More Recent Comments

Sunday, July 08, 2018

Nature falls (again) for gene hype

Nature is arguably the most prestigious science journal. Articles published in Nature are widely perceived to be correct, unbiased, and factual. This perception is certainly true of articles that appear in the News section of the journal since these article are presumably written by expert science writers who have evaluated the new study and decided that it's worth reporting.

Sandwalk readers know that this perception is false (fake news). It turns out that science writers who publish in Nature are not very much better than science writers in general and that's not good.

I recently published a post about an extraordinary claim concerning the number of human genes [Disappearing genes: a paper is refuted before it is even published ]. It concerns a paper posted on an archive site claiming to have found 4,998 new genes of which 1,178 are new protein-coding genes (Pertea et. al., 2018). About five weeks later another paper was posted that effectively refuted the claim of new protein-coding genes (Jungreis et al., 2018). In between publication of those two papers, a freelance science writer, Cassandra Willyard, wrote an article for Nature News that covered the original claim of 4,998 new genes [New human gene tally reignites debate].

Let's see how she handled the controversy.

She starts off badly by bringing up the old myth about gene predictions made by those who were working on sequencing the human genome.1 These non-experts were making guesses that ranged from 26,000 to 312,000. The myth is that experts were expecting about 100,000 genes and they were "shocked" to discover only 30,000 genes in the human genome [Facts and Myths Concerning the Historical Estimates of the Number of Genes in the Human Genome]. The reality is that the real experts were expecting about 30,000 genes based on work done in the '60s and '70s and their predictions were correct. No surprises as far as the experts were concerned.

Cassandra Willyard compounds her misunderstanding of the real history by publishing an extremely misleading (i.e. false) diagram that was published in 2010 (Pertea and Salzberg, 2010). Note that these two authors are the first and last authors on the recent bioRxiv paper (Pertea et al., 2018). The paper published in 2010 appeared long after the time when the original myth had been refuted but that didn't seem to bother those who reviewed the paper [see: False History and the Number of Genes 2010].2

I suppose we can't fault the Nature science writer for believing what the authors of the paper say in their publication. Can we?

Part of the problem in writing about this topic is the definition of a gene [Philosophers talking about genes]. This is 2018. We know there are two types of genes—protein-coding genes and noncoding genes. Experts have known about these two kinds of genes for more than half-a-century but the general public has been misled into thinking that all genes encode proteins. Science writers are also confused. Here's how the Nature article addresses the issue ...
Further confounding counting efforts is the imprecise and changing definition of a gene. Biologists used to see genes as sequences that code for proteins, but then it became clear that some non-coding RNA molecules have important roles in cells. Judging which are important — and should be deemed genes — is controversial, and could explain some of the discrepancies between Salzberg’s count and others.
In my opinion, we are long past the time when it's helpful to say that biologists once thought all genes encoded proteins. That hasn't been true of knowledgeable scientists for a very, very long time (>50 years). Perhaps it would be better to say that most people aren't aware of noncoding genes in spite of the fact they have been known since the 1960s. That's a better perspective.

As expected, the main emphasis in the Nature article is on the discovery of almost 5,000 new genes and the fact that scientists still don't know how many genes there are in the human genome. It quotes one scientist—a genome researcher and a former head of RefSeq—who says that the new gene count is probably due to the extra data that Pertea et al. analyzed. This lends considerable credibility to their discovery. On the other hand, the article quotes Adam Frankish of GENCODE who notes that they have already concluded that many of the predicted protein-coding genes are probably not genes.

Adam Frankish is one of the authors on the second paper posted on the archive site (Jungreis et al., 2018). It refutes the extraordinary claim of Pertea et al. that they discovered 1,178 new protein-coding genes. The second paper was posted only five weeks after the paper being promoted in the Nature News article but before Cassandra Willyard could have seen it.

In light of the challenge to the original paper as promoted in Nature, I wonder if the journal is going to publish a followup piece on the number of genes? The journal has a new editor, Magdalena Skipper, who apparently is committed to transparency and accuracy [A welcome from the new Nature editor]. This will be an interesting challenge since you might recall that as Senior Editor she played a prominent role in the ENCODE/Nature propaganda campaign back in 2012. Here she is in a video proudly proclaiming that ENCODE has discovered a function for 80% of the genome and debunking the idea that most of our genome is junk.

1. Mostly technical experts whose focus was sequencing technology and bioinformatics.

2. In their most recent paper, Pertea et al. start off by referencing their 2010 paper on the number of genes. The emphasis is on the high gene number estimates of the 1990s and the rapid decline in estimates as the sequencing project was completed. It's interesting that Steven Salzberg commented on my 2010 post by saying that he was well aware of the earlier low estimates from the 1960s and '70s but they decided not to mention it in their paper because of "space constraints" and because they didn't believe they were reliable. As it happens, the earlier evidence-based estimates by the experts turned out to be much more reliable than "educated" guesses by non-experts. Nevertheless, the early reliable estimates still aren't mentioned in their recent 2018 paper in spite of the fact that the authors were alerted to the problem eight years ago. I don't understand what they have to gain by promoting a false history of gene number estimates.

Jungreis, I., Tress, M.L., Mudge, J., Sisu, C., Hunt, T., Johnson, R., Uszczynska-Ratajczak, B., Lagarde, J., Wright, J., Muir, P., Gerstein, M., Guigo, R., Kellis, M., Frankish, A., and Flicek, P. (2018) Nearly all new protein-coding predictions in the CHESS database are not protein-coding. bioRxiv(July 2, 2018). [doi: 10.1101/360602]

Pertea, M., and Salzberg, S. (2010) Between a chicken and a grape: estimating the number of human genes. Genome biology, 11:206. [doi:10.1186/gb-2010-11-5-206]

Pertea, M., Shumate, A., Pertea, G., Varabyou, A., Chang, Y.-C., Madugundu, A.K., Pandey, A., and Salzberg, S. (2018) Thousands of large-scale RNA sequencing experiments yield a comprehensive new human gene list and reveal extensive transcriptional noise. bioRxiv(May 29, 2018). [doi: 10.1101/332825]


  1. The problem with assuming that the 1960's "genetic load" argument really was an accurate predictor of the number of human genes is that there doesn't seem to be much evidence that it was a generally applicable argument. Did it accurately predict the number of genes for other organisms? Somebody's guess for the number of jellybeans in a jar is bound to be closer to the real number, but it doesn't follow that the person has a generally applicable method for guessing the number.

    1. The genetic load argument is not really about precise gene number though, it is about how much of the genome is under constraint. After all, it was advanced before even people knew about things like splicing.

      It does give a good order-of-magnitude estimate though, and given that this is something that many did in fact get wrong in a major way based on other criteria, it is a remarkable success for me.

    2. The point I’m making is that there were many knowledgeable scientists who were predicting about 30,000 genes back in the 60s and 70s. They based their predictions on available evidence.

      The people who were making guesstimates of 100,000 genes in the 1990s were unaware of the genetic load argument and of the predictions based on the number of mRNAs from the earlier era. The 1990s guesses did not represent the views of the experts and they shouldn’t be given as much credit as they are getting in recounting the history of gene estimates.

      I was not the least bit shocked to learn we had 30,000 genes when the draft sequence was published in 2001. Were you?