After the ENCODE project found that some 80 percent of the genome is biochemically active and transcribed into RNA, many junk DNA defenders were not ready to give up their paradigm. They responded by claiming much of that transcription is random “noise” — little better than “junk RNA.” We’ve responded to this goalpost-shifting argument multiple times in the past — for example here, here, here, and here. Last year, Daniel Witt noted that one prominent skeptic of junk RNA is the Nobel Prize-winning biochemist Thomas Cech, who wrote:
[M]any scientists do not share my enthusiasm for these RNAs. They think that RNA polymerase, the enzyme that synthesizes RNA from DNA, makes mistakes and sometimes copies junk DNA into junk RNA. A more scholarly description of such RNAs might explain them away as “transcriptional noise” — the idea being, again, that RNA polymerase isn’t perfect. It sometimes sits down on the wrong piece of DNA and copies it into RNA, and that RNA may have no function. I readily admit that some of the lncRNAs may in fact be noise, bereft of function, signifying nothing.
However, I’ll point out that there was a time in the not-too-distant past when telomerase RNA and microRNAs and catalytic RNAs weren’t understood. They hadn’t been assigned any function. They, too, could have been dismissed as “noise” or “junk.” But now hundreds of research scientists go to annual conferences to talk about these RNAs, and biotech companies are trying to use them to develop the next generation of pharmaceuticals. Certainly one lesson we’ve learned from the story of RNA is never to underestimate its power.
“Pervasive Transcription”
Now, a new scientific paper in Genome Biology and Evolution, “Pervasive Transcription in the Human Genome Exceeds Background Noise,” provides strong evidence that in a large proportion of cases DNA transcription is not random “noise,” and the RNA produced is not “junk.” A summary article in the journal explains what the debate is all about:
A major debate in genome biology centers on whether the vast, non-protein-coding majority of the human genome consists of nonfunctional “junk” or functional “dark” DNA. This contention is fueled, in part, by the observation of pervasive transcription: while only roughly 2% of the genome encodes proteins, projects like ENCODE reveal that over 75% of it is transcribed into RNA. One view posits that these transcripts represent undiscovered functional noncoding elements, while others argue that this widespread transcription is merely a consequence of background noise—the chance production of many nonfunctional transcripts by a leaky transcriptional machinery.
This follows a previous 2024 paper in Nature which used an innovative experimental technique to test for whether transcription is randomly producing “junk RNA,” or whether it is functional and controlled. The 2024 paper took sequences of non-coding DNA in yeast and mice and reversed them. If transcription is indeed random and meaningless, then RNA should still be produced at the same level as it is by the “native” genomic sequence. But if transcription is meaningful, controlled by promoter sequences, and the RNA sequences are actually doing something useful, then reversing the sequence should scramble any transcription-regulation mechanisms and lead to far less transcription than normal. Their results were mixed, but interesting. In yeast, transcription remained high. But in mice, the reversed sequence was transcriptionally silent. This suggests that in a mammalian genome like the mouse, the RNA was not randomly produced, and not junk.
Transcription in the Native Genome
The 2026 paper reports on an AI trained on genomic data, including data from the 2024 paper, enabling it to predict when transcription would be initiated. The program, called Puffin-D, found that most of the genomic sequences analyzed would not be transcribed after having their sequences shuffled around — again, suggesting the transcription in the native genome is controlled and not random:
Upon applying Puffin-D to these altered sequences, researchers noted that while some background transcription exists, predicted initiation levels of transcription remained sparse. In fact, these predicted transcription initiation levels were at least four times lower than the levels detected in the native human genome. Such a significant difference suggests that most transcription in humans is not simply the result of stochastic background noise.
The technical paper explains these striking findings:
We applied the model to randomized (reversed or shuffled) versions of the human genome and found that transcription is predicted to be sparse across all randomization methods, initiating with at least four-fold lower frequencies than in the native human genome. This relatively low level of background transcription from the human genome suggests that most transcription is not a consequence of background noise, thus it requires other explanations.
A summary there explains the important implications of the data:
Our work provides an important step in resolving the long-standing debate over what proportion of the human genome is functional. A key aspect of this debate is whether pervasive transcription provides evidence for the existence of a large suite of uncharacterized functional elements or is the result of background noise. Our major finding is that while there is appreciable background transcription in the human genome, it is much lower than total transcription. We found this by employing machine learning to predict transcription initiation in randomized versions of the human genome as a way of estimating background transcription. We also found the excess transcription is not the result of transposon transcription. Our results indicate that much of the human genome pervasive transcription cannot be attributed to noise. This in turn suggests the high levels of transcription are either beneficial and thus attributable to selection, or result from mutation bias resulting in the creation of transcription initiation sites.
And again, the technical paper offers this striking conclusion: “These results suggest that the majority of observed transcription in the native human genome is not readily attributable to background noise, so other explanations for this excess of native genome transcription are required…. The conclusion that much of the transcription observed in the human genome cannot be explained by background noise requires alternative explanations for the excess transcription.” (emphasis added)
Functional RNA
Maybe the “alternative explanations” should include that obvious possibility that the RNA produced is probably highly functional, and thus its transcription is regulated and controlled. In fact, the paper explores this explanation:
Another possible explanation for why transcription levels are above background in the human genome is that transcription has been selected for because it provides benefits to the host. We think there are three main, non-mutually exclusive possibilities for what selection is acting on. The first is selection for a large suite of uncharacterized noncoding RNAs whose transcripts are functional, providing benefits to the host. The growing list of functions ascribed to noncoding RNAs supports this hypothesis. However, this explanation is hard to reconcile with observations that most of these elements do not show DNA sequence conservation and that some noncoding RNAs do not appear to encode functional transcripts.
But of course their argument that DNA elements can only be functional if they show “sequence conservation” is an entirely evolutionary view, which ignores the possibility that functional sequences can designed to be different in different organisms. We discussed this major blind spot in evolutionary thinking in a Long Story Short video last year.
In the end, the new 2026 paper explores various explanations for why transcription is so high — some of which involve function for the RNA and some of which would push it back into the realm of junk. Much of the paper’s discussion is speculative, but the bottom line is the data shows not only that the vast majority of the genome is transcribed, but that the transcription is not random. Taken at face value, this evidence directly challenges both the junk DNA and junk RNA paradigms, and is highly consistent with intelligent design’s prediction of mass functionality in the human genome.