Archive | Haystack Corner RSS feed for this section

The Needle Issue #25

14 Apr
Juan-Carlos-Lopez
Juan Carlos Lopez		 Andy-Marshall
Andy Marshall

The approval of multiple anti-amyloid monoclonal antibodies (mAbs) — aducanumab (Aduhelm; now withdrawn), lecanemab (Leqembi) and donanemab (Kisunla) — over the past five years has opened the era of disease-modifying Alzheimer’s drugs, albeit with only modest benefits in addressing cognitive decline (30% slowing) and associated serious safety risks, such as CNS inflammation and cerebral hemorrhages, which has limited clinical uptake. While many drug development programs target biological processes other than amyloid formation (e.g., tau and tangles, neurotransmitter receptors, neuroinflammation, autophagy, and mitochondrial or metabolic dysfunction), companies continue to optimize anti-amyloid monoclonals, but also look for alternative ways to therapeutically target Aβ.

One alternative therapeutic modality to antibodies is chimeric antigen receptor (CAR) immune cell therapy. In recent weeks, we have been thinking a lot about in vivo chimeric antigen receptor (CAR)-T therapies, which were one of the dealmaking trends in 2025, and we recommend readers check out an excellent summary of trends in the area from the consultancy firm Scitaris (you don’t even have to give them your details to download the report).

CAR-T treatments have established their clinical niche as last-ditch treatments for B-cell malignancies, with some remarkable outcomes for late-stage patients. In some cases, they have been shown to be at least twice as effective as T-cell engager bispecific antibodies in clinical studies. But they remain rather blunt instruments.

Despite advances in the clinical management of cytokine-release syndrome and immune effector cell neurotoxicity syndrome (ICANS), CAR-T treatments continue to be associated with serious risks. And while there have been advances in managing these adverse events, atypical non-ICANS neurotoxicities (NINTs) can also create serious clinical management issues, with risk factors predisposing patients to development still only poorly understood.

That said, over the past year, we have seen an increasing trend for the use of CAR-T treatments outside oncology. They have started to be applied with promising efficacy in various areas of autoimmunity (systemic lupus erythrematosus, lupus nephritis, systemic sclerosis, Sjögren’s syndrome, antisynthetase syndrome, myasthenia gravis and idiopathic inflammatory myopathies) and neuroinflammatory conditions (multiple sclerosis). In this respect, a recent paper in Science caught our attention. In it, Marco Colonna and his colleagues at Washington University in St. Louis harness astrocytes to clear amyloid plaques by promoting their ability to phagocytize Aβ.

To that end, they used in vivo gene therapy to generate astrocytes carrying chimeric antigen receptors (“CAR-As”), a strategy not unlike the one used in cancer immunotherapy. Although both macrophages (CAR-Ms) and conventional CAR-Ts have been tested in preclinical models of Alzheimer’s disease with limited success, this study reports the first attempt to directly engineer astrocytes in the body to generate CAR-As.

In broad terms, the construct used to generate CAR-As consisted of an Aβ-binding domain and the phagocytic signaling protein MEGF10 (multiple epidermal growth factor-like domains protein 10). The team examined a variety of constructs and chose two for in vivo testing. One of them combined a fragment from the Aβ-binding antibody crenezumab and MEGF10, which is primarily expressed in astrocytes. The second construct combined a fragment of aducanumab with the phagocytosis receptor Dectin-1, which is primarily expressed in microglia.

The authors packaged the constructs in an adeno-associated viral (AAV) vector under the control of an astrocyte-specific promoter and injected them intravenously into 5xFAD mice (which carry five familial Alzheimer’s disease (FAD) mutations, driving rapid Aβ plaque formation, synaptic loss, and cognitive decline starting around 2–4 months). Both CAR-As reduced amyloid burden and neuritic dystrophy, and the treatment worked both in the prophylactic and therapeutic settings.

Single-nucleus RNA sequencing and immunostaining showed that the CAR-As adopted the transcriptomic profile of activated astrocytes and readily clustered around amyloid plaques. Microglial cells, in turn, also responded to the treatment by showing a reduction of the disease-associated transcriptomic profile that is often seen after administration of monoclonal anti-Aβ antibodies. This is of interest because this disease profile of microglial cells has been suggested to contribute to the inflammatory reaction sometimes seen after Alzheimer’s immunotherapy.

A caveat of the study is that the authos saw no improvements in cognition following therapy, albeit behavioral results in mouse models have been notoriously poor at predicting outcomes in humans. However, the translational questions don’t stop there.

If in clinical practice the CAR-A approach would require an AAV vector, then immunogenicity of the treatment is going to be an issue. Pre-exposure to AAV is often a problem for gene-therapy programs, where patients are much younger. Given that Alzheimer’s is a disease associated with an elderly population, immunogenicity is likely to be exacerbated. Similarly, the delivery of 1013–1014 viral genomes to elderly patients living with Alzheimer’s—many of whom will already have a brain prone to neuroinflammation—makes the specter of unwanted side effects a major concern. In this respect, finding Alzheimer’s patients whose disease stage and age would be appropriate for a therapy with potentially highly toxic consequences for fragile recipients is also difficult to gauge.

That is not to say that CAR-immune cell therapy may not have a place in CNS disease. It just seems like neurological conditions, such as multiple sclerosis where patients are younger and potentially less fragile, are the place where much of the translational groundwork and clinical management for CAR-A or CAR-T therapies must be worked out before moving into neurodegenerative disease for elderly and cognitively compromised patients.

Tags: , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,

The Needle Issue #24

24 Feb
Juan-Carlos-Lopez
Juan Carlos Lopez		 Andy-Marshall
Andy Marshall

X-ray crystallography has long been the go-to workhorse for providing atomic structures of drugs interacting with their protein targets. Increasingly, those static snapshots are being complemented by readouts from experimental analytical tools based on nucleic magnetic resonance (NMR) spectroscopy and cryoelectron microscopy (cryo-EM), offering drug developers a broader window into proteins as dynamic, breathing molecules. This is spurring a raft of new service provider startups, including AIffinity (Brno-Medlánky, Czech Republic), NexMR (Zürich, Switzlerand), CryoCloud (Utrecht), and Intellicule (West Lafayette, IN), all of which aim to supply drug-discovery teams with state-of-the-art platforms providing structural data with rapid turnaround times and low cost.

As many of the most compelling ‘undruggable’ targets are renowned shape shifters — aggregation-prone proteins like Tau, amyloid precursor protein (APP) or huntingtin in neurodegenerative diseases, or transcription factors like P53, KRAS and c-MYC in oncology — a lot of therapeutic startup activity has recently focused around so-called ‘intrinsically disordered proteins’ (IDPs). The ability to attain markedly different conformations under different conditions allows IDPs not only to play moonlighting roles or serve as hubs in signaling networks, but also to localize into liquid- phase condensates (or membrane-less organelles — attributes that make them acutely sensitive to mutations that can compromise specificity and lead to nonspecific binding, resulting in toxicity and disease.

As IDPs frequently resist attack by conventional drug discovery approaches, a slew of startups has sprung up to try to go after this target class, many using new structural techniques. These include Peptone (London, UK), Dewpoint Therapeutics (Boston, MA), brainQR Therapeutics (Göttingen, Germany), and Kodiform Therapeutics (Oxford, UK). Just last month, Topos Bio secured a $10.5 million seed round to “tackle ‘undruggable’ proteins driving Alzheimer’s and cancer”. Dewpoint also just announced it has dosed its first patient in a phase 1/2a trial of its lead beta-catenin program in gastric cancer and elected its MYC development candidate to take forward.

An important postscript to the startup activity targeting undruggable IDPs is that more conventional ‘druggable’ target classes, like tyrosine kinases, may also represent a fruitful hunting ground for dynamic conformational states that may have been missed by traditional crystallographic approaches. Given that conventional drug targets have relatively well-trodden clinical and commercial development paths, they may also represent simpler starting points and testing grounds for commercial programs aiming to apply the new analytical approaches to support medicinal chemistry programs around validated targets.

In a paper recently published in Science, the team of Charalampos (Babis) Kalodimos at St. Jude Children’s Research Hospital use high-resolution NMR spectroscopy to gain structural insight into how SRC family tyrosine kinases (Src, Hck, and Lck) achieve processive phosphorylation of multisite substrates.

The SRC enzyme family is essential for rapid and coordinated signaling in processes such as cell migration and T-cell activation. In addition, SRC family kinases are frequently overexpressed in tumors, contributing to the activation not only of multiple scaffold or signaling proteins, such as receptor tyrosine kinases (e.g., EGFR, FGFR, PDGFR or IGF1R), but also of downstream effectors (e.g., MAPKs, FAK, paxillin, p130Cas, ELMO1 and RAC1). Although there are approved drugs like the multikinase inhibitor Sprycel (dasatinib) that bind the SRC active site, these drugs have such extensive off-target and adverse side effects that there is a pressing need for new paths to more-selective SRC inhibitors.

SRC enzymes share a conserved domain organization, with a disordered N-tail, a tandem SH3–SH2 module, a kinase domain, and a disordered C-tail. All can carry out processive phosphorylation — a phenomenon where the enzyme phosphorylates multiple residues in a substrate during a single encounter. Each of these catalytic cycles typically requires ATP binding, phosphate transfer and ADP release, and ADP release is often the rate-limiting step. So, a question that has long puzzled structural biologists is how ADP-release–constrained kinases achieve sufficiently rapid turnover to successfully perform their function.

Using NMR spectroscopy with cryogenic probes — which reduce electronic/thermal noise and increase sensitivity up to five-fold compared with room-temperature probes — the St. Jude team characterized the conformational ensemble of the Src kinase domain and identified three interconverting states: a predominant active state, a previously described inactive Src/CDK-like state, and a hitherto unknown low-populated intermediate state positioned linearly between the other two. Structural determination revealed that this intermediate state displays features that are distinct from the active and inactive states. Its activation loop is partially folded, the P-loop is displaced inward, and the αC helix is shifted upward. This conformation binds ADP poorly relative to the active and inactive states, suggesting that it facilitates nucleotide release.

Using mutational analyses, the researchers then confirmed the functional importance of this intermediate state. Variants that eliminated this intermediate state while stabilizing the active state showed slower ADP dissociation, reduced catalytic turnover and impaired processive phosphorylation of the multisite Src substrate p130Cas. Instead of generating a fully phosphorylated substrate in a single binding event, these mutants accumulated partially phosphorylated intermediates. Equivalent mutations in other kinases of the SRC family, Lck and Hck, similarly reduced catalytic efficiency and impaired multisite phosphorylation of their respective physiological substrates CD3ζ and ELMO1 in Jurkat cells. Furthermore, these mutations compromised cellular functions measured via in vitro assays, including T-cell activation using Lck-deficient Jurkat cells and migration of mouse embryo fibroblasts lacking Src, Yes and Fyn in the presence of fibronectin. These molecular and functional findings indicate that the intermediate state is evolutionarily conserved and essential for processive activity across the SRC family.

Mechanistically, the work establishes that rapid ADP release, enabled by transient sampling of a structurally constrained intermediate, is critical for sustaining catalytic turnover rates that exceed the speed of substrate dissociation. More broadly, it shows that kinase conformational landscapes are tuned not only for switching between active and inactive states, but also for optimizing specific kinetic steps within the catalytic cycle.

From a drug developer’s standpoint, because Sprycel and other inhibitors target the active or inactive conformations of the SRC active site, the identification of a low-populated, functionally indispensable intermediate suggests a completely new strategy to target tyrosine kinases: selectively stabilize or destabilize the intermediate state to fine-tune catalytic turnover and processivity rather than simply blocking activity. Targeting such transient conformations could enable more precise modulation of signaling output, potentially improving selectivity and reducing off-target effects in kinase-directed therapies.

We look forward to seeing how many more of these intermediate states are uncovered in other kinase targets and whether pharmacological inhibitors targeting this state have advantages over orthosteric or allosteric chemotypes that conventionally have been used to inhibit the kinase active site or lock it in an inactive conformation. What is clear is that ultrafast NMR measurements of binding and state behavior are a powerful differentiating tool for understanding kinase activity where static structures aren’t enough.

Tags: , , , , , , , , , , , , , , , , , , , , , , , , , , , ,

The Needle Issue #23

10 Feb
Juan-Carlos-Lopez
Juan Carlos Lopez		 Andy-Marshall
Andy Marshall

In our past issue, we took a look at all the financing deals that The Needle has covered since our inaugural issue. This week we turn our attention to last year’s deal making in the preclinical biotech space.

In 2025, preclinical dealmaking didn’t just slow — it polarized. Capital clustered around AI-enabled discovery, China-sourced assets, and in vivo CAR-T cell therapies, while entire therapeutic categories effectively disappeared from licensing activity. Based on the 131 publicly disclosed preclinical transactions in our sample, we reveal where early-stage risk capital is still flowing — and where it has quietly retreated.

Similar to the data we reported in our past newsletter, our analysis captures only publicly disclosed deals (partnerships, research collaborations, licenses, joint ventures, reverse mergers, equity investments and options) on business wires, industry news sites, and venture-fund sources. In the preclinical space, many deals are carried out in stealth, and companies in some important regions (like China) don’t use business wires or news sources traditionally available in the West. For these reasons, our estimates underestimate the true level of early-stage preclinical dealmaking.

In total, we tracked 131 preclinical deals over the year, of which 42 were licensing deals, 64 were strategic partnerships/collaborations and 14 were mergers and acquisitions (M&As). In keeping with early stage’s exploratory nature, the importance of stealth, and the non-compensatory nature of much of the work done, over half of the publicly announced strategic partnerships (35 deals; 55%) had no terms disclosed. As one would expect, a smaller proportion of the licensing deals failed to provide terms, but even for this category, 8 of the 48 transactions (17%) didn’t give financial details. Four of the 14 M&As that we tracked also made no mention of deal terms.

US-headquartered companies continue to dominate the dealmaking landscape, whether it is research collaborations, licensing or trade sales. One reason for the dominance of companies in the US — and the UK, which is second in deal activity — is likely simple math; a greater number of companies are financed and built in these countries compared with the rest of the globe (see The Needle Issue #22).

Strategic partnerships in 2025 favored platforms over products — and Western biotechs over Asian peers.

The 64 strategic partnerships we tracked had upfront payments that ranged from $5 million to $110 million, but the median ($35.5 million) underscores how concentrated value remains in a handful of outlier platform deals.

US companies accounted for 37 of the 64 deals (58%). Three notable partnering big-ticket deals involved biotechs splashing out large sums on preclinical collaborations, with the payers showing interest in branching out into new therapeutic modalities: last May, CRISPR Therapeutics (San Diego, CA) pivoted from gene editing to siRNA, paying $95 million to Sirius Therapeutics (Shanghai, China) to co-develop a long-acting siRNA designed to selectively inhibit Factor XI for thrombosis; in December, Regeneron Pharmaceuticals (Tarrytown, NY) spent $150 million (and made an equity investment) to jointly develop Tessera Therapeutics’ (Somerville, MA) target-primed reverse transcription therapy (TSRA-196), which uses lipid nanoparticles (LNPs) to deliver RNAs encoding an engineered reverse transcriptase (‘gene writer’), writer-recognition motifs, and a SERPINA1 template to correct a mutation in alpha 1 antitrypsin deficiency; and later the same month, peptide developer Zealand Pharma (Søborg, Denmark) announced a transaction with OTR Therapeutics (Shanghai, China), paying $20 million upfront for small-molecule programs centered around validated targets of Zealand’s franchise in cardio-metabolic disease.

For obvious reasons, target discovery and drug screening comprise about a third of collaborations and partnership agreements, but do not figure much in licensing and M&A. Mentions of machine learning in partnering deals (18.2% of 2025’s deals, with several in the top 10 grossing set) suggest large-language and other models are an increasingly established facet of preclinical development. Neurodegenerative disorders garnered the second largest number of partnering transactions in our 2025 sample. And, with all the noise around GLP-1s and other incretins, metabolic disease and obesity were the focus of 11% of deals.

Perhaps the most counterintuitive finding in the partnership data is the near-total absence of China-headquartered companies — despite their dominance in preclinical licensing. This may reflect geopolitical friction, IP risk tolerance or a Western preference for control in collaborations. Alternatively, the absence may reflect the limitations of Haystack’s methodology for collecting data. Certainly, the partnership data contrasts starkly with our licensing data, which show Chinese assets performing so well that they are biting at the heels of US companies and running far ahead of UK companies. In contrast, for strategic partnerships, it was UK-, and South Korea-based firms that were most prominent behind the US (15%, and 7% of dealmaking, respectively).

For licensing, the shift to Asia seen in later parts of the biotech pipeline is also manifest in the preclinical space.

Chinese companies were involved in nearly a quarter of all the licensing deals made last year, clinching 11 out of the 48 deals we tracked. This interest in early-stage Chinese assets mirrors last year’s banner deals for later-stage assets, such as Pfizer’s ex-China rights acquisition of 3SBio’s (Shenyang, China) PD-1 x VEGF bispecific antibody for $1.25 billion, or GSK’s $1.10 billion acquisition of Jiangsu Hengrui’s (Lianyungang, China) phosphodiesterase 3/4 inhibitor and oncology portfolio. Overall, deals seeking access to assets from Asian biotechs (companies based in China, South Korea, Singapore and Taiwan) comprised 33% of all preclinical licensing transactions in our sample.

Looking at the preclinical licensing as a whole, upfront amounts ranged from $0.7 million to $700 million, with a median value of $35 million. Most deals centered around cancer, followed by autoimmune, neurodegenerative and metabolic diseases.

What was perhaps most surprising is that we didn’t see any licenses for preclinical assets in the cardiovascular space, suggesting that the interest of a few years ago has somewhat diminished (although assets for heart disease still made up 4% of partnering agreements). Notably absent from preclinical licensing in 2025: cardiovascular, pulmonary, skeletomuscular, hepatic, pain, psychiatry, women’s health, sleep, hearing, and stroke. This pattern perhaps reinforces the industry’s retrenchment toward genetically anchored, biologically de-risked indications. Together, these licensing gaps underscore a 10-year low in early-stage risk appetite outside traditional blockbuster categories.

The top 10 licensing deals from last year are listed in the Table below. Of this elite tier of top-grossing deals, cancer and autoimmune comprised the lion’s share (70%), with neurodegenerative, neurodevelopmental, metabolic, and ophthalmic disease all represented. Only two of the top 10 deals involved traditional small molecules (with one additional license for a molecular glue), whereas biologics accounted for seven. While small molecules still comprise the biggest chunk of licensing activity (18.9%), deals trended toward bispecific and multispecific antibodies for cancer immunology and autoimmune indications — and biopharma was prepared to pay: Of the 8 licensing transactions for multispecifics in our sample, IGI Therapeutics’ (New York, NY) deal with Abbvie, and CDR Life’s (Zurich, Switzerland) agreement with Boehringer Ingelheim, ended among the top 10 grossing deals of the year.

Which leads us to mergers. Overall, we tracked 14 M&A deals last year in the preclinical space. According to Dealforma data presented at JP Morgan, private biopharma accounted for just over 55% of merger activity in 2025 on par with previous years. In the Haystack data, 12 of the 14 acquisitions for preclinical programs were for US-based private companies, reinforcing the historical trend of American biotechs outperforming those in the rest of the world in terms of negotiating successful exits for their investors.

The biggest story in early-stage mergers from last year, though, was biopharma’s ravenous appetite for in vivo CAR-T cell therapy, with Capstan, Orbital and Interius comprising 3 of the 14 acquisitions recorded by Haystack, all of which ranked among the top 5 highest upfront payments. As our sampling commenced in April 2025, we missed another deal: AstraZeneca’s acquisition of lentiviral in vivo CAR-T therapy developer Esobiotec, originally announced in March 2025 with an upfront of $425 million. All in all, in vivo CAR-T therapies claimed 4 of the top 5 acquisitions last year.

The use of lipid nanoparticles (LNPs) in many of these in vivo CAR-T platforms (Orbital, Aera Therapeutics, Stylus Medicine, MagicRNA, Orna Therapeutics, Byterna Therapeutics and Strand Therapeutics) and elsewhere (Tessera, Starna Therapeutics, Nanovation Therapeutics, United Immunity, Genevant Sciences, Pantherna Therapeutics, eTheRNA Immunotherapies, and Beam Therapeutics) also underlies a continuing theme of investment and dealmaking around drug delivery platforms.

Apart from LNPs, several drug delivery deals also centered around antibody shuttles that can take biologics and siRNAs across the blood–brain barrier into the CNS. These included Manifold Bio/Roche, Vect-Horus/Secarna, Ophidion/Neuronasal, JCR/Acumen and Denali/Royalty Pharma. This year will see more of these shuttles enter clinical testing, with Alector’s transferrin shuttle AL137, a subcutaneous anti-amyloid beta antibody, slated for an IND submission.

In sum, the preclinical dealscape in 2025 reveals an industry willing to fund innovation — but only when paired with platform leverage, delivery, or late-stage optionality. As Haystack tracks dealmaking through 2026, the key question will not be whether capital returns to early-stage biotech, but whether it broadens beyond today’s narrow set of ‘acceptable’ risks. We look forward to tracking deals throughout 2026 and identifying new emerging trends in biotech deals.

 

Tags: , , , , , , , , , , , , , , , , , , , , , , , ,

The Needle Issue #22

27 Jan
Juan-Carlos-Lopez
Juan Carlos Lopez		 Andy-Marshall
Andy Marshall

As is customary at the turn of the year, we have taken the opportunity to take a look back at financing deals we covered since issue#1, which went live in April last year. Together, these data offer a snapshot of how capital flowed into early-stage, preclinical therapeutic startups in 2025 — and where it did not.

Before diving into the numbers, it is worth qualifying that this analysis captures only publicly disclosed financing rounds, rather than the full universe of early-stage biotech funding. An increasing fraction of preclinical companies now operate in stealth, in part because of fast-moving competition from regions such as China. As a result, the figures presented here likely undercount the true level of early-stage activity.

From the start of our coverage in Q2 2025 through the end of December, we reported 195 preclinical financing rounds. Because Haystack Science focuses on discovery-stage and pre-IND companies, this number excludes financings for assets already in clinical development. Even so, the dataset provides a useful lens on early-stage investor behavior.

Independent industry analyses paint a consistent picture. Multiple sources indicate that 2025 was a year in which venture capital shifted toward later-stage, clinical-stage deals, which were fewer in number but larger in size. This trend was reinforced by ‘Q4 2025 Biopharma Licensing and Venture Report’, presented at the JP Morgan conference. According to JP Morgan, 2025 saw just 191 seed and Series A financings, the lowest total since 2020.

According to the Haystack Science data sample, no venture fund made a series A investment in more than three companies last year (these series A financings ranged from $8–300 million, with a median of $42.5 million). As the deals that Haystack tracks are only the publicly disclosed subset, we expect our sample is skewed to companies that raised larger sums. In the deals we tracked, the most bloated series A ($300 million) went to Cambridge, Mass.-based Lila Sciences, a generative ML model powered startup building “autonomous, closed-loop experimentation using generative ML models to generate drug mechanism hypotheses, test them robotically in the lab with minimal human intervention, and iteratively learn from results.” Lila was backed by megafund Flagship Pioneering and General Catalyst.

21 funds invested in more than one series A round. These were: Arch Ventures, Atlas Venture, Lightstone Ventures, 3E Bioventures, Access Industries/Biotechnology, BGF, BVF Partners, Canaan Partners, Cormorant Asset Management, Dementia Discovery Fund, Eight Roads, Johnson & Johnson Innovation – JJDC, Khosla, Omega Funds, Orbimed, Polaris Partners, Samsara, Santé Ventures, Sofinnova Partners, The Column Group, and Versant Ventures. No fund invested in more than 3 series A investments in last year’s sample.

Further back in the pipeline, we tracked 60 deals. These seed financings—which ranged from $1.1–54.5 million with a median of $10.45 million—were mostly for smaller amounts ($1–$30 million), with a few much larger financing amounts. Overall, 85 different funds, family offices, angels and individuals participated in funding preclinical therapeutic startups in 2025. Of these 85 sources of financing, only 7 financed more than one company. The takeaway from this is that most (>90%) of companies at the seed stage receive funds from a completely unique set of investors.

The 7 financing entities involved in more than one seed deal were: AdBio Partners, Kurma Partners, NRW Bank, Ackermans & van Haaren (AvH), Bioinnovation Institute (BII), ClavystBio and ExSight Ventures. It is noteworthy that two of these funds are based in Paris, France: AdBio Partners and Kurma Partners. AdBio specializes in early-stage investments across Europe with a ~€86 million ($102 million) fund raised in 2021 focusing on oncology, immunology, and rare diseases. Kurma is part of the Eurazeo group, managing >€600 million in assets across several funds focused on early-stage therapeutics and diagnostics.

NRW.BANK, based in North Rhine-Westphalia, Germany, invests in innovative biotech companies focusing on tech-driven healthcare, bio-digital integration, and novel platforms for data/discovery, aligning with broader innovation goals. They appear to be an important source for the small scattering of financing (13) deals in German-speaking countries. NRW works closely with AvH, an Antwerp, Belgium-based diversified holding company and investment firm, with AvH Growth Capital a proactive investor in early-stage companies like DISCO Pharmaceuticals and Evla Bio.

Another very interesting seed funder is BII in Copenhagen Denmark. The institute provides in-kind grants of up to €3 million for bridging translational studies in European academic institutions. For those projects that progress to a company build, a combination of convertible loans of €500K (Venture Lab) and then €1.3 million (Venture House) are made available to complete seed funding. As of January 2026, BII has supported over 130 early-stage life science and deep tech companies, with many attracting significant external funding. This month, there was news that Novo Nordisk has just plowed another $856 million of funding into BII.

Overall, in terms of the location of where most investment is occurring, our analysis reveals the capacity to host startups is expanding across the globe, with at least 19 countries hosting one preclinical startup that received funding in 2025. These countries were: USA, UK, France, Switzerland, China, The Netherlands, Canada, Denmark, Germany, Belgium, Japan, Spain, Israel, Australia, Ireland, Norway, Portugal, South Korea and Singapore. Perhaps the prominence of France as a location for preclinical therapeutic startups was most surprising from our sample. Interestingly, a lot of ex-US startups now also have a US (usually Cambridge, Mass.-based) headquarters. Digging deeper, 85 different cities around the world host a startup that obtained financing (pre-seed to series B) in 2025, with 20 cities hosting two or more. As expected, the Boston cluster led with 28 preclinical therapeutic startups, the Bay area hosted 19, and the UK’s Golden Triangle had 13. Of the following pack, some interesting standout cities were Paris, France (with 5 in our sample) and New York City (with 7), the latter long in the shadow of its Boston neighbor.

In terms of the disease areas attracting early-stage investor money, cancer dominates, comprising the focus for 34.4% of the funding raises. This is slightly lower than the biopharma sector as a whole, where cancer comprises up to 45% of pipelines. Following cancer, neurodegenerative disease, autoimmune disease and inflammatory disease all figured prominently. The uptick in deals for companies tackling CNS disorders has been a rolling theme recently, given the burden of neurodegenerative disease and dementia on public health systems and the paucity of disease-modifying treatments. With the continuing stampede around GLP-1s/incretins, there was also a healthy number of metabolic/ endocrine disease startups financed.

One last area we looked at was the type of therapeutic being financed by investment groups. Here again, the pharmaceutical industry’s traditional workhorse, the small molecule, remained pre-eminent in 2025, comprising 24% of financing deals in pre-seed, seed, series A and series B financings that were in the preclinical stage. Established modalities like monoclonal antibodies (mAbs) were a common focus. And there was a resurgence of interest in recombinant proteins and peptides (likely boosted by the focus on incretins and the metabolic disease and obesity space). Of new modalities, antibody-drug conjugates, bispecific and multispecific antibodies, antisense oligonucleotides (ASOs), small-interfering RNAs (siRNAs) and chimeric antigen receptor (CAR) immune cell (T cell and NK cells) also were to the fore, each making up around 6% of all the early-stage deals we tracked. A type of therapeutic gathering increasing attention is clearly the induced-proximity therapeutic sector (including the different flavors of PROTACs, DUBTACs and molecular glues). Finally, although a great deal has been mentioned about investor apathy for gene editing and gene therapy, these also captured 3-4% of the deals.

Tags: , , , , , , , , , , , , , , , , , , , , , , ,

The Needle Issue #21

6 Jan
Juan-Carlos-Lopez
Juan Carlos Lopez		 Andy-Marshall
Andy Marshall

On December 9, the Italian charity Fondazione Telethon made waves by becoming the first non-profit organization to obtain FDA approval for an advanced therapy: Waskyra (etuvetidigene autotemcel) is an ex vivo lentiviral gene therapy indicated for the rare immune deficiency Wiskott-Aldrich Syndrome. Fondazione Telethon’s accomplishment underscores the impact that philanthropic organizations can have on drug discovery and has rightly been celebrated by patient-advocacy groups working to develop therapies for other conditions of limited commercial interest. How can this wider universe of disease foundations emulate Fondazione Telethon’s achievement and leverage the lessons from Waskyra’s approval?

Drug development for rare and ultra-rare conditions faces multiple challenges: limited understanding of the disease, paltry funding, a lack of business models providing a return on investment, regulatory obstacles, manufacturing and distribution barriers, and so on. For all these reasons, venture capitalists and pharma companies have shied away from diseases that, like Wiskott-Aldrich Syndrome, affect small populations of patients. This is the unspoken dirty secret of modern medicine. Current commercial drug development is unfit for >90% of all known diseases.

With the biopharma industry steering clear of these conditions, patient advocacy groups and other charities are trying to fill the void. According to a recent study commissioned by the US Department of Health and Human Services (HHS), 585 advocacy groups fund “medical product development” activities in the United States. Why has it taken an Italian non-profit organization to be the first to cross the US FDA approval finish line?

The organization responsible for development of Waskyra is the San Raffaele Telethon Institute for Gene Therapy (SR-TIGET), a 30-year-old partnership between Telethon Foundation and Milan’s Ospedale San Raffaele. Over those three decades, SR-TIGET has raised over half a billion euros in philanthropic capital to build internal capabilities equivalent to those available in a clinical-stage biotech company: target discovery, preclinical modelling, regulatory strategy, phase 1/2 clinical trials and registration. In other words, unlike most patient foundations and groups, this organization has accumulated the resources to generate the data necessary to walk the full path to approval, independently of the need to collaborate with a pharmaceutical company.

Out of the 585 patient advocacy groups cited in the HHS report, only 11 operate with their own research staff and lab space. In contrast, 106 advocacy groups fund life-science companies, and 536 fund academic or medical institutions. This implies that, at most, only 1.9% of US patient groups use a model that shares at least some similarities with SR-TIGET’s. This is important to emphasize because having all these in-house capabilities makes an organization less dependent on industry partnerships, which can be difficult to secure in the first place and are subject to change if economic conditions and/or company priorities alter.

Of course, it would be disingenuous to expect all patient foundations to adopt the SR-TIGET model. According to the HHS report, the mean annual revenue of an advocacy group capable of funding clinical trials is ~$32 million, with their median annual revenue at ~$3.5 million. Most of the 585 charities have no hope of achieving these financing levels, particularly those advocating for patients living with ultra-rare conditions. At the same time, these figures represent the reality of commercial development, and they should be part of the calculus used by patient advocacy groups to define the scope of their activities and inform their fundraising strategy.

It is worthwhile noting that Waskyra is not Fondazione Telethon’s first rodeo. SR-TIGET was responsible for much of the work behind two other approved ex vivo lentiviral gene therapies for ultrarare conditions: Strimvelis (for ADA-SCID; European approval in 2016) and Lenmeldy (for metachromatic leukodystrophy; European approval in 2020; FDA in 2024). In both cases, the organization partnered with for-profit companies to take the drugs to market, providing SR-TIGET with crucial training in the drug-approval process before they achieved their recent independent success with Waskyra. At the same time, those early experiences made it painfully clear that the story does not end with regulatory approval, as many without experience of developing medicines assume.

In 2018, Strimvelis, which had been developed by SR-TIGET in collaboration with GlaxoSmithKline, was acquired by Orchard Therapeutics along with the rest of the pharma’s rare disease gene-therapy portfolio. After taking the therapy to approval, however, Orchard pulled the plug and decided to cease marketing of the therapy. Fondazione Telethon then stepped in and had to arrange the transfer of the marketing authorization from the company to the foundation. Although SR-TIGET has been able to make the therapy available in Italy, Strimvelis remains unavailable elsewhere in Europe. This is unsurprising as setting up distribution networks across continents requires deep expertise and investment, and has long been the sole purview of commercial organizations.

In the case of Waskyra, the manufacturing and distribution strategy for the United States is not yet clear, but a week after the FDA decision, Fondazione Telethon signed a memorandum of understanding with the Orphan Therapeutics Accelerator (OTXL) under which Orphan Therapies (an OTXL subsidiary) will become the exclusive commercialization partner for the therapy. OTXL is a separate, US-based, non-profit organization focused on the clinical development of “shelved” ultra-rare disease treatments. That two independent non-profit organizations have come together to deliver a life-changing therapy to patients is of great significance and perhaps underappreciated by the wider community. It will be interesting to see how this partnership evolves, particularly with regards to pricing.

Indeed, pricing has been another thorny issue for Fondazione Telethon. The cost of Strimvelis is reportedly ~€600K. Between July 2023 (when the foundation obtained the marketing authorization) and the end of 2024, SR-TIGET has treated only two ADA-SCID patients (~14 children are born every year with the disease in Europe). Of most concern, the associated costs for these two treatments were €4.7 million. Although Fondazione Telethon is a non-profit entity, multi-million Euro losses of this kind simply are unsustainable. It will therefore be important that the foundation sets a price of Waskyra on the US market where it can at least recoup the costs of its treatment — if not make a return that it can invest back in further R&D efforts.

Which brings us to perhaps the most important takeaway from SR-TIGET’s Waskyra approval. It is striking how this foundation has focused very heavily on the development of gene therapies, and in particular ex vivo lentiviral gene therapies. Luigi Naldini, leader of SR-TIGIT, is a pioneer in the study of lentiviral vectors, and a lot of the research conducted at the institute over the years has focused on the optimization of vectors and on understanding the biology of hematopoietic stem cells with the eventual goal of fixing disease-causing mutations. According to the SR-TIGET website, the organization has treated ~25% of patients who have received hematopoietic stem cell-based gene therapy worldwide.

In contrast, most patient groups have a starting point around a specific disease (or a subset of related diseases) for which drug-discovery projects are launched, often using multiple therapeutic modalities to have as many “shots on goal” as possible. These are two fundamentally different approaches. SR-TIGET has focused on one therapeutic modality and then deployed it across different diseases; most other foundations focus on one disease and then invest in many different therapeutic modalities.

Ultra-rare drug developers and patient groups should take note: an increasing body of data suggests that organizations achieving development success have adopted a similar platform-based approach to bringing therapeutics to patients. And the reason for this is simple: putting together an entire discovery, commercialization and distribution apparatus for more than one therapeutic modality is simply unaffordable for most independently funded non-profits.

There are now several examples to illustrate this point. In the field of antisense oligonucleotides (ASOs), n-Lorem Foundation has achieved success using solely the ASO modality, with >35 kids suffering from 17 different “nano-rare” diseases now treated: CHCHD10/ALS, TARDBP, LMNB1, ATN1, SCN2A encephalopathy, PACS1, ASXL3/Bainbridge Ropers, MAPK8IP3/ALS, hnRNPH2/ASD, H3F3/chondrosarcoma, KIF1A/KAND, UBTF/CONDBA, TUBB4A-related leukodystrophy, EPL1/familial dysautonomia, serum amyloid A amyloidosis, or FLVCR1 and PRPH2 retinopathies. Again, success has been achieved by developing a single modality across an incredibly wide range of nano-rare neurodegenerative, neurodevelopmental, autonomic nervous system, kidney and retinal diseases.

For adenoviral associated virus serotype 9 (AAV-9) gene therapy, social purpose corporation Elpida Therapeutics continues to make progress with its platform for ultra-rare conditions (recently receiving an $8 million grant from the Center for Regenerative Medicines) Again, Elpida is focusing on just one modality and developing it against multiple neurodevelopmental and neurodegenerative conditions: Charcot-Marie-Tooth disease type 4J, Spastic Paraplegia 50 (SPG50), and Neuronal Ceroid Lipofuscinosis 7 (CLN7). Similarly, Nationwide Children’s Hospital, which carried out the original work leading to approval of Novartis’ AAV-9 gene therapy (Zolgensma) for spinal muscular atrophy, has deep resources and expertise, enabling it to serve as a hub for this type of gene therapy. In recent weeks, it announced the start of a clinical AAV-9 program for SLC6A1 neurodevelopmental disorder.

Elsewhere, one might argue that, in base editing, we are also starting to see yet another example of a modality hub emerge. Following the success of base editing around CSP1 for baby KJ (highlighted in Issue #6 of The Needle), the Center for Pediatric CRISPR Cures is building a hub around gene editing R&D expertise — an initiative that the Innovative Genomics Institute’s Fyodor Urnov is also promoting.

What does all this mean? We would suggest that academic medical centers and patient foundations interested in developing ultrarare therapies should consider the platform-based approach as an efficient way to deploy their capital. Evidence is clearly building that focusing on one modality works. For therapies beyond that single modality, organizations might be better served by identifying another resource-rich ‘hub’ organization for development programs in their disease.

Another advantage of a large platform-based hub approach with a host of different disease spokes is that it would result in a diversified portfolio of projects in which each project is a separate shot on goal. This may achieve the scale to deliver a successful drug and, therefore, generate income. In fact, MIT economist Andrew Lo has used financial-engineering techniques to show that a portfolio of ultra-rare disease projects could generate a return on investment exclusively from the sale of FDA’s Priority Review Vouchers (PRVs), which pharma companies seek to acquire for a median >$100 million. Although the reauthorization of the PRV program by the US Congress is uncertain, we think this is a tantalizing insight because it points to a sustainable path for the development of ultra-rare therapies.

2025 has been a landmark year for ultrarare therapies. Besides the FDA approval of Waskyra, the successful use of base editing to treat CPS1 deficiency in Baby KJ in just seven months, the acceptance of >160 patients into n-Lorem programs, and the administration of several gene therapies to ultrarare patients (Urbagen, an AAV-9 gene therapy for CTNNB1 syndrome being yet another recent example) suggest that ultrarare disease treatments are finally gaining momentum. With SR-TIGET, n-Lorem, Nationwide Children’s and Elpida showing the way, perhaps a development model is finally emerging to treat these debilitating childhood diseases that devastate too many families around the world.

Tags: , , , , , , , , , , , , , , , , , , , , ,

The Needle Issue #20

9 Dec
Juan-Carlos-Lopez
Juan Carlos Lopez		 Andy-Marshall
Andy Marshall

By our count, there are now 15 bi-specific antibodies approved by the US Food and Drug Administration (the last peer-reviewed count from 2024 we found chalked up 13). This year has been a bumper year for bi-specifics — antibodies that recognize two molecular targets. Several of 2025’s largest deals have involved assets in this class, including Genmab’s $8 billion acquisition of Merus in September and Takeda’s $11.4 billion splurge on an anti-Claudin18.2 bi-specific antibody and antibody-drug conjugate (ADC) from Innovent Biologics.

Not only is this trend likely to continue, but we predict that it will expand to encompass tri- and multi-specific antibodies, the development of which is an area of intense research activity. Just a couple of weeks ago, South Korea’s Celltrion clinched a $155 million (biobucks) deal for TriOar’s tri-specific ADCs for cold tumors. And at the SITC meeting last month (which we covered in issue 19) tri-specifics were highlighted by no less than five companies: Nextpoint (B7-H7 x CD3 x TMIGD2), CrossBow (cathepsin G peptide x CD3 x CD28), TJ Biopharma (CDCP1 x CD3 x 4-1BB), Biocytogen (DLL3 x CD3 x 4-1BB) and Radiant Therapeutics (potentially tri-specific/trivalent).

Building an antibody that recognizes three or more targets at the same time is not trivial, though. There are multiple technical, clinical and regulatory hurdles that developers need to overcome before the antibody reaches patients. Why, then, go through the trouble of creating a multi-specific antibody when a bi-specific may show clinical benefit? As it turns out, there are several reasons why a multi-specific antibody may be worth the effort.

First, as tumors often escape by downregulating or mutating a single target epitope, a multi-specific antibody may reduce the likelihood of escape by simultaneously targeting multiple tumor antigens. Second, multi-specifics could increase safety and reduce toxicity of a therapy. For example, a multi-specific antibody can be designed to require co-expression of two or more antigens on the same cell to bind effectively. Healthy cells expressing only one antigen would be spared, thereby reducing off-tumor toxicity. Similarly, targeting multiple mechanisms with a single antibody may reduce the need to use several separate drugs, simplifying dosing and reducing risks for patients. Third, and perhaps most important, a multi-specific antibody can simultaneously block several disease pathways, yielding synergistic effects that a bi-specific might not achieve. In solid tumors, for example, tumor heterogeneity, limited immune-cell infiltration and an immunosuppressive microenvironment often result in therapeutic failure. Multi-specific antibodies could combine tumor targeting, immune-cell recruitment and checkpoint modulation in a single molecule.

Perhaps the best example of this comes from the field of T-cell engagers (TCEs). A tri-specific antibody can incorporate not only tumor-cell binding and CD3 engagement, but also a co-stimulatory domain, such as CD28. This can boost T-cell activation, persistence and potency more than a bi-specific that only binds to CD3.

In this regard, a recent paper in PNAS is an excellent example of the power of the approach. A research team from EvolveImmune Therapeutics reports on the development of EVOLVE, a next-generation TCE that integrates CD3 binding with CD2-mediated co-stimulation to enhance T-cell activation, durability and tumor-killing capacity, while avoiding target-independent toxicity.

Conventional CD3-bi-specific TCEs activate T cells through a stimulation signal but often fail to provide the complementary co-stimulation necessary for sustained effector function. This can result in T-cell dysfunction, reduced persistence and limited clinical durability. To address this, Jeremy Myers and his colleagues systematically compared multiple costimulatory pathways and identified CD2 as a superior target owing to its broad expression on naïve, activated and exhausted CD8⁺ T cells, and its sustained expression within tumor-infiltrating lymphocytes.

The team engineered tri-specific antibodies that fuse a CD58 extracellular domain (the natural CD2 ligand — Lymphocyte Function-Associated Antigen 3;LFA-3) to affinity-tuned CD3 binders within an IgG-like format. They showed that integrated CD2 co-stimulation substantially improves T-cell viability, proliferation, cytokine production and cytotoxicity across tumor types.

When optimizing the molecule, they found that CD3 affinity must be attenuated: high-affinity CD3 domains cause target-independent T-cell activation and cytokine release (superagonism), whereas intermediate-affinity variants retain potent tumor-directed killing with reduced off-target activation.

The EVOLVE tri-specifics outperformed matched bi-specifics targeting HER2, ULBP2, CD20 and B7-H4, with increases up to >50-fold in potency, depending on the target. The optimized tri-specifics also showed superior tumor control in vivo, achieving durable tumor regression in humanized mouse models even after cessation of the treatment.

Even though tri- and multi-specific antibodies could offer clear advantages over bi-specifics, they are not without problems. From the technical standpoint, multi-specifics combine multiple binding specificities and often non-natural architectures. This feature increases complexity at every step from discovery to manufacturing. The assembly of IgG-like multi-specifics can result in heavy/light and heavy/heavy chain mispairing leading to heterogeneous products. Although antibody engineers have come up with strategies to address this issue, each solution adds constraints to developability.

Multi-specific antibodies can also have lower expression, cause more host-cell stress and require more advanced cell-line engineering or multi-vector expression systems. Moreover, downstream purification often needs additional steps to separate mis-paired species. Similarly, multi-specific antibodies are often less stable, more aggregation-prone, and more sensitive to formulation conditions, impacting shelf life and immunogenicity risk.

It is also important to show identity, purity and functional activity for each specificity and for the multi-specific activity (that is, simultaneous binding, cell-bridging). So, establishing robust potency assays is often the greatest challenge. What is a good model system to design a development candidate going after several targets at the same time? With each additional binder, complexity in discovery and development increases.

From the clinical standpoint, although multi-specifics can potentially be safer than bi-specific antibodies, as we mentioned above, other toxicological risks exist.

TCEs have been known to trigger cytokine-release syndrome, neurotoxicity, or unexpected tissue toxicity if targets are expressed on normal tissues. First-in-human dosing strategies are therefore critical. Moreover, multi-specifics may have non-linear pharmacokinetics (target-mediated clearance for each target), and dual-target engagement can alter distribution and half-life; selecting a safe, effective dose requires integrated PK/PD modeling and biomarker strategy.

And the headaches don’t stop there. Efficacy of a multi-specific may depend on co-expression of two or more targets. Stratifying patients may therefore complicate trial enrollment and endpoint definition, not to mention that it may be necessary to develop companion diagnostics (already expensive and complex for conventional monoclonal antibodies). And related to this point, when multiple targets are engaged, it can be hard to know which specificity caused an adverse event, complicating risk–benefit evaluation and mitigation.

Finally, from the regulatory perspective, although expectations are still evolving, agencies expect a pharmacological package that reflects multi-specific mechanisms, particularly with regards to toxicology. Regulators routinely require robust control strategies to ensure product consistency. Again, this is going to be more complicated for multi-specifics because small changes in manufacturing can alter pairing or potency.

Multi-specific antibodies are gaining momentum. They represent a potentially powerful technology, but many questions still surround their development. Success may depend on striking the right balance between choosing the appropriate therapeutic indication, identifying the simplest effective format, heavy upfront developability and analytical work, and early interactions with regulators to align on pre-clinical packages.

Tags: , , , , , , , , , , , , , , , , , , , , ,

The Needle Issue #19

25 Nov
Juan-Carlos-Lopez
Juan Carlos Lopez		 Andy-Marshall
Andy Marshall

Although therapeutic antibodies represent a $160 billion-dollar annual market and comprise a third of all approved drugs, discovering new antibody molecules remains a labor-intensive process, requiring slow experimental approaches with low hit rates, such as animal immunizations and or the panning of phage- or yeast-displayed antibody libraries. The drug hunter’s dream would be to design an antibody to any target by simply entering information about that epitope into a computer. Now that dream is one step closer with a recent proof of principle peer-reviewed paper published in Nature on work disclosed last year from the team of 2024 Nobel Laureate David Baker. Baker and his colleagues at the University of Washington introduce the first generalizable machine-learning method for designing epitope-specific antibodies from scratch without relying on immunization, natural antibody repertoires, or knowledge of pre-existing binders.

Unlike small-molecule drug development, which has benefitted from an explosion of interest in the use of machine-learning models, in-silico design of antibody binders has lagged far behind. One reason for this is the paucity of high-resolution structures of human antibody–antigen pairs—currently only ~10,000 structures for 2,500 antibody-antigen pairs have been lodged in SAbDab (a subset of the RCSB Protein Data Bank). Most of these structures are soluble protein antigens, but there’s little data to model antibody binders to GPCRs, ion channels, multipass membrane proteins and glycan-rich targets, which are of most commercial interest. Overall, the antibody–antigen structural corpus is orders of magnitude smaller, noisier and narrower than that available for small molecules, lacking information on binding affinities and epitope competition maps via PDBBind/BindingDB/ChEMBL.

For these reasons, most companies have focused on machine learning prediction of developability properties—low aggregation, high thermostability, low non-specific binding, high solubility, low chemical liability/deamidation and low viscosity—for an antibody’s scaffold, rather than in-silico design of the six complementarity determining-regions (CDRs) on the end of an antibody’s two binding arms.

Even so, several recently founded startups have claimed to be using machine-learning models to predict/design antibody binders from scratch. These include Xaira Therapeutics, Nabla Bio, Chai Discovery and Aulos Bioscience.

Xaira debuted last year with >$1 billion in funding to advance models originating from the Baker lab. Nabla Bio also raised a $26 million series A in 2024, publishing preprints in 2024 and 2025 that describe its generative model (‘JAM’) for designing VHH antibodies with sub-nanomolar affinities against the G-protein coupled receptor (GPCR) chemokine CXC-motif receptor 7 (CXCR7), including several agonists. In August, Chai announced a $70 million series A financing based on its ‘Chai-2’ generative model disclosed in a preprint that details de novo antibodies/nanobodies against 52 protein targets, including platelet derived growth factor receptor (PDGFR), IL-7Rα, PD-L1, insulin receptor and tumor necrosis factor alpha, with “a 16% binding rate” and “at least one successful binder for 50% of targets”.

Finally, Aulos emerged with a $40 million series A in 2021 as a spinout from Biolojic Design. This program has generated computationally designed de novo CDR binders with picomolar affinities for epitopes on HER2, VEGF-A, and IL-2. The IL-2 antibody (imneskibart; AU-007)—designed to selectively bind the CD25-binding portion of IL-2, while still allowing IL-2 to bind the dimeric receptor on effector T cells and natural killer cells—reported positive phase 2 results in two types of cancer just last week. Absci, another more established company, has also been developing de novo antibodies, publishing a generative model for de novo antibody design of CDR3 loops against HER2, VEGF-A and SARS-CoV-2 S protein receptor binding domain.

Overall, though, computational efforts have largely optimized existing antibodies or proposed variants once a binder already exists. Recent generative approaches have often needed a starting binder, leaving de novo, epitope-specific antibody creation as an unmet goal. The Baker paper now provides a generalizable, open-source machine-learning approach that can find low nanomolar antibody binders to a wide range of targets.

To accomplish this task, the authors use RFdiffusion, a generative deep-learning framework for protein design, extending its capabilities by fine-tuning it specifically on antibody–antigen structures. Their goal was to enable the in-silico creation of heavy-chain variable domains (VHHs), single-chain variable fragments (scFvs), and full antibodies that target user-defined epitopes with atomic-level structural accuracy.

Their approach integrates three major components: backbone generation with a modified RFdiffusion model, CDR sequence design via the algorithm ProteinMPNN, and structural filtering using a fine-tuned RoseTTAFold2 predictor (the authors note that improved predictions can now be obtained by swapping out RoseTTAFold2 for AlphaFold3 developed last year by Google Deepmind and Isomorphic Labs). The refined RFdiffusion model can design new CDRs while preserving a fixed antibody framework and sampling diverse docking orientations around a target epitope. The resulting models generalize beyond training data, producing CDRs unlike any found in natural antibodies.

Baker and his colleagues created VHHs against several therapeutically relevant targets, including influenza H1 haemagglutinin, Clostridium difficile toxin B (TcdB), SARS-CoV-2 receptor-binding domain, and other viral or immune epitopes. High-throughput screening via yeast display or purified expression led to the identification of multiple binders, typically with initial low affinities in the tens to hundreds of nanomolar range. Cryo-EM confirmed near-perfect structural agreement between design models and experimental complexes, particularly for influenza haemagglutinin and TcdB, demonstrating atomic-level accuracy across the binding region and the designed CDR loops. To enhance affinity, the authors used OrthoRep, an in-vivo continuous evolution system, for the affinity maturation of selected VHHs. The affinity of the resulting VHHs improved by roughly two orders of magnitude while retaining the original binding orientation.

Baker and his team further challenged their method with the more difficult problem of de-novo scFv design, which requires simultaneous construction of six CDR loops across two amino acid chains. The team introduced a combinatorial assembly strategy in which heavy and light chains from structurally similar designs were mixed to overcome cases where a single imperfect CDR would compromise binding. This enabled the discovery of scFvs targeting the Frizzled epitope of TcdB and a PHOX2B peptide–MHC complex. Cryo-EM validation of two scFvs showed that all six CDR loops matched the design model with near-atomic precision.

Future work is needed to extend de novo antibody prediction via this method to tougher target classes, such as membrane proteins. Clearly, modeling across all six CDR loops and the heavy and light chains remains a hard problem; indeed, the paper’s marquee result was designing a single scFv where all six CDRs matched the designed pose at high resolution; more generally, scaling reliable heavy- and light-chain co-design beyond a few cases remains an open engineering challenge that future methods will need to solve. For the field to gather momentum, benchmarking efforts like the AIntibody challenge will be needed, together with public efforts to create datasets of negative binding data, akin to those described in a paper published earlier this year.

Overall, the Baker paper is seminal work that establishes a practical and accurate approach to designing epitope-specific antibodies from scratch. It represents a major advance in the development of therapeutic antibody discovery.

Tags: , , , , , , , , , , , , , , , , , , , , ,

← Older Entries