Did AI Create a Cancer Drug in 30 Days? What the Cited Paper Reports

What the post claims

A widely shared Instagram post says an AI system created a cancer drug in 30 days. It says the drug targets breast, lung and pancreatic tumors. It says the system screened more than 1.4 million compounds. It names a candidate, TH5427, said to block a protein called MTH1. It says the drug worked in lab and animal tests and spared healthy cells. It cites Chem. Sci. 2023, 14, 1443-1452.

What the cited paper is

That citation points to one paper. It is "AlphaFold accelerates artificial intelligence powered drug discovery: efficient discovery of a novel CDK20 small molecule inhibitor." Ren and colleagues published it in Chemical Science on January 10, 2023. We read its abstract in PubMed and Europe PMC records. We could not open the full text, so details found only in the full text are not confirmed here.

What the abstract reports

• Cancer type. The abstract names hepatocellular carcinoma, a common form of liver cancer. The target was a protein called CDK20. The abstract does not mention breast, lung or pancreatic cancer.
• The drug name. The abstract does not mention TH5427 or MTH1. The molecule it names is ISM042-2-048, aimed at CDK20. We did not confirm that TH5427 or MTH1 appear anywhere in the paper.
• The 30 days. The abstract says a first "hit" molecule was found "within 30 days from target selection and after only synthesizing 7 compounds." A hit is an early starting point. It is not a finished medicine. The 30 days covers finding that hit, not making a drug.
• The 1.4 million. The abstract gives no such number. We could not confirm it.
• What the AI did. The authors used AI software to pick the protein to study, to use AlphaFold (a program that predicts protein shapes), and to generate candidate molecules. People then made the selected molecules and tested them in lab assays.
• The tests. The abstract reports lab tests only. A second-round molecule bound to CDK20 and blocked its activity in a test tube. It also slowed the growth of a liver cancer cell line (Huh7) at a lower concentration than a comparison cell line (HEK293). The abstract reports no animal tests and no human trial.
• "Sparing healthy cells." The comparison was between two cell lines grown in a dish. That is a hint of selectivity. It is not proof that healthy tissue in a person would be spared.

The authors describe the work as the first use of AlphaFold in finding a hit.

What this means

The paper reports an early lab result for liver cancer cells. It does not report a cancer drug, a human trial or an animal study. Several details in the post (the cancer types, the drug name, the 1.4 million figure and the animal tests) do not match the abstract of the paper it cites. We cannot say where the post got them.

From lab result to approved treatment

NCI describes the clinical trial phases, which come after lab work:

• Phase 1 tests whether a new treatment is safe and what its side effects are. NCI says these trials are done in "a small group of people (around 15 to 30)."
• Phase 2 looks at whether the treatment works against cancer, in about 50 to 100 people, while safety is still watched.
• Phase 3 compares the new treatment with the current standard treatment. It can involve 100 to several thousand people, assigned by chance.
• FDA review. NCI says results from phase 1 to 3 trials are used to decide about approval by agencies such as the FDA.
• Phase 4 follows long-term safety and benefit after approval.

The NCI page we opened starts at phase 1. It does not describe the lab and animal work that comes first. Our sources show only that those steps come before human trials in this process.

A cell-line result sits before all of these steps. Many lab results do not lead to an approved drug. Nothing in this paper shows that the molecule is safe or works in people.

What this does not mean

This does not mean the AI work is fake or unimportant. Using AI to find early drug candidates faster is a real research topic. It does mean that a speed claim about finding a first molecule is not a claim about a treatment.

No one can get this molecule as a treatment. It is not approved by the FDA. It is not offered by any source we reviewed. Be wary of anyone who sells or promotes a product using this claim.

Questions to ask your care team

• Is there a clinical trial that fits my type and stage of cancer?
• This headline mentions a new drug. Is it approved, or still in testing?
• What phase is a trial I found online, and who runs it?
• How can I check a claim before I act on it?

Our pages on clinical trial phases and what clinical trials are explain more. For background on the cancer in the paper, see liver cancer.

How this article was prepared

An AI-assisted editorial system helped prepare this page. No named medical reviewer has reviewed it unless one is listed.

The National Cancer Information Foundation publishes Cancer Explained. This page is for learning. It is not medical advice and does not suggest a test or treatment.

See an error, old source, or unclear wording? Tell us.