Codeness: a measurable index that separates self-templating epigenetic codes from sequence echoes across the tree of life

For decades, biologists have debated whether the cell possesses a second layer of instruction beyond the DNA sequence itself. We know that DNA acts as the master blueprint, a fixed set of chemical letters that spells out the instructions for building an organism. But cells also carry a dynamic layer of chemical tags and structural changes that sit on top of that DNA, telling it which genes to turn on or off at any given moment. This layer, known as the epigenome, changes as cells divide and as organisms develop. The central question has long been whether this layer functions as a true "code"—a system of instructions that is copied faithfully from one generation of cells to the next, independent of the underlying DNA sequence, much like a written language that can be read and reproduced. For a quarter-century, this question remained a matter of qualitative argument, with scientists offering conflicting opinions based on different observations. Some argued that these chemical marks form a rigid cipher, while others insisted they were merely fleeting reflections of the cell's current activity, dissolving and reforming with every cell division.

A new study by independent researcher Davis W. Cheng transforms this decades-old debate from a philosophical argument into a measurable reality. The researcher proposes a way to quantify exactly how "code-like" different parts of the epigenome are. Instead of asking if the epigenome is a code in a vague, all-or-nothing sense, the study breaks the concept down into three specific, testable properties. First, it asks if a chemical mark is faithfully inherited when a cell divides, passing from a mother cell to its daughter. Second, it asks if that mark actually causes changes in the cell's behavior, rather than just following along. Third, it asks if the mark can be predicted simply by looking at the DNA sequence underneath it. By combining these three factors into a single score, the study creates a "codeness index" that allows scientists to rank different epigenetic layers based on how well they function as a self-sustaining memory system.

The results of this measurement reveal a clear split in how the epigenome works. The study analyzed seven different molecular layers found in mammals and found that they fall into two distinct categories. The first category consists of true self-templating codes. These are layers that carry their own instructions and copy themselves with high fidelity during cell division, largely ignoring the DNA sequence beneath them. The most prominent member of this group is DNA methylation, a chemical tag that attaches directly to the DNA letters. The study found that DNA methylation is the most code-like layer, with a high score for its ability to copy itself and influence the cell. Other layers in this group include certain repressive histone marks and a specialized protein variant called CENP-A, which helps build the centromere, the part of the chromosome that pulls cells apart during division. These layers act like a permanent record that is passed down, ensuring that a cell remembers its identity even as it divides.

In sharp contrast, the second category consists of what the study calls "sequence echoes." These are layers that are highly predictable based on the DNA sequence but do not copy themselves. Instead, they are re-derived from scratch every time a cell divides. The most sequence-predictable layer, chromatin accessibility—which determines how open or closed the DNA is to the machinery that reads genes—actually ranks as one of the least code-like layers. This finding overturns the idea that the most readable or predictable part of the epigenome is necessarily the one that carries the most memory. The study shows that just because a feature can be predicted from the DNA sequence does not mean it functions as a heritable code. In fact, the layers that are most dependent on the DNA sequence for their existence are the ones least likely to carry a stable, independent memory across generations.

To ensure these measurements were not just theoretical estimates, the researcher reanalyzed two major sets of primary data. One dataset tracked the inheritance of DNA methylation through cell division, confirming that the chemical tags are copied with a fidelity of 93.3 percent, a rate high enough to be considered a reliable memory system. The second dataset examined a genome-wide screen where scientists silenced genes and then removed the tool used to silence them, watching to see if the silence persisted. The study found that genes located in regions rich in specific DNA structures, known as CpG islands, were far more likely to remain silent after the tool was removed. This confirmed that the ability to "read" the state of the cell depends on the underlying DNA structure, specifically the presence of these islands, rather than on the repressive chemical marks themselves. This suggests that the DNA methylation substrate is the true carrier of the memory, not the histone marks that often accompany it.

The study also looked beyond mammals to see if this pattern holds true across the tree of life. By applying the same measurement logic to bacteria, fungi, plants, and invertebrates, the researcher found a consistent ranking of which layers act as codes. In bacteria, fungi, plants, and mammals, DNA methylation remains the most code-like layer. However, in nematodes, a type of roundworm that lacks DNA methylation, the role of the primary code is taken over by repressive histone marks and small RNA molecules. This suggests that the function of a heritable memory system is universal, but the specific molecule that carries it changes depending on the lineage. The study proposes a "universal epigenetic codeness law," a hypothesis that the ranking of these layers is conserved across all life, with each kingdom applying a single multiplier that adjusts the overall strength of the system. While this law is presented as a testable hypothesis rather than a proven fact, the data strongly supports the idea that the hierarchy of epigenetic memory is a fundamental feature of biology.

Ultimately, this work moves the field away from the idea of a single, rigid "histone code" that acts like a fixed lookup table for every gene. Instead, the findings support a view of the epigenome as a combinatorial language. In this language, different marks combine to create context-dependent instructions, with a core memory layer concentrated in DNA methylation. The study concludes that the question of whether an epigenetic code exists is no longer a matter of opinion but a measurable quantity. By separating the self-copying codes from the sequence-driven echoes, the research provides a clear framework for understanding how cells remember their identity and how that memory is preserved or lost across the diversity of life. The study does not claim to have solved every mystery of epigenetics, but it offers a precise tool to measure the strength of the memory systems that keep life running.