An Outcome-Independent Analytical Framework Reveals Systemic Discordance Between a Continuous Neural-Mesenchymal Programme and Consensus Molecular Subtyping in Colorectal Cancer: A Multi-Cohort Transcriptomic Study
This study introduces an outcome-independent, four-gene neural-mesenchymal (NEM) programme score derived solely from co-expression patterns that largely overlaps with but is distinct from the discrete CMS4 subtype in colorectal cancer, revealing a continuous biological programme that captures mesenchymal activity missed by current categorical classification systems.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Colorectal cancer is not a single disease but a collection of distinct biological behaviors, each leaving a different signature in the genetic material of tumor cells. For years, scientists have tried to sort these tumors into neat categories to better understand how they grow and how they might respond to treatment. One widely used system, known as Consensus Molecular Subtyping, divides these cancers into four groups. One of these groups, often called the "mesenchymal" type, is particularly aggressive. It is characterized by cells that have lost their rigid structure and gained the ability to move and invade other tissues, a process known as the epithelial-mesenchymal transition. This specific group is also known to be packed with supportive tissue from the body's own connective structures, which often makes it harder to treat. However, biology rarely fits perfectly into boxes. The question researchers have long faced is whether these rigid categories capture the full story of how a tumor behaves, or if there is a continuous spectrum of activity that the categories miss.
A new study by Anirudhha Roy at JIS University in India explores this gap by looking at colorectal cancer through a different lens. Instead of starting with the established categories or the patient's survival data, the researcher began with a simple observation of how genes talk to one another. He selected a specific list of fourteen genes known to be involved in nerve development and the ability of cells to change shape and move. By examining how these genes rose and fell together in a large set of tumor samples, he identified a tight-knit group of four genes that consistently moved in sync. This group included genes associated with nerve growth and genes known to drive the ability of cells to become invasive. The researcher then created a single score based on the activity of these four genes, treating it as a continuous measure of a "neural-mesenchymal" program—a biological state where a tumor is acting like a mix of nerve tissue and invasive, moving cells.
The study then tested this new score against the established four-category system across nine different groups of patients from around the world, totaling nearly 3,750 individuals. The results revealed a striking pattern of overlap and difference. The new score was excellent at finding the aggressive, mesenchymal tumors; it correctly identified nearly all of them. However, the reverse was not true. While the new score found almost all the tumors the old system labeled as aggressive, it also found a significant number of tumors that the old system had placed in the other, less aggressive categories. In the initial group of patients, about one-third of the tumors flagged as having this high neural-mesenchymal activity were not classified as the aggressive type by the standard system. This pattern held true across all the different groups of patients studied, suggesting that a substantial portion of tumors possess this invasive, nerve-linked activity even though they do not carry the official label for it.
The researchers also looked at whether this new score could predict who would survive longer. The data showed a slight trend where patients with high scores had a higher risk of death, but this link was not strong enough to be considered a definitive proof of prognosis. The study was careful to note that this score was built without looking at survival data, so the connection to patient outcomes remains an observation rather than a guaranteed rule. Furthermore, when the team checked the physical presence of the proteins made by these four genes in actual tissue samples, they found something unexpected. Two of the genes were clearly present in the tumor cells, but the gene associated with nerve growth was not found in the tumor cells themselves. Instead, it appeared to be located in the nerves surrounding the tumor. This suggests that the score might be capturing a conversation between the tumor and the nerves nearby, rather than just a change inside the tumor cells alone.
The core discovery of this work is that the continuous activity of this neural-mesenchymal program and the categorical labels used to sort tumors describe two different, though overlapping, aspects of the disease. The standard system successfully identifies the most aggressive tumors, but it leaves out a significant minority that still carry the biological hallmarks of invasion and nerve interaction. The study does not claim that the old system is wrong, but rather that it is incomplete. By using a method that looks at how genes coordinate with each other rather than how well they fit a template, the researcher found a way to see a biological reality that the standard labels miss. This finding implies that if doctors or scientists rely only on the standard categories to study the most dangerous forms of colorectal cancer, they may be systematically ignoring a large group of patients who share the same dangerous biological features but are classified differently. The work suggests that understanding cancer may require looking at the continuous flow of biological activity, not just the boxes we draw around it.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.