Artificial intelligence is driving cancer research and treatment.
Developing cancer drugs takes a long time, costs a lot of money, and often fails. According to a review published in the international scientific journal Cancers in September 2026, only about 4% of cancer drug candidates that enter human trials are approved. Developing a drug can take about 14 years and cost more than $1.2 billion.
Artificial intelligence (AI) is helping change how researchers discover and develop drugs. It can help them identify targets for treatment, detect diseases, and choose suitable patients for studies. AI may not yet be able to find a cure for every type of cancer, but it is already helping make real progress.
1. Finding treatment targets that were once too difficult to reach.
Many proteins linked to cancer have complex shapes. This makes it hard for scientists to find substances that can attach to these proteins and stop them from working. AI can help predict the three-dimensional shapes of proteins and simulate how possible drug molecules interact with their targets.
New ways to predict protein structures, especially Google DeepMind's AlphaFold, have opened up more possibilities for research in molecular biology. This work also helped earn Demis Hassabis and John Jumper the 2024 Nobel Prize in Chemistry.
Biotechnology companies are investing heavily in this area. Isomorphic Labs, a UK-based company that uses AI to discover drugs, aims to bring possible new drugs into human trials. However, finding a promising compound is only the first step. The drug must still pass strict rounds of testing.
2. Helping develop drugs with a better chance of being safe.
AI can screen millions of compounds, predict their properties, and remove weak candidates at an early stage. This helps researchers spend more time and money on compounds that look more promising.
A 2024 analysis by the Boston Consulting Group in the United States found that, among about 20 AI-designed drug molecules that had completed Phase I clinical trials, around 80–90% passed this stage. This was higher than the historical average of about 40–65%.
However, in Phase II, drugs must show that they can treat the disease. About 40% of this group succeeded, a rate similar to that of other drugs. This suggests that AI can improve the search for promising drug candidates, but it has not solved the biggest challenge: proving that a drug actually works in patients.
3. Helping develop personalized cancer vaccines.
AI and computer programs that study biological data are helping researchers develop treatments suited to each patient. These include cancer vaccines that use mRNA technology.
The process starts by reading the genes in both the tumor and healthy tissue to find changes that are specific to the tumor. Software then predicts which pieces of abnormal proteins are likely to trigger an immune response. Researchers use this information to design a vaccine that matches the features of each person's tumor.
This is an important step toward personalized medicine, which means choosing treatment based on each person's needs. However, more studies involving different groups of patients are still needed to find out how well these vaccines work.
4. Improving breast cancer detection with mammograms.
Finding cancer early is one of the main ways to improve treatment results. AI can help doctors examine mammograms, which are breast X-ray images, and spot unusual signs that the human eye might miss.
The MASAI trial in Sweden included 105,934 women. It compared mammogram readings assisted by AI with the standard method, in which two radiologists read the images. Results published in The Lancet in January 2026 showed that the AI-assisted group detected 80.5% of cancer cases, compared with 73.8% in the control group. The AI-assisted method did not increase the rate of false positives, meaning results that incorrectly suggest cancer is present.
However, the benefits of AI must be judged carefully. Finding more cancer cases does not automatically mean that fewer people will die from the disease. Long-term studies are still needed to understand the real effects on public health.
5. Looking for signs of cancer in blood tests.
Another area of research is blood testing for several types of cancer at once. These tests use data analysis and machine learning to look for DNA markers linked to tumors in the blood and predict which organs may have cancer.
However, the NHS-Galleri trial in the UK showed that this technology still needs more testing. The study followed around 142,000 people through three yearly surveys. It did not achieve its main goal of reducing the rate of cancer diagnosed at a late stage. The drop in cases of stage IV cancer was not large enough to make up for the rise in cases diagnosed at stage III.
At the annual ASCO (American Society of Clinical Oncology) conference in May 2026, the organization that sponsored the study announced some encouraging results from further analysis. These included fewer stage IV cancer cases detected for some types of cancer and more cases found at an early stage. However, these results still need to be checked independently. It is not yet possible to confirm whether blood tests for several types of cancer can reduce deaths.
6. Predicting tumor gene changes from tissue samples.
Finding gene mutations often requires detailed tests that take time and use limited tissue samples. AI may make it possible to predict some genetic features of tumors from images of tissue samples that doctors already examine under a microscope.
One example is EAGLE, an AI model designed to identify EGFR mutations in people with lung cancer by examining tissue images. A study published in the international medical journal Nature Medicine in July 2025 found that this tool could reduce the need for rapid genetic testing in some cases. It could also help save tissue samples for more detailed analysis.
However, predictions made by AI cannot fully replace tests that confirm the results. The model's accuracy may also differ between population groups and types of samples. It must therefore be tested on many different groups of patients before it can be widely used.
7. Reducing the time needed to find patients for clinical trials.
Clinical trials often take a long time to recruit enough people. Doctors must check each patient's medical records, health history, and test results against the study's detailed requirements.
AI can help automate this work. TrialGPT, a tool developed by researchers at the U.S. National Institutes of Health, can assess whether a patient meets each requirement for a clinical trial.
In early studies, TrialGPT was 87% accurate, close to the 89–90% accuracy of human reviewers. It also reduced screening time by 42.6%. However, the tool was tested using combined patient profiles, so its performance in real-world settings still needs to be checked.
AI can also help researchers review trial rules that may be too strict. Trial Pathfinder, a tool developed by researchers at Stanford and Genentech, simulated lung cancer trials using records from more than 61,000 patients. The results showed that changing some rules that exclude patients could greatly increase the number of people eligible to take part, without changing the results of the simulation.
However, matching patients to trials more quickly does not necessarily mean that recruitment will finish sooner. Costs, travel, the number of trial sites, and doctors' decisions can still create major barriers.
AI cannot yet replace biological testing.
Although AI offers many promising possibilities, it cannot replace laboratory research, tests before human trials, or trials involving people. Only carefully designed studies can show whether a drug is safe, works as intended, and truly helps people live longer.
Regulators are also developing rules for assessing how AI is used in drug development. The U.S. Food and Drug Administration (FDA) issued draft guidance in January 2025. By January 2026, the FDA and the European Medicines Agency (EMA) were to publish general principles for using AI in drug development.
In Vietnam, AI could help strengthen cancer research, develop personalized medicine, analyze medical images, and improve the process of choosing patients for clinical trials. To use this technology effectively, the country needs to invest in high-quality medical data, protect privacy, test algorithms on Vietnamese people, and keep medical professionals in charge of important decisions.
The realistic goal is not for AI to find a cure for every type of cancer. Instead, it is to help reduce failures in drug development, find diseases earlier, and organize research more efficiently. AI can speed up the search for treatments, but scientific evidence and patient safety must always guide the final decisions.