I recently attended AACR 2025 in Chicago from April 25-30. This post contains takeaways and disclosures that I found interesting.
This year, there seemed to be less basic science and a larger focus on drugs and therapeutic platforms. Many academic groups spent time over the last year studying a drug’s resistance mechanisms, potential synergies, or effects on the tumor microenvironment. There is greater appreciation that real progress is being made with therapeutics entering clinical trials, which perhaps means that any sort of basic science being done is becoming more and more tenuous. For example, far fewer people care about CML basic biology after the advent of imatinib, and most of the interesting research progress in CML has shifted to studying the onset of disease, rather than its progression. A similar thing is happening in other tumor types and you start to wonder whether there really need to be this many scientists (21k+ at AACR this year) in cancer.
This year, there was a new requirement to upload posters. All posters are searchable on a web portal until October 31 2025. The search engine doesn’t work very well, so to search for pharma programs, I took this approach:
Over the last couple of weeks, I have watched most of the talks that I have deemed to be immediately relevant to drug development and picked through the 6500+ posters for interesting disclosures. Some of the talks were “Do not post” so I won’t be sharing any screenshots. Here are some patterns I found interesting:
While not every shiny new tool was discussed at AACR, platforms utilizing novel technology are most certainly back.
Tumor specific T cell engagers (both protease activatable, or using tumor specific antigens), have made exciting progress since the early days of Blincyto. More IgG-like formats in combination with FcRn binding mutations are enabling longer half life to make dosing more flexible. This not only provides convenience benefits to patients, but also enables greater control over PK/PD and the therapeutic window. Protease activatable binders (e.g. Cytomx and Janux platforms) have greatly enhanced both the target landscape and the dosing capabilities by limiting on-target, off-tumor activity. Never write off a drug class (or target) without fully exploring the exposure-efficacy relationship. Finally, novel formats that engage the immune system in more optimal manner than the crude 1+1 format (i.e. 1 CD3 binding arm and 1 tumor antigen binding arm), have started to further push the boundaries on therapeutic window by decreasing the risk of cytokine release syndrome (CRS). One of which, that was disclosed a year ago but had more data presented at AACR, is AstraZeneca’s TITAN platform, which selectively engages CD8 T cells, rather than CD4 T cells which are the main drivers of CRS risk.
ADCs (dual payload, bispecific, etc) have also greatly improved through engineering approaches. Hundreds of clinical trials have told us what the best payloads are (TOPi), what the mechanisms of resistance are (paylod resistance), and what defines a strong target (e.g. high antigen density, like HER2). Continuing research is telling us additional mechanisms of action, how to reduce off target toxicity (e.g. Fc engineering), how to reduce on target, off tumor toxicity (e.g. Cytomx EPCAM ADC, bispecific approaches), and how to improve potency (e.g. dual payload, Mythic’s pH dependent binding approach, engineering internalization, binder affinity, bystander activity, etc). Agile manufacturing approaches like Sutro’s cell free platform with non-natural amino acids can be used to rapidly iterate on drug antibody ratios (DAR), payload choices, biophysical properties, and payload positioning.
Receptor internalizing approaches (e.g. bispecifics, biparatopics, certain small molecules) can provide more potent inhibition of a receptor signaling addicted tumor. Targets include EGFR, cMET, HER2, and FGFR2. The biology here is that binding can induce conformational changes that mimic ligand engagement, promoting endocytosis. Amivantamab (cMET x EGFR), Merus’ Petosemtamab (EGFR x LGR5), Zanidatamab (HER2), a set of academic FGFR2 antibodies from Bill Sellers and Nabeel Bardeesy, and Enliven’s (now former) HER2 small molecule ELVN-002 are all capable of receptor internalization. This property is useful intrinsically, but also may be useful for the design of ADCs, which often require internalization for efficacy. A new approach published in Nature Chem Bio earlier this year claims that simply conjugating PEI to an internalizing antibody can promote degradation and prevent recycling. Combining this approach with an already internalizing antibody could result in more sustained receptor inhibition.
Molecular glues/PROTACs have been transformative for several years now. We will get into this, but for KRAS inhibitors, molecular glues that sterically inhibit interactions with downstream partners have been remarkable. Below is a slide from Jian Jin at Mount Sinai, who chaired a session on the discovery of new PROTACs and molecular glues. His lab, and others including Shaomeng Wang have seemingly made it their mission to make a degrader for every oncogenic protein.
Improved small molecule inhibition chemistry (allosteric, covalent, and more selective inhibitors) is unlocking new pharmacology. Several interesting molecules described in later sections including a more selective CDK4i, a Y220C covalent activator of p53, an allosteric mutant selective PI3K inhibitor, a highly selective ABL1 kinase inhibitor, and a selective WT HER2 inhibitor that induces receptor internalization.
Radiopharmaceuticals continue to make strides, just slower. We’ve written extensively on radiopharm before. One interesting target AstraZeneca is exploring for prostate cancer is STEAP2. They have 3 approaches: an ADC, a radiopharmaceutical from Fusion Pharma, and a T cell engager. With STEAP2, the idea is to avoid xerostomia observed with PSMA targeting, as PSMA is expressed within salivary gland tissue.
If I had to guess, there are maybe 100+ KRAS targeting agents in development. These range from the 1st generation covalent inhibitors, to molecular glues, and even to exosome RNAi based approaches. The flurry of activity in KRAS is the clearest evidence to me that target crowding can be a good thing! The pace of improvement in pharmacology is terrific and has provided definite benefits. The resistance mechanisms to 1st generation covalent G12C inhibitors are overcome by the molecular glue approaches. Even improvements in the binding kinetics for the covalent class of inhibitors can yield clinical benefits (Sotorasib -> Adagrasib -> Divarasib -> D3S-001, which achieved an 89% ORR in CRC as a monotherapy, compared to just 9.7% for sotorasib). If we were to have written off D3 Bio’s molecule as another Chinese me-too, we would have lost out on serious clinical benefit. The Monday morning quarterbacking has now told us that disappointment from the first generation of covalent G12C off state inhibitors will be short lived.
A result of the remarkable progress in KRAS inhibitor development is that PDAC 5-year OS will likely triple in 5 years (13% -> ~40%). Daraxonrasib will be approved in the 1L and 2L metastatic settings, and likely also in the resectable adjuvant setting. Combinations with PRMT5 inhibitors in MTAP deleted patients (22%), HER2 ADCs (2%; but higher in KRASi resistant patients), and CLDN18.2 targeted approaches (60% of pts) will serve as the obvious additional tailwinds that already have biomarkers. If KRAS + chemo in the first line is tolerable, improvements in OS could be even more spectacular. Several groups have reported that KRAS inhibitors and chemotherapy target complementary cell states (basal and classical respectively) in heterogeneous PDAC tumors.
KRAS combination studies will be a very fruitful avenue of clinical research, so much so that I think that it is rate limiting for biology research in certain contexts. Everyone can speculate on resistance mechanisms or combos (there are maybe 150-200 groups working on combos with KRAS), but ultimately, clinical trials will provide answers. More sophisticated strategies like combining multiple KRAS targeting strategies (e.g. KRAS inhibitors + degraders, or allele specific + panKRAS), FAK inhibitors, CDK4/6 inhibitors, PI3K/mTOR pathway inhibitors, immunotherapy (e.g. CTLA-4 or cancer neoantigen vaccines) and everything else that is trying to compete for an ‘all comers’ population (vs above mentioned approaches which will most likely be approved in biomarker defined populations) will need to play out in clinical trials. No one is incentivized to run the head to head comparisons in mice.
There are a few new strategies I found interesting from AACR:
Not just with KRAS but targeting oncogenes with combinations of two agents against the same protein has been shown to be effective (e.g. HER2 targeted ADCs with HER2 targeted small molecules). Revolution Medicines had clinical data showing the efficacy of this strategy, combining a G12C or G12D specific glue compound (Elironrasib or Zoldonrasib) with their panKRAS G12X glue Daraxonrasib.
BridgeBio had a talk discussing combining BBO-8520, a first-in-class direct dual inhibitor of GTP-bound (ON) and GDP-bound (OFF) KRASG12C, with a PI3K breaker compound or cetuximab. In mice, they didn’t observe toxicity with either approach. It may be obvious to say, but therapeutic window with KRAS targeting agents is already proving to be important for monotherapy activity, and will be even more important for combinations. Having a variety of chemistries and targeting mechanisms with varied off-target activity and potency I’m sure will be incredibly useful for chemists designing the next generation of molecules.
I and many others are rooting for FAK inhibition/degradation. The preclinical rationale is very strong and emerging data from Verastem at ASCO validates the anti-tumor activity of targeting FAK in combinations. Targeting KRAS fundamentally changes the signaling of addicted tumors, opening up opportunities for agents that didn’t show activity as monotherapy.
Allorion has a selective CDK4 that has ~2x better selectivity than Pfizer’s selective CDK4 inhibitor. Though their HQ is in Natick, MA, this is yet another Chinese drug company.
5 different LY6G6D targeting agents had poster presentations. Cartography’s 1+1 TCE CB21, Antengene’s (China) 2+1 TCE ATG-110, Genentech’s 1+1 TCE Linclatamig, HEC Pharma Group’s (China) LY6G6D / 41BB bispecific HEC-921, and EpimAb’s (China) tetravalent Fabs-In-Tandem Ig (FIT-Ig) EM1034. Genentech got to LY6G6D first and had disappointing clinical data. Out of 765 that consented to pre-screening, only 604 submitted a tumor sample, 182 (31%) had a LY6G6D+ tumor, and only 46 were enrolled in the study. Most patients very quickly progressed (mDOR = 1.6 months) and only 1 patient had a cPR (2.2%). Skin tox (dry skin, rash, pruritus) was highlighted by investigators as potential on-target off-tumor activity. There are perhaps two lessons for those who spent their academic careers building single cell atlases; A: Anchoring on novelty is not necessarily a good thing. Here, Cartography picked a target that Genentech got to first, but without access to a trove of preclinical data showing the limits of anti-tumor efficacy, they are holding the bag on a lead program that is no longer exciting because the target is now considered weak. B: There needs to be some level of engineering to surpass the first round of copycat molecules. Ultimately, you need competence engineering many different constructs and the ability to develop high(er) throughput assays to measure them, to be competitive.
AstraZeneca Titan platform for TCEs. This is ~1 year old so I’m a little late, but it looks exciting. The construct contains a TCR and CD8 nanobody binding domain, as well as two antigen binding arms. This selectively activates CD8 T cells and has reduced CRS compared to the typical 1+1 approach (largely because CRS derives from CD4 T cells.
New synthetic lethal pathway was published in Nature by two groups (Broad & Abbvie) and Tango had a preprint and abstract (funny how everyone always converges at once).
Mythic Therapeutics is developing pH dependent antibodies for ADCs. This disclosed data at ASCO. The platform is similar to sweeping antibodies, where there is binding at neutral pH and release (no binding) at pH 5.5. However, instead of trying to degrade an extracellular protein, the goal is to enable greater trapping of antibody in the cell (sometimes antibody gets returned to the cell surface when the receptor is recycled) and reduced TMDD. The result is the ability to 3-5x exposure in tumor cells with ~1/2 the exposure in normal tissues, enabling targeting of tumors with lower antigen expression. Perhaps disappointingly, this approach has enhanced on-target toxicity. cMET is expressed in the eyes, and 49% of pts dosed had blurred vision, 44% had keratopathy. The choice of an MMAE payload likely made things worse given the previously observed association of ocular toxicity with MMAE payloads.
A couple new mutant selective PI3Ka inhibitors: TYK Medicines, STX-478, Ensem Therapeutics ETX-636 allosteric mutant selective PI3Ka inhibitor, LAE118 improved activity on helical domain mutants, but no hyperglycaemia data (Laekna Therapeutics), GSC002639, a brain penetrating and mutant-selective PI3Kα inhibitor (Changchun GeneScience Pharmaceutical). Just like KRAS inhibitors, this will be a class that continues to get better over time from improvements to the therapeutic window.
Sutro’s Site specific conjugation technology for non-natural amino acids provides a high throughput and flexible optimization platform to rapidly explore payloads and empirically optimize DAR ratios. As new formats of ADC continue to get explored, the platform could become extremely valuable. The market cap is tiny (<100 mn and deeply negative EV) and the company is burning a bunch of cash on seemingly undifferentiated ADCs.
The Broad PRISM team (multiplexed cellular screening platform) had two interesting posters. First, they found that CD32a expression correlated with ADC efficacy in cell lines not expressing the target antigen. The upshot is that engineering out Fc engagement might reduce off-target toxicity from non-specific uptake. Secondly, they showed that the efficacy of Datopotamab deruxtecan (Datroway) is dependent on TROP2 expression, while Sacituzumab govitecan (Trodelvy) is not. Trial data from these two drugs will continue to read out over the next several years.
HER2 inhibitors (e.g. neratinib) are effective in HER2 low cells. HER2 high and HER2 low cells grow faster in heterogeneous cultures. Provides rationale for combining HER2 small molecule inhibitors, and surface targeted approaches like ADCs which have improved efficacy in HER2 high cells
LUT014 is a topical WT BRAF inhibitor for EGFR inhibitor induced rash. This is being developed by a company called Lutris Pharma. The idea is that inhibition of WT BRAF causes paradoxical MAPK pathway activation. This can offset the effects of MAPK pathway inhibition in the skin, which would otherwise cause acneiform lesions.
Several interesting degraders had oral presentations. These included PLX-513, a CDK2 degrader with greater selectivity than CDK2 inhibitor, developed by Plexium, CM-486, an ALK degrader from BMS, and NRX-4972, an AURKA degrader from Nurix.
DHX9, an ATP dependent RNA helicase, is now druggle thanks to Accent Therapeutics. This is being trialed in tumors with high levels of replication stress (ie. BRCA1/2 mutated or MSI-hi/dMMR). The DepMap profile indicates that DHX9 is pan-essential so tolerability will be a key question.
Frontier Medicines is a company applying covalent fragment based drug discovery approaches to cancer. Their FMC-220 program is a p53 Y220C covalent activator, that induces p53 gene expression in cancers that have that specific p53 mutation.
Dostarlimab in MSI-hi/dMMR cancers across all histologies (NEJM) continues to impress. This trial split patients into a rectal cancer cohort and a broad solid tumor cohort. The efficacy in rectal cancer was again a remarkable 100% CR rate. Strong activity was also observed in urothelial and colon cancer, with reduced activity in gastric, GEJ, prostate, and endometrial tumors.
In the academic world, there were two interesting posters to me. First, a group has claimed to identify potential new cFLIP (aka CFLAR) degraders? The data shown thus far is somewhat unconvincing so TBD, but would be awesome to see. Secondly, HNF1B emerged in TF ORF drug anchored screens as a transcription factor capable of promoting resistance to KRAS inhibition.
I specifically searched through posters and talks for new drugs that pharma or smaller oncology focused biotechs are developing.