90% of Molecules Fail Before They Reach Their Target

The Delivery Gap: Why Great Molecules Fail

90% of drugs that enter clinical trials fail. Lack of efficacy and safety toxicity account for most of these losses, but 10-15% trace directly to pharmacokinetic and delivery failures -- molecules with strong in vitro activity that cannot reach therapeutic concentrations at their intended site of action in vivo.

This delivery gap is expensive. Billions spent on drug discovery produce candidates that never reach patients because they cannot cross the biological barriers between administration and action. The molecule may be flawless; the delivery system fails it. In practice, a potent drug with poor delivery is indistinguishable from a failed drug.

The advanced drug delivery market tells the story clearly. Valued at $299.6 billion in 2025, it is projected to reach $487.4 billion by 2030 -- a compound annual growth rate of 10.2%. Pipeline complexity keeps rising, and delivery innovation is increasingly the rate-limiting step between discovery and therapy.

The First-Pass Metabolism Challenge

For oral drugs, the GI tract and liver are unforgiving gatekeepers. First-pass metabolism -- hepatic degradation of absorbed drug before it reaches systemic circulation -- can slash bioavailability to the point where oral dosing becomes impractical.

The numbers are stark. Propranolol, one of the most widely prescribed beta-blockers, achieves only 26% oral bioavailability due to extensive hepatic metabolism. Morphine reaches just 30% bioavailability orally, forcing dose escalation that amplifies side effects. These are not niche compounds -- they are frontline therapeutics held back by delivery constraints.

The damage goes beyond reduced efficacy. Variable first-pass metabolism drives inter-patient variability, complicates dose optimisation, and raises adverse event risk. A patient with hepatic impairment may see wildly different drug exposure than the healthy volunteers in pivotal trials, producing real-world safety gaps that controlled studies cannot predict.

The Tissue Targeting Complexity

First-pass metabolism is only part of the problem. Getting drug to the right tissue at therapeutic concentrations is its own challenge. The blood-brain barrier (BBB) is the clearest example: CNS disorders affect hundreds of millions of people worldwide, yet approximately 98% of small-molecule chemotherapeutics cannot cross the BBB.

Solid tumours pose distinct obstacles. The enhanced permeability and retention (EPR) effect -- once expected to passively funnel nanoparticles into tumours -- has been inconsistent in clinical translation. Tumour heterogeneity, irregular vasculature, and high interstitial pressure mean systemic administration frequently fails to deliver meaningful intratumoral concentrations.

The result: pipelines crowded with molecules that showed compelling activity in isolated systems but collapsed in the complex biological environment of actual patients. Tissue targeting is not a minor engineering detail. It is a fundamental bottleneck in therapeutic development.

The Patient Compliance Crisis

Even when drugs reach their targets, sustained therapeutic effect depends on patient adherence -- and the data here is blunt. Approximately 50% of patients with chronic conditions do not take their medications as prescribed. Non-adherence accelerates disease progression, drives hospitalisation and mortality, and burns healthcare resources on preventable complications.

The causes are well documented: dosing complexity, multiple daily administrations, difficult procedures, and side effects triggered by peak-trough concentration swings. A therapy dosed once monthly will, on average, outperform a daily equivalent in the real world -- not because the molecule is better, but because the delivery system accounts for human behaviour.

Delivery systems that extend dosing intervals, simplify administration, and flatten plasma concentrations attack compliance at its source. The evidence is already in: long-acting injectables and extended-release formulations consistently show better real-world effectiveness than conventional alternatives, even when their pharmacokinetic profiles are comparable.

The Delivery Technologies Landscape

Different delivery problems demand different engineering solutions. Each technology below targets a specific limitation -- first-pass avoidance, sustained release, targeted accumulation, or stimuli-responsive activation -- and each carries its own development, manufacturing, and regulatory requirements.

Spotlight: Antibody-Drug Conjugates

Antibody-drug conjugates (ADCs) merge biologics-grade targeting with small-molecule potency. Cytotoxic payloads linked to tumour-targeting antibodies achieve selective delivery that conventional chemotherapy cannot. The ADC market reached $12.26 billion in 2024 and is projected to hit $32.11 billion by 2033 at an 11.3% CAGR.

The technology has moved well past its early failures. First-generation ADCs were undermined by linker instability and payload limitations. Current platforms use site-specific conjugation, cleavable linkers tuned to tumour microenvironment conditions, and payloads with wider therapeutic windows. Enhertu (trastuzumab deruxtecan) illustrates the progress -- delivering efficacy in breast cancer patients who had exhausted other HER2-targeted options.

Emerging Frontier: Exosome and Extracellular Vesicle Delivery

Extracellular vesicles (EVs) and exosomes exploit natural cellular communication pathways. These nanoscale vesicles already shuttle proteins, nucleic acids, and lipids between cells -- making them candidates for therapeutic cargo delivery with lower immunogenicity than synthetic carriers.

EV-based delivery is still in early clinical development, but preclinical results for both small molecules and nucleic acid therapeutics are encouraging. Manufacturing scalability and characterisation standardisation remain hard problems, yet the approach continues to draw significant research funding.

AI/ML Breakthroughs in Delivery System Design

Delivery system optimisation requires balancing drug loading, release kinetics, stability, targeting efficiency, and manufacturability -- a multidimensional design space that empirical screening cannot cover efficiently. Machine learning changes this calculus by enabling predictive formulation design that cuts experimental burden by an order of magnitude.

The accuracy is already actionable. LightGBM models have achieved R-squared values exceeding 0.87 for predicting nanoparticle delivery efficiency, allowing researchers to rank formulation candidates computationally before synthesizing a single batch. The same approaches predict drug release profiles from polymer matrices, transdermal permeation rates, and microsphere degradation kinetics.

Beyond prediction, ML accelerates optimisation. Bayesian optimisation algorithms identify optimal compositions with far fewer experiments than traditional design-of-experiments methods. For ADCs, ML models predict payload-linker-antibody combinations most likely to hit target therapeutic indices -- compressing development timelines from years to months.

Patent and Regulatory Considerations

Dense patent coverage defines delivery technology development strategy. Platform technologies -- PLGA microspheres, OROS osmotic systems, transdermal matrix designs -- sit behind extensive patent portfolios. Any organisation entering this space must map freedom-to-operate constraints early, or risk late-stage derailment of otherwise viable programmes.

Regulatory frameworks are still catching up. The FDA's 505(b)(2) pathway allows reformulated drugs to reference prior approval of the active ingredient, reducing some development burden. But demonstrating bioequivalence for complex delivery systems often demands clinical studies that approach the scope of traditional NDAs.

For novel delivery technologies, regulatory acceptance depends on showing that added complexity serves a clear therapeutic purpose. Agencies are open to AI-driven development approaches but expect full documentation of model development, validation, and decision-making. Early regulatory engagement has become critical for complex delivery programmes.

The DeepC Approach: Integrated Delivery Intelligence

DeepC works on the delivery problem with studies that share one conversation: one designs the formulation, one improves it, one maps the patents around it. Delivery system design can then run as a parallel workstream from day one, shaping compound selection and development strategy instead of trailing behind discovery.

The formulation study opens with a QTPP, then designs 3 to 5 candidates from distinct strategic angles, working from the API's physicochemical properties and the route. For a poorly soluble CNS-targeted compound, nanoparticle, lipid-based, and intranasal candidates can sit side by side, weighed in a multi-criteria decision analysis.

The formulation optimization returns up to three variants of a formulation you already have, balancing competing objectives such as drug loading against stability, or release rate against manufacturing robustness, and states what each variant gives up. For a campaign that learns from bench results round by round, DeepC Optimize (in early access) proposes each next experiment with constrained Bayesian optimisation.

The freedom-to-operate analysis maps the patent field before you commit: claim charts for the blocking patents, a low, moderate, or high risk rating, and design-around strategies. In delivery areas with heavy patent coverage, early design-arounds prevent costly late-stage pivots and surface openings for new IP.

Key Strategic Priorities

• Integrate delivery early: Run delivery system design as a parallel workstream to discovery, not a downstream afterthought. Early delivery assessment shapes compound selection and heads off late-stage failures.
• Build AI-driven formulation capabilities: Predictive models at R-squared > 0.87 are a step change from empirical screening. Organisations that lack them face widening competitive disadvantage.
• Audit pipeline rescue opportunities: Shelved compounds with delivery limitations are undervalued assets. Modern delivery systems can unlock therapeutic potential that was previously out of reach.
• Map the patent field proactively: Dense IP coverage in delivery technologies demands early FTO assessment and design-around planning. Discovering a blocking patent late can kill an otherwise successful programme.
• Design for compliance: The 50% non-adherence rate in chronic conditions is a massive real-world effectiveness gap. Extended-release and long-acting formulations close that gap at the product level.