When Lung Cancer Research Matters to Appendix Cancer Patients

AI, KRAS, and Why Your Data Belongs in the Appendicure Registry

A new review in Expert Opinion on Drug Discovery covers a moment in appendix cancer drug discovery worth paying attention to, even though the paper itself focuses on KRAS mutations in non-small cell lung cancer.

Here is why.

Before I get into it, a confession. I am wired to believe that good things happen sooner than the timeline says. That bias colors how I read papers like this one. I have learned to talk myself down when needed. I still think the optimistic read is closer to right than wrong on this story, and I will tell you where I am applying the brakes as we go.

The KRAS problem is our problem too

KRAS is the most commonly mutated gene in the RAS family. It drives some of the most aggressive cancers, including pancreatic, colorectal, and lung. For decades it was considered undruggable. The protein is small, smooth, and gives a drug molecule very little to grab onto.

That changed in 2021 when the FDA approved sotorasib for non-small cell lung cancer with the KRAS G12C mutation. Adagrasib followed in 2022. Both drugs target one specific version of mutated KRAS, the G12C form.

Here is where appendix cancer enters the picture. KRAS mutations are reported in a large majority of mucinous appendiceal tumors and pseudomyxoma peritonei cases, often in the 70 to 80 percent range depending on subtype and cohort. The most commonly reported variant is not G12C. It is G12D. Other patients carry G12V or other variants. The KRAS G12C drugs that work in lung cancer do not help most appendix cancer patients, because most of us do not have that specific mutation.

That sentence used to be the end of the story. It is not anymore. The same researchers who built the G12C drugs have spent the last few years working on the next mutations on the list, and our most common variant is one of them.

Every one of these drugs traces back to research that started in other cancers. Lung. Pancreatic. Colorectal. The science moves first in cancers that have large patient populations and big research budgets. The work then extends into rare diseases like ours, sometimes slowly and sometimes faster than anyone expected. The KRAS G12D story is closer to the second case than the first, and that is worth sitting with for a minute.

What this means for appendix cancer drug discovery

The review by Samudrala and colleagues at the University at Buffalo walks through how artificial intelligence is now central to the search for new KRAS drugs. Their expert opinion is straightforward. AI plus laboratory work is faster and produces molecules with greater structural diversity and fewer off-target effects than traditional drug screening. They expect AI to become standard practice in drug discovery for aggressive driver mutations across multiple cancers.

The same research group built a computational platform called CANDO that screens drug compounds against large protein libraries. In earlier work they used it to predict that osimertinib, a drug approved for EGFR-mutant lung cancer, would synergize with several KRAS inhibitors. Lab testing confirmed the prediction. That is the loop the new review is describing. AI proposes, the lab confirms, and the cycle repeats much faster than it used to.

AI drug discovery has already crossed real milestones

This is the part where my optimistic streak shows. I have watched AI go from a buzzword pharma companies put in their slide decks to a tool that has actually produced a drug now in late-stage clinical trials. The work is not finished. It is also not science fiction anymore.

This is not theoretical anymore. A few examples that are worth understanding.

Insilico Medicine and the first AI-designed drug in late-stage trials

Insilico used generative AI to design INS018_055, a drug for idiopathic pulmonary fibrosis. The compound went from project start to a preclinical candidate in 18 months. The company has since reported positive Phase 2a results, with patients treated with the drug showing dose-dependent improvement in lung function compared to placebo. This is widely cited as the first drug discovered and designed by generative AI to reach this stage.

Pancreatic cancer drug combinations

A 2025 study from researchers at NCATS, MIT, and the University of North Carolina built machine learning models on data from 496 drug combinations tested in pancreatic cancer cells. The models then predicted synergy across 1.6 million possible combinations. The best-performing model achieved an 83 percent validation rate in laboratory testing of predicted synergistic combinations, identifying 307 verified hits. The point is not the specific drugs and the point is not clinical efficacy. The point is the scale. No human team could have tested 1.6 million combinations in the lab.

AlphaFold and protein structure

AlphaFold is the AI system that dramatically improved researchers’ ability to predict how proteins fold. It does not itself design drugs. What it does is provide the foundational step for structure-based drug discovery. Knowing a protein’s three-dimensional shape is what makes it possible to design molecules that bind to it. AlphaFold made that step faster and more accurate, which makes everything downstream, including the work on KRAS, more tractable.

Why none of this works without patient data

Here is the part of the story I have the most hope about, because it is the part we can actually control.

Appendix cancer drug discovery, like all AI-driven research, is only as good as the data the models learn from.. For KRAS drug discovery in lung, pancreatic, and colorectal cancer, the input data exists in enormous quantities. Tumor sequencing data. Treatment records. Outcomes. Imaging. Tissue samples in biobanks. Decades of clinical trial data.

For appendix cancer, those datasets are tiny. The disease is rare. The histologic subtypes are split across five categories under the NCCN framework, including LAMN, HAMN, appendiceal adenocarcinoma, goblet cell adenocarcinoma, and undifferentiated carcinoma not otherwise specified. Each subtype has its own biology and its own treatment patterns. When you split a rare cancer five ways, then split it again by grade and KRAS variant, you end up with patient subgroups that may number in the hundreds nationally.

This is the reason the same KRAS G12D drug that gets developed for pancreatic cancer may take years longer to be properly tested in appendix cancer. There are simply not enough patients in any single institution’s records to power the analyses that AI tools need.

Rarity is not the only problem. Appendix cancer is also unusually hard for computational analyses because the underlying data is messy in ways specific to our disease. Some centers report grade as a three-tier system, G1 through G3. Others collapse it into low and high grade. Pathology reports may classify the same tumor as mucinous adenocarcinoma at one institution and as a goblet cell adenocarcinoma variant at another. Treatment sequences vary widely, especially around CRS and HIPEC, and the outcome reporting from those procedures is not standardized. Even when a patient has full sequencing, the record may sit in a format that machine learning tools cannot use without significant manual cleanup.

That is what the Appendicure data registry is built to change.

Every patient who contributes data makes the next analysis sharper. If you have a KRAS G12D mutation and your record sits in one hospital, it is a single data point in that hospital. If your record is in the registry alongside hundreds of others, it becomes part of a signal a researcher can actually study. This is not theoretical. We are doing this right now with CDK4/6 inhibitors, working alongside Dr. Shen at MD Anderson and Dr. Lowy at UC San Diego to push NCCN to recognize these drugs for appendiceal disease with GNAS and KRAS mutations.

The honest caveats

This is the part of the post that hurts to write. I lean optimistic on these stories and I have to make myself slow down. An old sales book I read years ago taught me “Hope is Not a Strategy“, and our community deserves the real picture, not the rosy one. The caveats below are me applying the brakes on myself, because hope is not a strategy and our community deserves the actual picture and not the rosy one. Both things can be true at once. The science is genuinely promising, and I have to tell you exactly where it is and is not.

First, the lung cancer review does not mention appendix cancer. The science is relevant because we share the KRAS biology, not because the authors are working on our disease. We should not overstate the connection.

Second, the appendix cancer data on KRAS inhibitors is still very early. The April 2026 preprint reports clinical observations from six patients. That is a starting point, not a clinical practice change. The MRTX1133 work in appendiceal models is preclinical. The zoldonrasib trial included appendiceal patients but did not report disease-specific efficacy. As noted earlier, organoid and xenograft results often do not translate to humans, which is why these early signals need to be interpreted cautiously.

Third, AI in drug discovery has produced impressive early signals but has not yet delivered a generation of approved cancer drugs designed entirely by AI. The Insilico fibrosis drug is the furthest along, and it is for a non-cancer disease. The field is moving fast. It is not finished.

None of this changes the core argument about appendix cancer drug discovery.. KRAS mutations drive most appendiceal adenocarcinoma. Drugs targeting KRAS G12D and multi-selective RAS inhibitors are now in clinical testing and include appendix cancer patients. AI is accelerating the pace at which new candidates reach those trials. Our data is what makes our disease visible to those efforts. That is the part I will let myself feel hopeful about, with both eyes open.

What you can do

• Ask your medical team about next generation sequencing if you have not already had it. Knowing your tumor’s mutation profile, including whether you carry a KRAS mutation and which variant, matters for current and future treatment options.
• Ask whether trials of multi-selective RAS inhibitors or KRAS variant-specific inhibitors are open at academic centers that treat appendix cancer.
• Add your data to the Appendicure Patient-Led Data Registry. It is the single biggest contribution you can make to appendix cancer drug discovery. It takes very little time. It is one of the highest-impact things any of us can do for patients who come after us. If you can’t figure out how to access your medical records, Contact Us One of us will respond the same day.
• Share this with anyone in our community who is weighing whether registry participation is worth the effort. The argument is no longer abstract. It is the difference between our disease being included in the next wave of drug development or being left behind.

“Hope is not a strategy. It is also not optional. The strategy is the registry. The hope is what keeps us building it.”

Keep Reading: What a New AI Biomarker Could Mean for Appendix Cancer Patients

Sources

Samudrala R, Bruggemann L, Falls Z, Mahajan SD. Combining cutting edge computational and experimental methods for targeting KRAS mutations in non-small cell lung cancer. Expert Opinion on Drug Discovery, 2026.

doi.org/10.1080/17460441.2026.2654614

KRAS inhibition is an effective therapy for appendiceal adenocarcinoma. bioRxiv preprint, April 2026.

biorxiv.org/content/10.64898/2026.04.07.717107v1

Vazquez-Borrego MC et al. Antitumor effect of a small-molecule inhibitor of KRAS G12D in xenograft models of mucinous appendicular neoplasms. Experimental Hematology and Oncology, 2023.

doi.org/10.1186/s40164-023-00465-4

Bruggemann L et al. Multiscale Analysis and Validation of Effective Drug Combinations Targeting Driver KRAS Mutations in Non-Small Cell Lung Cancer. Pharmaceuticals, 2023.

pmc.ncbi.nlm.nih.gov/articles/PMC9867122

Insilico Medicine. Positive Topline Results of ISM001-055 for Idiopathic Pulmonary Fibrosis, November 2024.

Insilico Medicine press release

OncLive. RAS(ON) Inhibitor Zoldonrasib Is Safe, Shows Early Activity Signals in KRAS G12D Mutated PDAC and other histologies including appendiceal cancer, May 2026.

onclive.com (zoldonrasib coverage)

Artificial Intelligence-Driven Innovations in Oncology Drug Discovery. Drug Design, Development and Therapy, 2025.

pmc.ncbi.nlm.nih.gov/articles/PMC12232943