What 368,000 patents reveal about where drug discovery is heading
Patents sit at a useful point in the innovation timeline: late enough that a technology is worth protecting, early enough to signal where a field is moving before that direction becomes obvious. In a recent CAS Insights webinar co-produced with ACS Webinars, three CAS scientists used that signal to map the pharmaceutical landscape, drawing on more than 368,000 pharmaceutical patent families published between 2020 and 2025 across over 100 patent offices worldwide.
- Angela Zhou introduced the methodology and the industry-level view of who is filing.
- Kavita Iyer walked through emerging therapeutic modalities and molecular targets.
- Janet Sasso traced where disease-area innovation is accelerating and where it has stalled.
The single clearest signal across every layer of the analysis is diversification: more modalities, more targets, and more disease areas drawing serious research attention.
A data foundation built for landscape analysis
Angela Zhou opened with the data itself. The analysis drew on 368,000 patent families extracted from the CAS Content Collection™, the largest human-curated repository of scientific information. Patent families are consolidated so that a single innovation filed across many offices counts once rather than many times. That consolidation matters: it measures research effort rather than legal activity. Every patent had already been human-indexed, yielding roughly 85,000 distinct scientific concepts and 7.6 million distinct chemical or biological substances across the corpus.
The method paired that data with natural language processing (NLP) and AI-assisted analysis, guided throughout by domain experts who arranged, deduplicated, and validated the extracted topics. Angela framed the combination as the core of the work: comprehensive data, advanced analytics, and scientific judgment working together, with the results visualized so readers can take in a landscape at a glance.
At the industry level, total pharmaceutical patent output has held steady across the six years studied. Angela also examined ownership. Among the top 50 pharmaceutical filers, most hold patents they filed themselves, but a few acquire a notable share from others. Novo Nordisk led on acquired patents at roughly 20%, followed by Sino Biopharmaceutical and China Resources. For a small research organization looking to advance technology toward the clinic, that pattern points to the companies most likely to license or buy in.
Therapeutic modalities: Diversification, not replacement
Next, Kavita Iyer picked up the modality thread. Small molecules still dominate patent volume, unsurprising given their century-long advantage in chemistry, scale, cost, and oral bioavailability. The more telling story is in growth rate: cell and gene therapies, RNA therapeutics, and antibodies are all outpacing small molecules, while small categories such as AI and machine learning in drug discovery and radiopharmaceuticals are accelerating fastest of all.
To read the analysis, Kavita walked through the CAS TrendScape maps that anchored the report. In each map, the size of a hexagon reflects how many patent documents sit behind a topic, and its color reflects growth rate, with warmer oranges and reds signaling faster increases. Early topics tend to be small and can grow quickly; larger, established areas grow more slowly. The most compelling opportunities usually show meaningful accumulated research and a fast growth rate, though a researcher hunting for uncrowded white space might look for faster growing topics with few patents designated by small, brightly colored hexagons instead.
Several modality signals stood out:
- Targeted protein degradation. The field reached a milestone in May, when the U.S. FDA approved vepdegestrant (Arvinas/Pfizer), the first proteolysis-targeting chimera (PROTAC) cleared for treatment in the world, in this case for breast cancer. Patent activity concentrates on the components that determine selectivity and efficacy, the E3 ligase binders and linkers. Targets for PROTACs are diversifying beyond the original hormone-receptor targets into kinases and others. Molecular glues remain a comparatively open, fast-growing space, and linkers recur as a critical shared component across modalities.
- RNA therapeutics. mRNA therapeutics continues to grow well beyond mRNA vaccines during the pandemic, and small interfering RNA (siRNA) remains a large, maturing segment, where activity centers on conjugation chemistry and delivery. Newer subtypes, including self-amplifying RNA, circular RNA, and guide RNA tied to CRISPR, are attracting rising interest. BioNTech emerged as a clear patenting leader, with activity rising since 2022.
- Antibody therapeutics. The maps here show larger hexagons in cooler colors, a sign of substantial, ongoing activity in a maturing area. Monoclonal antibody growth has plateaued, while bispecific and multispecific antibodies and antibody-drug conjugates (ADCs) are growing faster, particularly in immuno-oncology. Nanobodies and single-domain antibodies are better represented in academic settings, reflecting their earlier stages.
- Cell and gene therapies. Chimeric antigen receptor (CAR) therapy dominates, driven by CAR-T success in blood cancers, with patent activity focused on CAR structures and components. A second milestone came in June, when China's NMPA approved satricabtagene autoleucel (Satri-cell by CARsgen), the first CAR-T therapy for a solid tumor, targeting Claudin 18.2 in gastric cancer. CAR-NK and CAR-macrophage approaches are emerging as off-the-shelf alternatives with potential advantages in solid tumors and toxicity.
Across company portfolios, the diversification is structural: most leading firms now patent across four to five modalities rather than the one or two common a decade ago. Bristol Myers Squibb and AstraZeneca are extending from small molecules into antibodies and cell and gene therapies, Eli Lilly into RNA, and Regeneron, historically antibody-first, into RNA as well.
Molecular targets: Beyond the established playbook
The analysis identified more than 2,000 molecular targets, with enzymes and receptors leading the number of emerging topics. Kavita focused on three areas: receptors, epigenetic targets, and CD antigens, chosen because receptors underpin most approved drugs, epigenetic drugs represent a mechanistically distinct and recently expanding class, and CD antigens matter across CAR-T, ADCs, and related modalities.
Within receptors, attention is moving to targets distinct from the well-established ones: TRP channels in pain and inflammation, C-type lectin receptors expressed on immune cells for cancer, and ASGPR, less a conventional target than a delivery receptor enabling liver-directed siRNA. Among the major receptor families, somatostatin receptor subtype 2 (SSTR2) anchors peptide receptor radionuclide therapy, melanocortin receptors feature in metabolic and rare obesity disorders, and cannabinoid receptor work points towards a desire to create peripherally restricted agents that avoid central nervous system side effects. Angela noted that the appearance of ACE2, heavily studied during the pandemic, was a positive control confirming the method surfaces genuine signal.
For epigenetic targets patent activity now extends to SWI/SNF complex subunits, which are frequently mutated in cancer, and to protein arginine methyltransferase 5 (PRMT5), with several inhibitors in each class reaching clinical trials. Patent activity indicates considerable diversification as compared to first generation epigenetic therapeutics which were centered on HDAC and DNMT inhibitors. Even for the established HDAC targets, priorities have shifted from pan-inhibitors toward isoform-specific designs.
Among CD antigens, the well-known markers CD19, CD20, CD38, and CD47 remain central, but activity is expanding to other immune-cell targets: CD7 for T-cell malignancies (prompting CRISPR-edited, CD7-deleted CAR-T programs to avoid self-targeting), CD25 as a target that can be exploited in opposite directions in autoimmunity versus cancer, and CD133 on treatment-resistant cancer stem cells.
Disease areas: Where innovation is accelerating, and where it is not
Janet Sasso closed the presentations with the disease-area view. Cancer remains the single largest category at roughly 15% of filings, followed by infectious disease, neurological disorders, and immune-mediated disease. Across nearly every category, corporate entities account for most filings and academic organizations for the remainder, a split Janet read as a pipeline continuum rather than a measure of commercial interest. Academic groups tend to lead early discovery, while companies carry work through development, scale-up, and commercialization. A high academic share often marks an earlier-stage field.
The analysis mapped more than 1,700 growth topics across 17 disease categories, and the fastest-moving areas are not the largest markets:
- Fibrotic disease had one of the most expansive maps in the report, spanning seven organ systems. Fibrosis behaves less like a single disease than a recurring mechanism addressed with shared pathways across organs, favoring platform strategies over organ-by-organ programs.
- Genetic disease stood out for how many rare conditions now show a visible innovation footprint, made tractable by gene therapy, antisense oligonucleotides, and enzyme replacement. Huntington's disease leads (in terms of growth) with gene-silencing programs targeting the HTT messenger RNA.
- Metabolic disease is diversifying past diabetes and lipid clusters, with rare lipid disorders driving the deepest activity. Obesity and weight disorders are seeing renewed interest alongside GLP-1 success, including new delivery routes such as Eli Lilly's oral, once-daily GLP-1 pill.
- Respiratory disease showed the highest proportion of fast-growing topics of any category, concentrated to asthma, allergic airway disease, and occupational and radiation-induced fibrotic conditions.
Janet was equally direct about the quiet zones. Infectious disease sits on the low, steady side of growth, reflecting mature antibacterial markets. Urological disease shows few high-growth areas, with innovation shifting toward devices, surgery, and diagnostics. Autoimmune conditions such as rheumatoid arthritis and psoriasis sit nearly flat. These areas are flat, she argued, because commercial incentives, not the underlying science, are the constraint, which is where novel therapeutic options are needed.
She left the audience with four signals to carry into prioritization: follow the mechanism, since growth concentrates where biology is well-defined and matched to the right modality; treat rare diseases as addressable; think in platforms rather than single programs; and read the flat areas as signals of unmet need rather than stalled science.
What comes next
Kavita closed with a preview of the team's next analysis. Rather than examining modalities and targets separately, the group is studying how they co-occur, plotting modality-target combinations by patent volume against growth rate. The combinations in the low-volume, high-growth corner point to lower patent saturation and open opportunity. As one example, ADCs paired with FGFR2 represent a validated, fast-growing combination, while ADCs paired with CXCR2 mark a genuine white space with no such conjugate yet in development.
The throughline across all three presentations was the value of the map itself. As Angela put it in closing, individual research is often narrow and project-focused, and a landscape view helps researchers see where their work sits relative to nearby fields, where emerging targets and modalities are drawing attention, and how the past five years compare with the next.
Watch the full webinar
To hear directly from Angela Zhou, Kavita Iyer, and Janet Sasso, including the full methodology, the CAS TrendScape maps, and the audience Q&A, watch the recording on demand via ACS Webinars.
For the complete analysis, see the CAS Insights Report, Unlocking the future of pharmaceutical innovation through patent intelligence, including the full set of CAS TrendScape maps and the underlying data.
Questions from the audience
At the end of our panel discussion, we had the chance to answer quite a few questions from the audience. Here’s a summary.
Question: Compared with a similar report five years ago, what would be the biggest differences?
Answer: The clearest difference is diversification, across both therapeutic modalities and molecular targets. Five years ago, the analysis would have shown far less development in these areas. Technology has moved quickly since then, particularly in biologics, gene therapy, and cell therapeutics, which feature far more prominently on the map now than they would have five to 10 years ago. Small molecules still account for a large share of overall effort because of their distinct advantages, but biologics bring their own value, including suitability for rational design, potentially faster development timelines, and often greater target specificity. The period also opened up more of the druggable universe: targets once considered undruggable became reachable through approaches such as PROTACs , which can degrade a protein regardless of its size rather than relying on a small molecule to inhibit it directly. Genetic diseases caused by single-gene mutations have gained momentum as well, supported by advances such as recent CRISPR-based therapeutics with one of them being approved in 2025.
Question: How does the CAS TrendScape map categorization correspond to the CAS lexicon, and how often is the lexicon updated in the pharmaceutical field?
Answer: The CAS Lexicon is a standardized vocabulary and ontology of controlled terms (CAS Concepts) used to index scientific literature and patents. The analysis drew on that Lexicon while also applying natural language processing (NLP) to capture how researchers and authors describe their work in current terms. Comparing the two, roughly 80% of the extracted topics matched existing lexicon concepts, and the analysis also surfaced new terms. The Lexicon is updated continuously, both through systematic analysis of this kind and through suggestions from individual analysts who flag new terminology as they read new articles and patents. A useful byproduct of this project is that it feeds the newest terminology from the field back into the Lexicon.
Question: Small molecules still dominate patent volume despite heavy investment in biologics, RNA, and cell therapies. What does that tell us about the pace of transition in drug discovery?
Answer: The data indicates that small molecules are not going to be replaced. They carry inherent advantages in scalability, chemistry, and cost, which means they are not going away. Biologics are not displacing small molecules. The shift is one of diversification rather than replacement.
Question: What is the clinical and strategic significance of the first PROTAC approval?
Answer: The approval provides clinical proof of concept that PROTACs work in humans and can achieve the necessary bioavailability and therapeutic efficacy. It also establishes a regulatory path: with one PROTAC having completed all three phases, a pipeline is now in place that may accelerate future approvals. The design itself is significant. Because degradation requires bringing two proteins together, a single small molecule would struggle to do the job, whereas a PROTAC links two small molecules, one binding the target protein and one binding the degrader, to bring them into proximity. That gives medicinal chemists a more tractable starting point for rational design, and AI may further advance this work, particularly for the harder problem of molecular glue design, by improving prediction of binding affinity within protein pockets.
Question: Are composition-of-matter patents and separate formulation patents counted as part of the same patent family?
Answer: It depends on how the patents were filed. If a formulation patent is filed as a continuation or division of the original composition-of-matter patent, it claims priority back to the same parent application, and the analysis groups them together as one family. If the formulation patent is filed as an independent application with its own new priority date, it is counted as a separate family.
Question: When the analysis refers to corporate patents versus university patents, what does that mean?
Answer: That distinction refers to patent ownership, specifically the patent assignee. A patent filed by a company as the assignee is counted as a corporate patent. Ownership can change: if a company purchases a patent from a university, that patent then counts as corporate, because the analysis reflects the current owner.
Question: What does co-occurrence analysis reveal that traditional patent counting cannot?
Answer: Counting how many patents are growing for one modality or another does not provide context about the targets involved. Co-occurrence analysis adds that context. A target might be well-validated for a disease and heavily explored for small molecules, yet have no biologics activity around it, which points to a potential opportunity. Because the analysis is data-supported, significance comes from repetition: a single co-occurrence says little, but a pairing that recurs hundreds of times signals a pattern worth investigating. The aim is to identify emerging combinations and white space where research and patenting opportunities exist.
Question: What is the significance of the data analysis for next-generation biological interventions, particularly for genetic diseases?
Answer: Many genetic diseases are rare, and AI adds substantial capacity for literature mining and connecting scattered evidence. By linking findings across the world, including published scientific data, clinical data, and electronic patient records, it becomes possible to build a more comprehensive picture of conditions that are individually rare, potentially making them more tractable, and more viable for industry, than before.
