rates and cell division kinetics, may substantially influence achievable duration and magnitude of target suppression independent of siRNA exposures alone. The model further demonstrated that prolonged silencing may emerge from sustained intracellular persistence of active RISC complexes despite declining total intracellular siRNA concentrations, reinforcing the importance of intracellular pharmacology as a major determinant of response. ¹¹⁷

QSP Modeling

QSP approaches are increasingly being applied to connect oligonucleotide‐mediated target knockdown with downstream pathway biology, disease mechanisms, and clinically meaningful therapeutic outcomes. ⁹⁰ Unlike traditional PK/PD or PBPK/PD frameworks that primarily characterize exposure and target engagement, QSP models mechanistically integrate proximal target modulation with systems‐level biologic response. Because much of oligonucleotide pharmacology occurs within intracellular compartments that are difficult or impossible to measure clinically, QSP, particularly when integrated with PBPK, provides a suitable framework for quantitatively linking intracellular disposition and pharmacology with downstream pathway biology and therapeutic response. Importantly, these models can help distinguish whether heterogeneous treatment responses arise from differences in intracellular delivery and target engagement or from downstream pathway regulation and disease biology, thereby informing optimization of delivery technologies, target selection, and rational combination strategies. For example, a comparative QSP framework for anti‐PCSK9 modalities, including siRNA‐mediated PCSK9 suppression by inclisiran, boinell incorporated PCSK9 regulation, LDL‐C turnover, and statin pharmacology to quantitatively predict LDL‐C lowering following siRNA‐mediated PCSK9 suppression while enabling comparative evaluation against monoclonal antibody‐based PCSK9 inhibition strategies. ¹³⁰ Such frameworks also support generation of biologically informed virtual patient populations for evaluating interindividual variability, biomarker strategies, combination therapies, and clinical trial design, while facilitating benchmarking across therapeutic classes and quantitative integration of biologic and clinical knowledge during oligonucleotide development. ¹³¹

Altogether, these emerging multiscale frameworks highlight a broader evolution of oligonucleotide MIDD from characterization of exposure–response relationships alone toward integrated systems‐level understanding of tissue disposition, intracellular pharmacology, pathway biology, and therapeutic response. As oligonucleotide therapeutics continue expanding into more heterogeneous diseases, tissues, and patient populations, successful development will likely depend on quantitative frameworks capable of integrating mechanistic information across biologic scales and systems rather than isolated PK or biomarker observations alone.

Future Opportunities and Remaining Challenges for MIDD of Oligonucleotide Therapeutics

Steady advances in oligonucleotide chemistry, delivery technologies, and mechanistic understanding of tissue and intracellular pharmacology have increasingly expanded MIDD for oligonucleotide therapeutics from predominantly empirical approaches toward more predictive and mechanistically integrated translational frameworks. As conceptually illustrated in Figure 6, different quantitative modeling approaches may provide unique value across distinct stages of development, spanning preclinical translation, FIH dose selection, clinical dose optimization, special population analyses, lifecycle management, and indication expansion.

Figure 6.

Applications of MIDD frameworks across the oligonucleotide development lifecycle, illustrating the use of mechanistic modeling, population modeling, MBMA, and AI/ML approaches to inform dose selection, characterize efficacy, optimize clinical study design, and support decision‐making from preclinical development through registration and lifecycle expansion. AI/ML, artificial intelligence/machine learning; E–R, exposure–response; FIH, first‐in‐human; K‐PD, kinetic‐pharmacodynamic; MBMA, model‐based meta‐analysis; MIDD, model‐informed drug development; PBPK‐PD, physiologically based pharmacokinetic‐pharmacodynamic; PD, pharmacodynamics; PK, pharmacokinetics; PK/PD, pharmacokinetic/pharmacodynamic; PK‐TE, pharmacokinetic‐target engagement; QSP, quantitative systems pharmacology.

The workflow depicted in Figure 6 emphasizes that oligonucleotide MIDD is not a single‐model exercise, but rather an iterative and integrated knowledge‐building paradigm. Early in development, mechanistic PBPK/PD and QSP frameworks can integrate preclinical plasma PK, tissue biodistribution, intracellular pharmacology, and downstream biologic response to support translational PK/PD characterization, projection of pharmacologically active human dose ranges, and rational Phase 1/2 study design. As clinical datasets mature, more empirically based population PK/PD, exposure–response, and model‐based meta‐analysis (MBMA) approaches may complement these mechanistic frameworks by characterizing variability, efficacy, and safety relationships across broader patient populations. Importantly, the framework also highlights how quantitative models may continue informing decisions well beyond initial dose selection and regimen optimization, including combination strategies, special population analyses, competitor benchmarking, indication expansion, and development of next‐generation molecules (reverse translation). Collectively, these approaches increasingly support a continuous “learn‐and‐confirm” paradigm in which mechanistic and data‐driven frameworks iteratively refine development decisions across the therapeutic lifecycle.

A major future opportunity and challenge for oligonucleotide therapeutics involves successful clinical expansion beyond liver‐directed applications toward extrahepatic tissues such as the lung, kidney, muscle, adipose, and immune cells. While GalNAc‐mediated ASGPR targeting has enabled highly efficient hepatocyte delivery, many extrahepatic tissues present substantially greater barriers related to vascular permeability, tissue penetration, cellular heterogeneity, intracellular trafficking, and endosomal escape. Consequently, it is still unclear to what extent translational assumptions established for liver‐targeted siRNA and ASOs would generalize directly to emerging delivery platforms. Mechanistic understanding of tissue‐specific uptake pathways, intracellular sequestration, stability, and release kinetics, and cell‐specific lifespans will therefore become important determinants of successful translation and clinical response.

In parallel, human‐relevant experimental systems are being incorporated into translational workflows. Advanced in vitro platforms utilizing primary human cells, organoids, and organ‐on‐chip technologies are being developed to characterize oligonucleotide uptake, intracellular trafficking, and tissue‐specific pharmacology under more physiologically relevant conditions. ⁹⁵ , ⁹⁶ , ⁹⁷ Although these systems hold substantial promise, their successful incorporation into predictive MIDD frameworks will require rigorous qualification against in vivo and clinical outcomes, as robust IVIVC relationships remain incompletely established for many intracellular and tissue‐level processes. Accordingly, future translational paradigms will likely depend on integrated multiscale frameworks capable of quantitatively bridging cellular systems, tissue pharmacology, and whole‐body responses.

Increasing convergence between mechanistic PBPK/PD and QSP frameworks with AI, ML, and NAMs may substantially transform future oligonucleotide development paradigms. ⁹⁴ Automated high‐throughput screening platforms already enable simultaneous evaluation of hundreds of oligonucleotide sequences and chemical modification strategies for stability, potency, and intracellular activity, thereby narrowing candidate selection prior to animal testing. ¹³² Likewise, NAM‐derived systems may provide human‐relevant datasets to inform mechanistic translational frameworks. Integration of these data streams into adaptive multiscale modeling architectures may ultimately support more efficient hypothesis generation, translational prediction, and iterative optimization across development stages. AI/ML‐enabled approaches may further facilitate virtual population generation, adaptive clinical trial simulation, and continuous integration of multimodal datasets across compounds and therapeutic platforms.

Nevertheless, despite growing enthusiasm surrounding AI and NAM‐enabled paradigms, their near‐term role will likely remain complementary rather than replacement‐oriented. Isolated in vitro systems remain limited in their ability to fully recapitulate tissue‐selective biodistribution, intracellular trafficking, and whole‐body physiologic responses. Similarly, purely data‐driven approaches may struggle to extrapolate reliably outside observed experimental domains without incorporation of mechanistic biologic constraints. Consequently, the most impactful future frameworks will likely be hybrid systems integrating mechanistic modeling, NAM‐derived experimental data, clinical pharmacology observations, and explainable AI‐enabled analytics within quantitatively interpretable translational paradigms.

Another important emerging challenge involves improving mechanistic understanding of interindividual variability in oligonucleotide PK, PD, and clinical response. ³³ , ³⁴ Unlike many traditional therapeutics, for which differences in plasma exposure may explain a substantial portion of interindividual variability, key determinants of oligonucleotide response often reside at tissue and intracellular levels. Variability in receptor expression, endocytic activity, intracellular trafficking efficiency, fibrosis, inflammation, target turnover, and disease‐associated tissue remodeling may all contribute to differences in therapeutic response despite similar plasma PK profiles. ³³ , ³⁷ Mechanistically informed virtual patient populations offer a powerful means of quantitatively dissecting these otherwise difficult‐to‐measure sources of clinical variability, enabling evaluation of responder subpopulations and biomarker strategies prior to clinical investigation. As oligonucleotide therapeutics increasingly expand into heterogeneous diseases and precision medicine applications, future MIDD frameworks will likely require deeper integration of tissue biology, mechanistic biomarkers, systems pharmacology, and patient‐specific disease characteristics.

More broadly, oligonucleotide therapeutics remain a rapidly evolving translational science in which many foundational biologic and quantitative principles are still emerging. The coming years will likely yield major advances in tissue‐selective delivery, intracellular pharmacology and IVIVC, understanding variability in therapeutic response, and translation of novel delivery technologies into clinical use. Importantly, the future of oligonucleotide MIDD will likely depend not on any single modeling paradigm, but rather on integrated PBPK/PD, QSP, NAM‐derived experimental systems, AI/ML‐enabled analytics, and iterative clinical learning frameworks. Together, these advances position MIDD to play a much more central role in enabling rational and precision‐driven development of next‐generation oligonucleotide therapeutics.

Conclusions

Therapeutic oligonucleotides have evolved into a transformative therapeutic modality capable of selectively modulating previously difficult‐to‐drug biological targets. Yet their continued successful development depends on quantitative frameworks that move beyond conventional plasma PK‐centric paradigms toward mechanistic understanding of tissue disposition, intracellular pharmacology, and systems‐level biologic response. As highlighted throughout this review, MIDD approaches are helping bridge these complexities through integration of mechanistic PK/PD modeling, PBPK, QSP, and emerging multiscale translational frameworks across preclinical and clinical development. The field now appears positioned at a broader inflection point in which mechanistic modeling, human‐relevant experimental systems, and data‐driven methodologies may converge into adaptive, continually evolving computational ecosystems for translation. As oligonucleotide therapeutics continue expanding across diseases, tissues, and technologies, such integrated approaches will likely play a pivotal role in enabling more predictive, efficient, and patient‐focused development of next‐generation oligonucleotides as precision medicines.

Author Contributions

Paridhi Gupta and Vivaswath S. Ayyar contributed equally to this work. All authors read and approved the final manuscript for publication.

Conflicts of Interest

Paridhi Gupta is currently employed by Bristol Myers Squibb. Vivaswath S. Ayyar is a former employee of Johnson & Johnson. Vivaswath S. Ayyar and Mindy Magee are employees of GSK and may hold stock or stock options in the company. The authors declare no other conflicts of interest.

Funding

No funding was received to assist with the preparation of this manuscript.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.