More Recent Comments

Tuesday, February 16, 2021

The 20th anniversary of the human genome sequence:
6. Nature doubles down on ENCODE results

Nature has now published a series of articles celebrating the 20th anniversary of the publication of the draft sequences of the human genome [Genome revolution]. Two of the articles are about free access to information and, unlike a similar article in Science, the Nature editors aren't shy about mentioning an important event from 2001; namely, the fact that Science wasn't committed to open access.

By publishing the Human Genome Project’s first paper, we worked with a publicly funded initiative that was committed to data sharing. But the journal acknowledged there would be challenges to maintaining the free, open flow of information, and that the research community might need to make compromises to these principles, for example when the data came from private companies. Indeed, in 2001, colleagues at Science negotiated publishing the draft genome generated by Celera Corporation in Rockville, Maryland. The research paper was immediately free to access, but there were some restrictions on access to the full data.

I'm not sure that access to data is the most important issue but it's interesting that both Science and Nature chose to emphasize it. What's even more interesting is that both journals chose to ignore the very real controversies surroundng the function of the human genome.

From my perspective, the most important article in Nature is titled "A wealth of discovery built on the Human Genome Project — by the numbers" and it makes a big deal about the fact that there are many more publications on certain genes following the publication of the draft sequence (duh!). That's where the figure on the cover comes from. But here's the part about the real science behind the human genome sequence.

Not junk

A great debate pre-dated the start of the HGP: was it worth mapping the vast non-coding regions of genome that were called junk DNA, or the dark matter of the genome? Thanks in large part to the HGP, it is now appreciated that the majority of functional sequences in the human genome do not encode proteins. Rather, elements such as long non-coding RNAs, promoters, enhancers and countless gene-regulatory motifs work together to bring the genome to life. Variation in these regions does not alter proteins, but it can perturb the networks governing protein expression.

This figure (right) was included in the article. It is misleading because it implies that all non-coding DNA was thought to be junk and that's untrue. Lots and lots of noncoding elements (including regulatory DNA) were known before the draft sequences were published. It's also misleading because it implies that there are 130,629 noncoding genes in the current version of the human genome when, in fact, there are only a few thousand—most of those 130 thousand transcripts are almost certainly junk RNA.

An accompaning article on milestones in genomic sequencing covers the ENCODE publicity campaign in 2012 [Our most elemental encyclopaedia]. As most Sandwalk readers know, this publicity campaign was full of misrepresentations of the ENCODE data and emphasized the false narrative that junk DNA was dead. In spite of significant blowback from very knowledgeable scientists, Nature never admitted that it had made a mistake, althought the ENCODE researchers did publish a partial retraction.

Up until today, I assumed that Nature editors were aware of their mistakes in 2012 but were just too embarrassed to admit it. It looks like I was wrong because here's how they describe the ENCODE "milestone."

The second version of the encyclopaedia, now applying next-generation sequencing technologies to the entire genome, defined not only a set of 20,687 protein-coding genes but also how their expression is controlled in 147 different cell types. About 80.4% of the genome was associated with at least one biochemical event (that is, RNA- or chromatin-related), and 95% was within reach of a DNA–protein interaction. Technologies to probe long-range physical interactions between distinct chromosomes revealed a plethora of promoter–enhancer interactions that are critical for gene activation.

ENCODE 2 was a landmark for the understanding of molecular genetics and a major feat in data standardization, analysis and integration. Predictive models of gene expression were developed based on epigenetic marks or transcription factor binding patterns. Machine learning methods were trained to cluster functional regions across the genome that were afterwards associated with concrete biological processes, such as immune response or neural activity. There was enough statistical power to quantify the impact of negative selection on genomic regulation, and the integration with large databases of annotated variants enabled the exploration of individual functional alterations, with implications for diseases such as cancer.

That's extemely misleading because we now know that most of those 80.4% of biochemical events are spurious and have nothing to do with function. Nature includes this cartoon video, narrated by Tim Minchin, where he says, "[ENCODE is] a map that reveals the whole genome is a jungle buzzing with activity, even the parts we used to think of as junk."

I believe that Nature made a serious mistake back in 2012 when they promoted the idea that most of our genome is functional—an idea that was contrary to the evidence available at the time. Since then, that claim has been effectively refuted and the evidence that most of our genome is junk has been accepted by most knowledgeable scientists. Nature is being disingenuous (or worse) by continuing to promote this false idea in 2021. It's about time that the journal is held accoutable for their misinformation campaign of 2012 and their contunuing refusal to accept real science.


  1. I tried to figure out what exactly they mean by "130,629 non-coding elements" and digging into the supplemental data, it all looks like random numbers to me. Such a lack of rigor is sad.

    1. I sent an email message to all four authors asking them to explain what they mean by "not junk" and what they mean by 130,629 genomic elements.

      Note that one of the authors (Kellis) was the lead author on the 2014 "retraction" paper and he argued at that time that the original ENCODE paper was misunderstood.

      "Kellis says that ENCODE isn't backing away from anything. The 80% claim, he says, was misunderstood and misreported. Roughly that proportion of the genome might be biochemically active, he explains, but some of that activity is undoubtedly meaningless, leaving unanswered the question of how much of it is really 'functional'."

  2. I think you may be missing some of those "elements" when you equate them with transcribed RNAs. That leaves out a large class of non-transcribed regulatory sequences (bonus! they actually do exist and are functional!), i.e. non-spurious transcription factor binding sites. Of course they are all very short and so comprise a negligible fraction of the genome. But they do add up if you're just counting "elements".

    1. Yes, but they claim in their supplements that they have identified >800k enhancers and >58k promoters. Apparently, (according to their supplemental data) not a single enhancer had been discovered before 2001.

      After this, I thought maybe they exclude enhancers but do count ncRNAs? But they never really list how many lncRNAs they think there are. (Apparently, they list 9.6k lncRNAs that have SNPs that associate with disease.)

      These non-nonsensical statements do not inspire much confidence.

      You can go see for yourself:

    2. Oh, I have no clue where the figure of 130,629 comes from. I was just pointing out that RNA genes, spurious or not, are not the only possible "elements".

  3. Hello Larry, thanks for the neat post. Nice call.

    I was wondering if you have also seen this?

    "The water paradox and the origins of life"

    It seems to argue in favour of a land-based "warm little pond" as the likely origin of life since water has a tendency to break the bonds in nucleotide- and amino acid polymers. So life must have started where water was limited. There is a need for "wet-dry cycles". Thus life did nnot in the ocean, but on land were water was only present intermittently.

    I don't particularly buy such "soup" hypotheses for the origin of life. I lean more towards of deep sea alkaline hydrothermal vents, as this seems way more thermodynamically plausible. From reading some of your old posts, I know you have a similar position. Do you have anything to say about this article? I think porous hydrothermal vents could potentially allow for the same conditions that allow polymers to persist and not breakdown, so water doesn't need to be absent (intermittently) all together. But I am not sure. What are your thoughts on this?

    1. For a different perspective there's this:
      The ambivalent role of water at the origins of life