What's different today

Drug discovery has always moved in waves, but currently several previously distinct scientific revolutions are converging simultaneously. The promise is a discovery environment in which biology can be interrogated at greater depth, targets can be modulated through more diverse mechanisms, and candidate molecules can be designed with increasingly intentional control over function and developability.

Historically, pharmaceutical R&D was constrained by a narrow set of druggable target classes and limited capacity to connect disease heterogeneity to mechanism-based intervention. The newest discovery platforms matter because they directly reduce those constraints. Degraders convert occupancy pharmacology into event-driven pharmacology. Conjugates turn selective binding agents into targeted delivery vehicles. Gene editors permanently rewrite sequence. And AI-based design, where proprietary experimental data and platform chemistry are strong, is being applied to shorten the design cycle in hit generation, optimization, and biomolecular engineering.

The strategic environment in which these platforms have to develop has also changed. Biotech went through a prolonged funding contraction from 2022 through 2024; rising interest rates, a largely closed IPO window, and compressed public-company multiples made long-duration R&D assets harder to finance. Recovery through 2025 and into 2026 has been real but uneven: strategic capital and Big Pharma BD remain active, while generalist venture money has been selective. The bar for getting a deal done has gone up. Pharma BD teams are funding fewer platforms, more deliberately, and late-stage failures and clinical holds now draw sharper scrutiny than they would have in 2020 or 2021. Regulatory conditions have shifted too. Accelerated-approval standards, the rigor expected of confirmatory trials, and post-marketing enforcement have all tightened. Well-positioned modalities still get through; marginal ones face a tougher road.

For large pharmaceutical companies, this changes the question from whether to engage these fields to how to stage investment across them. The usual mistake is to treat every exciting modality the same way; better decisions depend on distinguishing what is validated, what is emerging, and what is enabling. The groupings that follow reflect not just whether a field has approved medicines, but how much commercial traction is behind them, how robust the supporting infrastructure is, and whether many large pharma organizations now treat the field as a capability they need to be in rather than one they can sit out.

ADCs have moved from a specialist oncology format into a major strategic class, and the reason is straightforward: improvements in linker chemistry, payload diversity, site-specific conjugation, and patient-selection strategies have improved the therapeutic index. ADC performance is determined not simply by antigen binding but by an integrated system involving target biology, internalization kinetics, linker stability, payload properties, biodistribution, and bystander effects. For every ADC that has worked, many others have failed in the clinic or reached approval without differentiating commercially; the format works, but not any program built in the format.

What is increasingly appreciated is that ADC innovation is no longer confined to first-generation cytotoxic payload delivery. The field is diversifying into novel payloads, innovative linker chemistry, immune-modulating designs, and non-oncology applications. The Enhertu trajectory is the canonical case in point. AstraZeneca and Daiichi Sankyo's 2019 collaboration, worth up to $6.9 billion, built on an ADC whose topoisomerase-I payload, cleavable tetrapeptide linker, and high drug-to-antibody ratio together redefined what "HER2-positive" meant clinically.¹ The DESTINY-Breast04 readout in 2022 extended the drug into HER2-low disease, creating an entirely new patient segment that did not exist as a commercial category before the molecule. That is the kind of outcome that justifies treating ADCs as a platform rather than a format. The competitive moat in this field is no longer antibody engineering alone. Daiichi's proprietary DXd payload-linker platform now spans HER2, TROP2, HER3, and B7-H3; the defensible asset is the integrated chemistry-biology system, not any single candidate, as reflected in Merck's $22 billion potential total commitment to three Daiichi DXd candidates in 2023.²

Incretin therapeutics have become a systems pharmacology platform rather than a metabolic success story, and that distinction matters a great deal for how one approaches partnering or in-licensing decisions in this space. The peer-reviewed evidence base supports clinically meaningful benefits across obesity, cardiovascular disease, kidney disease, and cardiometabolic risk reduction, while combination agonism is extending the frontier of mechanism-based metabolic control.

From a discovery strategy standpoint, the interesting opportunity is not another undifferentiated GLP-1 agonist. It is the broader biology that the field has opened up: polyagonist design, tissue-selective signaling, oral peptide engineering, lean-mass preservation strategies, metabolic inflammation modulation, and expansion into indications where weight loss is only one component of therapeutic effect.

The practical question for late entrants is which differentiation actually matters. Programs that offer structural rather than incremental advantages (triple agonism, oral small-molecule formats, alternative combination mechanisms) are where credibility is being established; assets built on modest tolerability or dosing-convenience advantages over tirzepatide or semaglutide have struggled to attract Big Pharma partnering interest at economics their investors needed. Novo Nordisk and Eli Lilly have also built peptide manufacturing capacity, supply chains, and formulation IP that competitors cannot easily replicate. The $16.5 billion Novo Holdings acquisition of Catalent in 2024, with onward transfer of three fill-finish sites to Novo Nordisk,³ and Eli Lilly's increase of its Lebanon, Indiana API site investment to $9 billion in the same year,⁴ illustrate the scale of that structural capacity build.

Genuine next-generation differentiation, not incremental improvement, is what opens the door to serious partnering. How far the category expands from here also depends on what payers will cover and on what kind of long-term outcomes evidence they end up requiring, both of which are still moving.

Radiopharmaceuticals have re-emerged as a high-conviction area because targeted radionuclide delivery combines biomarker-driven precision with a mechanism largely orthogonal to conventional systemic therapy. Radiopharmaceuticals can exploit targets that are already validated for imaging or binding, including the repertoire of mAb targets and many already validated extracellular target molecules.

The field's expanding relevance across solid tumors depends critically on isotope selection, chelation chemistry, dosimetry, and companion diagnostics. The strategic value depends not only on candidate molecules but on whether the organization can build or access the required operational ecosystem. Novartis's position in this field is the clearest illustration. The $3.9 billion acquisition of Advanced Accelerator Applications announced in October 2017⁵ and the $2.1 billion acquisition of Endocyte in October 2018⁶ built an integrated isotope-supply, manufacturing, and clinical capability that competitors have struggled to replicate. Pluvicto's subsequent FDA approval in March 2022⁷ and commercial trajectory, alongside the persistent supply constraints that shaped its early commercial profile, make the point: having the drug is not the same as being able to deliver it at scale. Isotope supply, reactor access, and cold-chain logistics act as independent constraints on how fast and how widely any radiopharmaceutical can reach patients, which is why the manufacturing and logistics capability matters as much to the investment case as the clinical data. Pluvicto's commercial trajectory, the multiple Phase 3 programs beyond PSMA, and sustained partnering activity around isotope-based therapeutics are why we place this field in Wave 1 despite its earlier inflection point. The field is not fully mature, but many large pharma organizations now treat it as a capability they need to be in rather than one they can sit out.

Targeted protein degradation shifts the underlying pharmacology from sustained occupancy to induced elimination of the protein target, a genuinely new paradigm. PROTACs, molecular glues, and broader induced-proximity architectures enable pharmacologic access to targets that have historically resisted classical inhibition, including scaffolding proteins and non-enzymatic disease drivers.

For pharma R&D organizations, the significance of degradation extends well beyond the PROTAC modality itself. It marks the emergence of induced proximity as a general design principle. Once a portfolio adopts event-driven pharmacology, future extensions naturally include stabilization, delocalization, deubiquitination, and context-dependent control of protein fate. That makes targeted protein degradation a route into a broader redefinition of how intracellular targets are manipulated.

The technology is not as evenly validated as some of its proponents claim, and significant work remains before the modality is mature. The Arvinas and Pfizer collaboration on vepdegestrant (ARV-471) is the most visible test case. Pfizer's $650 million upfront commitment in 2021 signaled that degraders had crossed into serious Big Pharma portfolio territory,⁸ but the Phase 3 VERITAC-2 readout in 2025, which met its primary PFS endpoint in the ESR1-mutant subpopulation with a hazard ratio of 0.57 but did not show a statistically significant benefit across the full intent-to-treat population, is a useful reminder that even well-engineered degraders are not a free pass past the usual selectivity, PK, and patient-segmentation challenges.⁹ The degrader thesis survives this readout intact, but as a modality requiring the same translational discipline as any small molecule, not as a shortcut around it. Across the field, outcomes are driven more by the specifics of each program — the choice of E3 ligase, where that ligase is actually expressed in tissue, and how the molecule behaves pharmacokinetically — than by any single architectural principle of the platform. The ecosystem extends well beyond Arvinas. Gilead's 2025 option-and-license agreement with Kymera for oral CDK2 molecular glue degraders (up to $750 million total)¹⁰ and Roche's 2023 strategic collaboration with Monte Rosa (up to approximately $2 billion total)¹¹ show that partnering interest in induced-proximity pharmacology remains broad even where individual readouts disappoint.

Gene editing is crossing a critical threshold from theoretical platform promise to clinical reality. Prime editing, base editing, and CRISPR-derived systems are increasingly discussed not simply as enabling technologies but as therapeutic classes in their own right, with editing windows, repair profiles, and delivery strategies that are becoming tractable enough for serious clinical development planning. In the longer term, the future may be written rather than edited: gene writing platforms aim to insert, replace, or rewrite larger DNA segments (up to whole genes plus regulatory elements) in a programmable way, often without classical double-strand breaks, using systems such as Cas-transposase fusions, retroelement-based "writers," or related synthetic biology constructs.

The promise lies in a changing definition of durable intervention. Even where editing will not dominate near-term standard of care, it can reshape expectations around durability, patient segmentation, and platform valuation. The hope-versus-hype gap in this field is still wide. The industry has seen earlier one-and-done therapies struggle commercially after approval: Glybera's European marketing authorization was not renewed in 2017 after minimal commercial use,¹² and Zynteglo was withdrawn from the European market in 2021 following pricing and reimbursement disputes.¹³ The field has its landmark approval in Casgevy (exagamglogene autotemcel), the Vertex and CRISPR Therapeutics sickle cell and beta-thalassemia therapy cleared by the FDA in December 2023, the first approved CRISPR-based medicine.¹⁴ But the cautionary counterpoint arrived quickly: Verve Therapeutics halted enrollment in the Heart-1 Phase 1b study of its PCSK9 base-editing candidate VERVE-101 in April 2024 after a patient in the 0.45 mg/kg cohort experienced Grade 3 transaminase elevation and Grade 3 thrombocytopenia; the program pivoted to a reformulated lipid nanoparticle in VERVE-102.¹⁵

That sequence, a transformative approval alongside a clinical hold in an adjacent program within the same eighteen months, is why we remain cautious about extrapolating from the hemoglobinopathy success into broader cardiometabolic applications where the risk-benefit math is very different. Delivery is the main thing holding the field back. Editing works well enough ex vivo or when aimed at the liver; getting edited cargo safely into other tissues is the problem that will decide which diseases gene editing can actually address.

The success of mRNA vaccines accelerated interest in RNA as a programmable therapeutic modality, but the long-term strategic value lies in improved delivery and expression control. The key determinant of future expansion is more the chemistry, formulation, biodistribution, and tolerability of the delivery system rather than the coding payload alone.

Oligonucleotide therapeutics and mRNA therapeutics share information-based design principles but face different delivery, stability, and manufacturing realities. Novartis's Leqvio (inclisiran), FDA-approved in December 2021, established siRNA as a viable chronic-dosing modality with payer-acceptable economics.¹⁶ Moderna and Merck's personalized cancer vaccine V940 (mRNA-4157) is now in Phase 3 in adjuvant melanoma (INTerpath-001) and non-small cell lung cancer (INTerpath-002), testing whether the mRNA platform extends meaningfully beyond prophylaxis.¹⁷ Delivery remains the principal shared ceiling across these modalities. The strongest opportunities are likely to emerge where transient or sustained expression is clinically advantageous and where organ-targeting can be made sufficiently precise. RNA therapeutics should be viewed as an adaptable information platform whose ceiling is set by delivery specificity and capability.

Bispecific and trispecific antibodies are transitioning to an engineering discipline in which geometry, valency, immune synapse formation, cytokine tuning, and half-life control are used deliberately to shape biological outcome. Their importance in oncology is already established, but the deeper significance is that multispecificity allows biologics to encode logic rather than merely affinity.

Many disease networks are not well-modulated through a single node alone, and that single-node constraint has long limited what drug discovery could address. Multispecific designs offer a route to conditional activation, dual-pathway suppression, cell redirection, and more selective immune engagement. The flip side of engaging the immune system more precisely is having to manage what happens when it responds; cytokine release and on-target off-tumor toxicity shape which formats and combinations can realistically progress through development.

Amgen's Imdelltra (tarlatamab), a DLL3-targeted T-cell engager, received accelerated FDA approval in May 2024 for small cell lung cancer and full approval in November 2025 on the basis of the Phase 3 DeLLphi-304 trial, which demonstrated survival benefit versus standard-of-care chemotherapy.^18,19 Johnson & Johnson's Tecvayli (teclistamab) and Talvey (talquetamab) have established multiple myeloma as a serious indication for the T-cell engager format. And Jazz Pharmaceuticals' $1.76 billion license of zanidatamab from Zymeworks in 2022 signaled that the asset class had moved firmly into partnering territory for Big Pharma BD teams.²⁰ Multispecifics competition is not confined to Western developers. Akeso and Summit's ivonescimab, a PD-1/VEGF bispecific, showed a statistically significant improvement in progression-free survival over pembrolizumab monotherapy in the Phase 3 HARMONi-2 trial in first-line PD-L1-positive NSCLC, on data presented at WCLC and reported by Akeso and Summit in September 2024,²¹ and Summit's $500 million upfront license from Akeso puts Chinese-originated assets squarely into Western BD strategy.²²

A hallmark of AI in drug discovery is that hype often outpaces the operational reality, but the genuine utility is material and growing. The fields where AI has become operationally useful, including de novo molecular design, protein engineering, property optimization, structure-guided design, and closed-loop design-make-test-analyze workflows, are expanding rapidly. The recognition of protein structure prediction by the 2024 Nobel Prize in Chemistry underscores the scientific significance of AI-enabled structural biology.

The partnering pattern is informative. Isomorphic Labs' simultaneous 2024 collaborations with Eli Lilly²³ and Novartis²⁴ brought in roughly $82 million in combined upfront payments against up to $2.9 billion in contingent milestones, and Novartis expanded the arrangement in February 2025. Structurally, these look like discovery partnerships with milestone-weighted economics rather than platform acquisitions, which tells you how most established pharma companies are pricing AI-discovery risk. The broader pattern is that pharma is signing multi-target deals with AI-first companies rather than attempting to rebuild those capabilities internally, while AI-first companies themselves are concentrating rather than broadening their portfolios. The deal structure says something about how pharma is thinking: AI as a way to make existing drug discovery faster and better, not as a completely new place drugs come from. So far there are few examples of AI-derived programs showing clinical differentiation that is clearly attributable to the AI approach itself, and that is the bar the field still has to clear.

The key is not to ask whether AI will replace medicinal chemistry or biologics engineering. The more relevant question is how AI changes cycle time, hypothesis quality, and portfolio breadth. AI is best understood not as a standalone therapeutic area but as an amplifier across every other area in this paper. Its value compounds where proprietary biological data, translational assays, and platform chemistry are already strong, and that is where the capability-building investment should be focused.

Cellular senescence has evolved from a basic aging concept into a credible therapeutic framework for chronic disease, tissue dysfunction, fibrosis, and age-associated pathology. Senolytic and senomorphic strategies are advancing.

Aging will not become a registrational indication per se. Rather, senescence biology increasingly intersects with established disease areas including pulmonary disease, metabolic dysfunction, oncology, and fibrosis. The framing that has intellectual rigor and rational investability is "diseases of the elderly" rather than "longevity."

UNITY Biotechnology's Phase 2 clinical setback in osteoarthritis, reported in August 2020, illustrated how difficult it is to translate senolytic concepts into registration-quality efficacy without careful mechanism-matching and indication selection; UBX0101 failed to separate from placebo at the 12-week primary endpoint, and UNITY discontinued the program and shifted focus to ophthalmologic and neurologic applications.²⁵ Funding around Life Biosciences, Altos Labs, and others continues, though capital availability in this space is bifurcated: strategic and foundation-backed money is present, while traditional venture underwriting of aging-as-indication theses is not. Clinical validation across the category is still thin, which is why we keep it in Wave 3 despite the scale of attention and capital it attracts. Near-term commercial weight will most likely come from disease-specific applications rather than aging-as-indication programs.

Spatial transcriptomics and related single-cell methods may fundamentally improve target identification and mechanism resolution. Spatially resolved molecular profiling allows discovery teams to identify cell states, microenvironment interactions, and treatment-response signatures that are difficult to resolve with bulk analysis.

Target validation remains one of the scarcest commodities in R&D. Better resolution of where a target is expressed, how diseased tissues are organized, and which cell-cell interactions drive pathology can improve both candidate selection and biomarker design.

Spatial biology's commercial traction sits largely on the tools side rather than in therapeutic deals. Its strategic value is felt upstream, in the quality of target selection and mechanism interpretation feeding every other modality. It belongs less in a list of "next therapeutic classes" and more in the set of capabilities whose adoption materially improves the output of everything else.

Neuroinflammation has moved from a secondary effect to a central disease axis across neurodegenerative, neuroimmunological, and neurovascular conditions. Convergent mechanisms involving microglia, astrocytes, inflammasome signaling, peripheral immune crosstalk, and blood-brain barrier dysfunction now provide a mechanistic framework that cuts across multiple CNS diseases.

CNS portfolios have historically suffered from weak target validation and poor translational fidelity. Neuroinflammation-anchored approaches offer a more mechanistically grounded route into disease modification, but only if programs are built around precise cellular hypotheses and biomarker-defined patient subsets rather than broad anti-inflammatory assumptions.

The 2025 evidence base is mixed in exactly the way the framing predicts. Sanofi's brain-penetrant BTK inhibitor tolebrutinib delivered a positive Phase 3 result in non-relapsing secondary progressive MS (HERCULES, HR 0.69; 95% CI 0.55-0.88; p=0.003), published in the New England Journal of Medicine in April 2025,²⁶ but subsequently missed in primary progressive MS (PERSEUS, December 2025),²⁷ and the FDA issued a Complete Response Letter declining to approve the secondary progressive MS indication in late December 2025.²⁸ The CRL arguably carries more implications for the broader BTK-in-neuroinflammation thesis than either trial result alone. TREM2 programs have absorbed real failures in parallel. Yet Sanofi's $470 million acquisition of Vigil Neuroscience in 2025 for the oral small-molecule TREM2 agonist VG-3927 indicates that the mechanism still commands real Big Pharma conviction despite the setbacks.²⁹ The mixed readouts across both targets are a reminder that a mechanism making biological sense is not the same as the drug working in patients.

We find this field compelling not because it guarantees success, but because it may finally offer a tractable framework for rational CNS intervention. The 2024-2025 pattern of sharper targets alongside continued late-stage failures is why careful program design, not mechanism conviction alone, tends to separate winners from expensive misses.

Implications

The core discipline is to avoid confusing scientific novelty with strategic fit. The most valuable opportunities usually sit at the intersection of three conditions: strong disease biology rationale, a modality with genuine mechanistic advantage, and an executable development path covering delivery, biomarkers, and manufacturing. Modalities that meet only one or two of these conditions can be worth watching, but they deserve smaller bets and longer timelines than modalities that meet all three. And pharma's behavior is a signal, not proof. Deals get done on science that later fails, and programs get walked away from on science that later works. Across all three waves, what tends to kill programs is rarely whether the underlying biology is real; it is whether the biology can be turned into a drug with workable delivery, safety, manufacturability, and commercial positioning.

What tends to kill programs is rarely whether the underlying biology is real.

The staging matters as much as the criteria. For Wave 1 platforms (ADCs, GLP-1, radiopharmaceuticals), the game is access and differentiation. The categories are validated; the moats sit in payload-linker chemistry, manufacturing capacity, and integrated supply ecosystems more than in target novelty. Late entrants with incremental improvements face a harder path to Big Pharma attention; genuine next-generation advantages are usually what earn it.

For Wave 2 platforms (targeted protein degradation, gene editing, RNA therapeutics, multispecifics, AI-enabled design), the opportunity is positioning ahead of inflection. The companies most likely to win are those that pick modalities where the translational evidence is real and the supporting infrastructure actually exists, not those chasing whatever platform has the most momentum that year. The VERITAC-2 and Verve VERVE-101 readouts are reminders that premature commitment to any single example within these platforms can be costly, even where the broader thesis holds.

For Wave 3 frontiers (senescence, spatial biology, neuroinflammation in neurodegeneration), the cleanest way in is disease-specific rather than platform-wide. Senescence as a biology of chronic disease has more capital access than aging as an indication. Spatial biology is most valuable as an enabling layer across other modalities, not as a therapeutic class in its own right. Neuroinflammation rewards precise mechanistic hypotheses tied to biomarker-defined subsets, not broad anti-inflammatory theses.

The companies worth funding are the ones that know where their modality actually sits on this curve and build accordingly. Early-stage science with unproven clinical translation ("Wave 3") sold at the valuation of a commercially validated platform ("Wave 1") will disappoint; a mature, deal-ready category pitched as if it were a visionary frontier will too. The framework above is meant to help evaluate and value emerging science so that effective investment and partnering decisions can be made.