How AI Is Speeding Up Cancer Research

The American Cancer Society estimates there were 2,041,910 new cancer diagnoses and 618,120 cancer deaths in the US alone in 2025 (PMC). These numbers show why speed matters so much in cancer research. Every month saved in the research process can help doctors confirm new biomarkers sooner, detect cancer earlier, or run better clinical trials that reach answers faster.

Artificial intelligence is not a cure for cancer. It does not replace biology, clinical trials, or regulation. What AI does is help speed up parts of cancer research that usually take a long time. This is especially true for tasks that involve finding patterns in huge datasets or reviewing complex documents that people normally handle by hand.

In this article, we explain four ways AI is helping cancer research move faster. We also look at cancer rates by state and explain where progress is still slow.

What “speeding up cancer research” actually means

Cancer research involves many slow steps. Some are biological, like figuring out how cancer works or proving that one thing causes another. Others are more practical, like labeling pathology slides, reviewing medical images, or matching patients to clinical trials by reading long eligibility rules.

When researchers say AI is “speeding things up,” they usually mean improvements like:

• Faster analysis of pathology slides, medical images, and genetic data
• Quicker generation of research ideas from large public datasets
• Less time spent finding and screening patients for clinical trials
• Faster updates to medical software tools, while still meeting safety rules

In general, AI works best when tasks are repetitive, high-volume, and full of data.

1) Turning pathology slides into faster, reusable research signals

Pathology is one of the most data-rich parts of cancer care. A single digital pathology slide can be enormous, sometimes containing billions of pixels. In the past, analyzing these slides at scale took a lot of time and manual effort.

In recent years, researchers have developed large “foundation models” for digital pathology. These models are trained on huge numbers of real pathology slides and can be reused for many different tasks. One prominent example is Prov-GigaPath, described in Nature as an open-weight whole-slide foundation model that achieves state-of-the-art performance across a broad range of digital pathology tasks (Nature).

Why this speeds up research:

• Researchers can start new studies using an existing model instead of building one from scratch.
• Large numbers of slides can be analyzed automatically, which helps with biomarker discovery and validation.
• Human experts can focus on difficult cases and study design instead of repetitive labeling work.

Scientific papers now describe these foundation models as a way to speed up work in diagnosis, prognosis, and biomarker prediction (Nature).

2) Accelerating target discovery by predicting protein structures

Cancer drug development often starts with a biological target, usually a protein or pathway. To design effective drugs, scientists need to understand how proteins fold and interact.

In 2021, DeepMind’s AlphaFold work demonstrated a method that can regularly predict protein structures with very high accuracy, including hard cases where no similar structure is known (Nature). That breakthrough did not “solve cancer,” but it changed the tempo of early-stage biology by making high-quality structural hypotheses available far more quickly than traditional experimental pipelines could for many proteins.

There is also evidence that the broader structural biology ecosystem shifted after AlphaFold. A Nature analysis has examined how AlphaFold affected structural biology and found measurable changes in downstream scientific activity consistent with faster access to structure information (PMC).

Why this matters for cancer research:

• Scientists can move faster when they have early clues about protein structure.
• Lab experiments can be planned more efficiently, saving time and money.
• Protein structure data helps researchers understand cancer-related mutations.

Taken together, faster access to reliable protein structures helps researchers move from ideas to experiments more quickly, which can shorten the earliest stages of cancer drug discovery.

3) Faster translation into real tools, especially in cancer imaging

One way to see research turning into real-world tools is through regulation. The U.S. Food and Drug Administration (FDA) keeps a public list of AI-enabled medical devices approved for use in the U.S.  (U.S. Food and Drug Administration). Many of these tools are in medical imaging, which plays a big role in cancer screening, detection, and monitoring.

This list does not mean every tool improves outcomes. It does show that AI tools are increasingly being built into regulated medical products.

The FDA is also creating rules for how AI tools can change over time. In August 2025, the agency released guidance on Predetermined Change Control Plans (PCCPs). These plans explain how AI tools can be updated safely after approval (U.S. Food and Drug Administration).

Why this matters for speed:

• AI tools can standardize measurements and reduce manual image review.
• Clear rules make it easier to improve tools while maintaining safety.
• Real-world use creates data that can improve future studies, if privacy rules are followed.

As more AI tools make it through regulation and into clinical use, the path from research to real-world impact becomes shorter and more repeatable.

4) Speeding up clinical trial matching, one of the most persistent bottlenecks

Clinical trials are critical for improving cancer care, but enrolling patients is often slow. Matching patients to trials requires reading long and complex eligibility rules and comparing them to medical records.

In 2024, researchers reported TrialGPT, a large language model approach to patient-to-trial matching. In a user study, TrialGPT reduced screening time by 42.6% (PMC). NIH also summarized this work publicly, emphasizing that the tool could identify relevant trials and provide a plain-language summary of how a person meets criteria (National Institutes of Health (NIH)).

Why this matters:

• Trials can reach enrollment goals faster.
• Patients may learn about suitable trials sooner.
• Research teams can spend more time on patient care and oversight instead of paperwork.

By reducing the time and effort needed to match patients to trials, AI can help studies move faster while giving patients quicker access to potential new treatments.

Cancer Burden by State

Sources:

• Cancer cases & deaths: American Cancer Society, Cancer Facts & Figures 2025 (state estimates). (Cancer.org)
• Population: U.S. Census / FRED 2024 state population estimates. (FRED)

What AI still cannot do, and why validation stays slow

Any serious discussion of AI in cancer needs to include limits. Even when AI looks promising in early studies, real-world use is slowed by:

• Data quality and repesentation: Models trained on limited populations can fail when deployed elsewhere.
• Clinical validation requirements: Oncology is high-stakes, and prospective validation takes time.
• Regulatory oversight: Even with PCCPs, iterative updates require predefined plans, monitoring, and evidence (U.S. Food and Drug Administration).
• Human factors: Many systems perform best as part of a human-AI team, and workflows need to be designed accordingly. The FDA and partners have emphasized principles of good machine learning practice and transparency to support safe use (U.S. Food and Drug Administration).

Why this matters for patients, not just researchers

Cancer remains a leading cause of death. SEER reports an age-adjusted cancer incidence rate of 445.8 per 100,000 people and a death rate of 145.4 per 100,000 people in recent years (SEER).

AI’s real value is not replacing biological breakthroughs. It is helping researchers move faster from questions to solid evidence:

• Pathology foundation models can reduce time spent building and adapting models for new tasks (Nature).
• Protein structure prediction can accelerate upstream discovery and prioritization (Nature).
• AI-enabled medical devices show a growing pipeline of translation into regulated tools, especially in imaging (U.S. Food and Drug Administration).
• Trial matching systems like TrialGPT can reduce screening time and help connect patients to appropriate studies (PMC).

For patients, these gains in speed can translate into earlier diagnoses, better-matched treatments, and faster access to clinical trials, even if progress still happens in careful, measured steps.

AI does not change the need for strong evidence, clinical judgment, or oversight, but it can reduce delays that have long slowed cancer research. Over time, shaving months or even weeks off discovery, validation, and trial enrollment adds up, increasing the chances that new findings reach patients sooner, when timing can matter most.