Tuesday, June 15, 2010

False History and the Number of Genes

Mihaela Pertea and Steven L Salzberg have just published a paper in Genome Biology with an interesting title: Between a chicken and a grape: estimating the number of human genes. Part of their paper covers the history of gene number estimates and it includes the figure shown here.
Figure 2. The trend of human gene number counts together with human genome-related milestones. Individual estimates of the human gene count are shown as blue diamonds. The range of estimates at different times is shown by the two vertical blue dotted lines. Note how this range has narrowed in recent years.
This is really annoying because it perpetuates a myth that needs to be debunked. I've addressed it in an earlier posting [Facts and Myths Concerning the Historical Estimates of the Number of Genes in the Human Genome].

Mihaela Pertea and Steven L Salzberg have completely ignored a substantial literature on the subject. First, there's the genetic load arguments of King and Jukes from 1969. They estimated that there had to be fewer that 40,000 genes in our genome. Ohno summarized the estimates in a 1972 paper and came up with an estimate based on current knowledge of 30,000 genes (Ohno 1972).

Then there's the substantive literature on expressed sequences from the 1970's These were mostly hybridization experiments showing that human tissues had a core of about 10,000 genes expressed in the most complex tissues. The estimate was that there were probably no more than double that number of genes in total. Benjamin Lewin was the expert on this subject and his early books (especially Gene Expression II) covered all the bases. By 1983, Lewin was able to conclude in Genes II ...
Given some uncertainties about estimating the numbers of genes present in multiple copies, we might say that the mammalian genome looks to be of the order of 30,000 - 40,000 gene functions.
He published the same estimate in Genes IV in 1990.

Lewin was not alone. Most textbooks contained similar estimates in the 1980s. In Molecular Biology of the Cell by Alberts et al. (1983) the estimate was also 30,000 genes (p. 406). These estimates were not dismissed as unreliable. Quite the contrary. In my circles, the general impression was that humans had to have fewer than 50,000 genes and the number was likely to be less than 30,000.

It's true that Walter Gilbert had "guesstimated" 100,000 genes and it's true that the early estimates from the Human Genome Project used a number like this (based on a false assumption). But that doesn't mean that everyone agreed. Indeed, among those who had really studied the problem, a much lower number was preferred.

During the 1990s, the preliminary results from EST cloning and sequencing started to come in and it looked like there were at least 100,000 genes based on this data. However, this was a controversial estimate precisely because so many people knew that it conflicted with a lot of data. Sure, there were those who believed the EST data over everything else but they did not represent everyone who was interested in gene numbers. It is very misleading to suggest that there was a consensus in favor of more than 50,000 genes as the figure implies.

That's false history and it does a great disservice to those who turned out to be correct.


[HatTip: Carl Zimmer]

King,J.L. and Jukes,T.H. (1969) Non-Darwinian evolution. Science 164; 788-798.

Ohno, S. (1972) So much "junk" in our genome. Brookhaven Symp. Biol. 23:366-310.

16 comments:

  1. Hi Larry,

    This is great stuff! But it seems to me this has gotten to the point where only a very public debunking in someplace like Nature will get scientists to start correcting themselves on the history -- most scientists don't read the right blogs! What do you think?

    ReplyDelete
  2. I notice that all of the really high numbers in their chart are associated with the human genome project. It seems to me there is an interesting historical point to be made there, about the use of hype to get attention and grants paid to a big science project.

    ReplyDelete
  3. hi Larry,
    Actually we were well aware of the Lewin estimate, and we had included it in an early draft. But for space constraints we decided to focus on sequence-based estimates, rather than those based on other methods. I realize that this is a bit genome-centric, but it wasn't as misleading as you seem to imply. I thought Lewin's estimate and his reasoning were somewhat tenuous and Ohno's even more so (given the state of knowledge at the time), and he could have turned out to be wildly off too.

    Our main point is supposed to be different: we wanted to draw attention to the fact that the number of genes is still unknown, and that the uncertainty surrounding the number is still surprisingly high, at least 1000 genes plus or minus.

    ReplyDelete
  4. @PZ,

    In this case I don't think it was hype.

    The silly estimates were made by people who didn't understand—or didn't bother to read—the scientific literature. They didn't understand the genetic load arguments and they didn't understand the hybridization studies done in the 70s.

    They didn't understand the results pouring out of developmental biology labs. If they had assimilated that data they would know that large phenotypic changes can be due to small changes in regulation. Thus, there's no need to postulate more genes for humans than for Drosophila.

    Of course, they didn't understand evolution either.

    So, the errors in estimating the number of genes can best be attributed to ignorance and not deliberate hype. That ignorance was reinforced by a false notion of the supremacy of humans on the "ladder of life."

    Many people were disturbed when the actual number of genes came in closer to the predictions of the experts than the amateurs. They felt that something must be wrong 'cause humans have to be a lot more special than other animals.

    This is what I call The Deflated Ego Problem and it's still with us today. It's what gives rise to all that nonsense about abundant regulatory RNAs, alternative splicing, etc. etc. These are all attempts to salvage something from the embarrassing situation and prove that humans really are different after all.

    ReplyDelete
  5. @Steven Salzberg,

    I'm sorry but that excuse (lack of space) doesn't cut it. You choose to focus on estimates of 100,000 genes that clearly turned out to be very wrong and you ignored the estimates of those scientists who were working directly in the field of estimating gene numbers.

    You choose to dismiss Lewin's estimate, and those of other experts, as "somewhat tenuous" but included some ridiculous estimates that, even by your own admission, were incorrect—and known to be incorrect at the time.

    That's not very scientific when you are supposed to be writing an accurate review of the history of a subject.

    BTW, how much extra space would it have taken to include a few extra points on your graph?

    ReplyDelete
  6. we wanted to draw attention to the fact that the number of genes is still unknown, and that the uncertainty surrounding the number is still surprisingly high, at least 1000 genes plus or minus.

    Other than illustrating difficulties with predicting genes, what difference does it make? Why even care about the exact number? From my POV, exactly nothing changes in a way we think of humans and their genomes whether it is 21191 or 23834. Seems like pointless numerology.

    ReplyDelete
  7. Not only is it pointless numerology, but isn't it idiotic to expect that all humans have the same number of genes? Shouldn't there be a range with variations? Now, we can talk about how wide that would be before it's surprising...

    ReplyDelete
  8. @Larry, there was a widespread view through much of the 1990s - and a very reasonable one - that once we knew the precise sequence, we would know where the genes were and (at least) how many there were. That turned out to be wrong, and that's what our short paper is trying to convey.

    Simply asserting that the (many) scientists who used ESTs to estimate gene content "didn't understand evolution" is a rather pointless ad hominem attack, and it's not true anyway.

    Now that we're learning that the genome has thousands of microRNAs, which were unknown in the 1970s, and thousands (maybe tens of thousands) of alternative splice variants, the number of distinct genetic elements will likely be much higher than current gene counts. We chose to ignore all this in our paper too - yes, space constraints are a real issue. We wrote a short perspective, not a book.

    We also wanted to draw attention to the fact that it's hard to find an agreed-upon gene list for the human genome right now. Instead you find multiple inconsistent gene lists.

    ReplyDelete
  9. Steven Salzberg says,

    @Larry, there was a widespread view through much of the 1990s - and a very reasonable one - that once we knew the precise sequence, we would know where the genes were and (at least) how many there were. That turned out to be wrong, and that's what our short paper is trying to convey.

    The early results from bacteriophage lambda, Hemophilus, E. coli, C. elegans, yeast, and especially Drosophila melanogaster soon dispelled that notion. The first large scale human sequences also revealed that it wasn't easy to find genes.

    I think everyone in the know understood that serious annotation had to take place before we had a good handle on what was a gene and what wasn't. I don't think anyone believed that an exact number would materialize given all the possible problems.

    Simply asserting that the (many) scientists who used ESTs to estimate gene content "didn't understand evolution" is a rather pointless ad hominem attack, and it's not true anyway.

    It's a lot closer to the truth than you seem to realize. In any case, they were wrong and the scientists who were experts in evolution and genetics turned out to be far more accurate.

    You can draw your own conclusions.

    Now that we're learning that the genome has thousands of microRNAs, which were unknown in the 1970s, and thousands (maybe tens of thousands) of alternative splice variants, the number of distinct genetic elements will likely be much higher than current gene counts. We chose to ignore all this in our paper too - yes, space constraints are a real issue. We wrote a short perspective, not a book.

    There may, indeed, be thousands of microRNAs and alternative splicing may be as prevalent as its strongest proponents claim. However, the debate over the validity of those claims is very much ongoing and a resolution isn't likely to happen soon.

    I note that the people on the two sides of this debate are, for the most part, the same as those on opposite sides of the EST gene number debate. Isn't that interesting?

    We also wanted to draw attention to the fact that it's hard to find an agreed-upon gene list for the human genome right now. Instead you find multiple inconsistent gene lists.

    That's interesting but you could have made the point just as effectively by introducing a more accurate history. Or by ignoring the history altogether. Once you choose to draw your figure and publish it, you implied that it was an accurate representation of the facts.

    It wasn't.

    ReplyDelete
  10. Genome Biology offers to add readers comments. Why don't you and Dr. Salzberg continue the discussionover there?

    ReplyDelete
  11. Steven et al--

    This reminds me of a Peanuts cartoon. Here is how I remember it:

    In the first frame, one sees Linus looking at a speck on the sidewalk below him and disclaiming something like this: "A Monarch butterfly! Ah, the noble wanderer! Having endured its journey of thousands of miles from Mexico, it breathes its last here on our humble sidewalk."

    In the next frame, Lucy walks up and says "Its a leaf".

    To which, Linus asks "How did that leaf get here, all the way from Mexico?"

    I think its curious, but not unusual, for scientists to seek physical causes for the discrepancy between their errors and reality, the way that the genomicists' generated inflated gene number estimates, then expected that there would be some special biological explanation for the "low" number of human genes.

    We could call it the "flying leaf" principle.

    Arlin

    ReplyDelete
  12. Larry,

    WOuld you happen to have the citation for those melting and re-hybridization experiments where the authors conclude that we have at most 10 times as many genes as E. coli? I remember the funny curves, and how the form of ours also showed that we had loads and loads of repetitive DNA. Old paper, probably late 60s or 70s.

    I have also been disappointed that articles estimating genes at such early stages in molecular biology did not make it into the citations of many articles of the post PDF era. These old articles are quite convincing, and would have saved us a lot of arguing. The elegance too. So little technology, such beauty of experimentation. So convincing a result.

    --Gabo

    ReplyDelete
  13. Gabo asks,

    WOuld you happen to have the citation for those melting and re-hybridization experiments where the authors conclude that we have at most 10 times as many genes as E. coli? I remember the funny curves, and how the form of ours also showed that we had loads and loads of repetitive DNA. Old paper, probably late 60s or 70s.

    There are dozens and dozens of papers from the 1970s. It's not possible to pick out one or two as representative of all the rest since the groups worked with different systems: sea urchin, Drosophila, tobacco, chicken, mouse, human.

    Benjamin Lewin published an excellent summary of the work in Gene Expression-2 in a chapter called "Complexity of RNA Populations."

    The basic idea is this. You take mRNA from any species or organ and hybridize it exhaustively against labeled DNA. What you measure is the amount of DNA that's driven into hybrid form by the excess RNA. These curves, which anyone over 50 should remember well, are called Rot curves.

    The experiments all converge on the same numbers. Only a few percent of the DNA is complementary to expressed RNA. A given tissue (e.g. embryos, liver cells in culture) will have an RNA complexity corresponding to about 10,000 genes. Different tissues will express different genes so the total RNA complexity of a typical vertebrate is equivalent to about 20,000 genes.

    Some mRNAs will be very abundant (thousands of copies per cell) and some will consist of less than one copy per cell. That's why the hybridization has to be carried out over a very long time (days).

    Lewin concludes ... (p. 718)

    The general conclusion suggested by these results is that the somatic tissues of higher (sic) organisms, or individual cell lines perpetuated in culture, have of the order of 10,000 - 20,000 active structural genes.

    The only exception was studies on mammalian brain tissue where the number tended to be higher—perhaps up to 70,000 genes expressed. Lewin relegates this to a footnote, mentioning that it seems anomalous. That was the general consensus at the time and it turns out to be correct. Brain cells seem to be very sloppy about transcribing DNA for some reason.

    I don't know how anyone could write a history of gene number estimates without featuring this data and the gene number estimates. Especially since they turned out to be correct. The mythology is based on "guesstimates" by people who were not familiar with the literature. They were not based, in the beginning, on any data at all.

    The only interesting part of the history is why some people fell for the guesstimate when it conflicted with a large amount of real scientific data.

    Steven Salzberg suggests (above) that they knew about all this data but they choose to ignore it for lack of space. On the other hand, they did find enough space to include a totally ridiculous estimate from 1964. That's the first point on their graph.

    I find that to be a very strange way of describing the history but that's his story and he seems to be sticking with it

    ReplyDelete
  14. Not just 1970. John Hawks uncovered one from 1948 predicting 20,000 genes:

    http://johnhawks.net/weblog/topics/history/genetics/spuhler-gene-number-2010.html

    ReplyDelete
  15. Larry writes:

    Lewin concludes ... (p. 718)

    "The general conclusion suggested by these results is that the somatic tissues of higher (sic) organisms, or individual cell lines perpetuated in culture, have of the order of 10,000 - 20,000 active structural genes."

    The only exception was studies on mammalian brain tissue where the number tended to be higher - perhaps up to 70,000 genes expressed. Lewin relegates this to a footnote, mentioning that it seems anomalous. That was the general consensus at the time and it turns out to be correct. Brain cells seem to be very sloppy about transcribing DNA for some reason.


    "...sloppy..." "...for some reason..." Larry, I am quite willing to believe that "junk" really is junk and that sometimes "shit happens". But anomalous transcription or alternative splicing in the brain, of all places, has to have gotten some people excited about looking for a function. This is OT, but can you recommend a good review of what, if anything, these excited people found? And what they still hope to find?

    ReplyDelete