AI Hunts Down Cancer 'Shape Shifting' Cells

In the fight against cancer, the primary tumor is rarely the deadliest threat. Rather, it is metastasis — the spreading of cancer cells. This lethal process begins with "escapers," or Circulating Tumor Cells (CTCs), that break away from the original tumor.

While capturing these "escapers" has long been a technical nightmare, a new frontier has been breached. Professor Learn-Han Lee of the University of Nottingham Ningbo, China, a scientist and director of the Asia Pacific Microbiome and Biomedical Research Network (AMBeR), is leading this charge. His team's recent work, which systematically reviews strategies using single-cell ribonucleic acid (RNA) sequencing (scRNA-seq), offers a blueprint for intercepting these cells before they spread.

By combining molecular biology with AI, Lee and his team are turning the tide in the battle against metastasis.

The shape-shifting disguise

The first challenge in catching a CTC is recognizing it. These cells are masters of deception, often hiding in plain sight among billions of normal blood cells. When asked how these cells manage to evade detection, Lee described them as biological "shape-shifters." He acknowledged that the cells employ a dual strategy of disguise and dormancy.

"The CTCs use a process called Epithelial-Mesenchymal Transition (EMT). Essentially, they shed their 'normal' cellular ID tags, like the EpCAM marker, which many of our standard detection tools look for," he said.

However, their survival strategy goes beyond molecular camouflage. Lee noted that these cells are socially complex. Instead of traveling alone, they form clusters or even hide within the body's own immune cells, such as neutrophils and platelets, using them as shields. Furthermore, they possess the patience to wait. Lee explained that many CTCs will not immediately metastasize. Instead, they can remain dormant for long periods, quietly surviving until environmental signals, often from the tumor microenvironment, trigger them to "wake up" and seed metastases.

AI as the molecular detective

To analyze the massive datasets generated by single-cell sequencing, Lee's team has turned to AI. In this context, AI is not merely a calculator but also a detective. Lee asserted that, "AI has absolutely become a detective, and a very sharp one." While scRNA-seq generates massive, complex datasets that are too large for manual analysis, AI, specifically machine learning, goes beyond just crunching numbers. From his perspective, researchers have successfully used AI to identify CTCs among millions of blood cells. Furthermore, using an algorithm called CTC-Tracer, they can even predict which specific lesion or organ a CTC originated from. Lee explained that while a CT scan typically reveals a tumor only after it has grown, AI possesses the capability to analyze the transcriptomic profile of a single CTC to indicate that the cell carries a high risk of metastasizing to the bone or brain, thereby providing a molecular warning system.

This precision is vital because not all cancer cells are created equal. Lee emphasized that treating them as a monolith is a fatal error in precision medicine. He likened treating all CTCs as one group to assuming every person in a crowd has the same intention. If doctors treat all cancer cells with the same therapy, they might kill the majority but leave the dangerous, therapy-resistant subtypes behind to proliferate.

Lee pointed out that his recent article provides concrete examples, such as in non-small cell lung cancer, where one discovered CTC cluster was highly proliferative while another was mesenchymal (cells that develop into connective tissue, blood vessels, and lymphatic tissue) with immune evasion properties. He noted that if researchers only examine the average, they would miss that immune evasion signature entirely.

By analyzing individual cells, researchers can spot critical differences that would be missed in bulk samples. Lee highlighted that they would also miss hybrid cells — fusions of tumor and immune cells — which are implicated in metastasis and resistance.

Bridging the gap

Beyond the lab, Lee focused on the practical application of these technologies. He identified the biggest barrier to global health not as the cost of technology, but the lack of standardized workflows. He noted that while one could buy a sequencer, without a validated, step-by-step guide to handle rare CTCs from blood draw to data interpretation, the technology sits idle.

Lee observed that while the cost was decreasing and open-source bioinformatics tools were becoming available, the actual bottleneck lay in training and infrastructure to ensure consistency. He argued for the necessity of moving away from "bespoke" science in wealthy nations to create robust, simplified kits that could function effectively in labs with fewer resources.

To solve this, Lee advocated for international collaboration to create robust, simplified kits that can be used in labs with fewer resources. He stressed the need to tackle the "last mile" problem: ensuring that the data generated leads to an actionable clinical decision. Without that, he asserted, even the best scRNA-seq technology cannot fulfill its role.

University of Nottingham Ningbo also contributed to this article.

Source: Science and Technology Daily