Tuesday, May 12, 2009

The Human Genome Sequence Is not Complete

The latest version of the human genome sequence is called Build 36 or GRCh37. Here's an overview from the Genome Reference Consortium.

The large red triangles represent regions where there is a lot of variability so that no single representation of the genome sequence will describe a majority of humans.

The black regions represent parts of the chromosomes that have not been sequenced and assembled into long stretches (contigs) of reliable sequence. Most of the unsequenced regions are at centromeres, or telomeres, or on the Y chromosome. These regions consist of thousands of copies of highly repetitive DNA. It is impossible to assemble these repetitive sequences.

Scientists are urging that more attention be focused on completing the chimpanzee and macaque genome sequences. We have been waiting a long time for the draft sequences of those genomes to be finished. The explosion of data on the human genome can only be realistically evaluated by comparing it to our closest relatives. (For example, are human non-coding RNAs conserved in primates?)

The fact that the human genome is not complete is not a problem. We know what's in the repetitive sequence regions even though we don't know exactly how it is arranged. The effort required to finish of the last bit is probably not as important as getting a final draft of other sequences.

Sandra Porter wonders Why don't we finish the human genome first?.


  1. Back in 1991 I heard Charles Cantor -- then head of the human genome project -- give a lecture in which he said that 90% of the work would be needed for the last 10% of the results. He added that once the first 10% of the work had been done we would be hearing that the genome had been completed. In the last few years I have had the impression that he was right.

  2. Even though many of the gaps remaining in the human genome contain some repetitive sequence, many of them are also in phenotypically important regions. This doesn't mean that it is not important to finish other non-human primates, because it is. But, it is also important to get these last bits of the human sorted out well- especially in the regions of high diversity.

  3. Oh- and the last build was Build 36. The just released build is GRCh37.

  4. The problem with this discussion is that two very different things are being considered that occur on very different time scales. Producing a DRAFT sequence of a genome is (now) a relatively quick task and provides access to broad brush answers such as an estimate of how many genes, types of repeats, broad sequence conservation statistics, and fodder for broad brush comparison with other genomes, including other individuals etc. If you are lucky it will give you the structure of specific genes. However for many regions the sequence will be incomplete or inaccurate. There are different definitions of what might be considered draft, and enhancing the draft by adding BAC clone based sequences is beneficial in the long term to allow better completion, but takes more time. New sequencing technologies are changing the approaches, but it is still true to say that getting to a "finished" reference sequence is a whole different ball game because it requires an additional step that is time consuming (see the introduction of my groups article at PMID: 18477386 for a brief discussion of methods). However there are also different degrees of finished sequence, and at some point the sequence is in a state that is sufficient for the vast majority almost all conceivable uses, is accurate to a tested standard, and contains the overwhelming majority of sequence that the overwhelming majority of scientists are interested in. The human genome is currently verging on that state, and while one can consider that it would be nice to fill in some of the remaining gaps, this is a non trivial process and takes time. Some gaps may be easily filled but many are there for a reason such as that they border locations of paralogous sequence repeat or that they are problematic for current cloning systems. It is possible that these sequences will eventually be extracted from large scale genome sequencing such as the 1000 genomes project, but in my view getting finished sequence will still require hard graft at the coal face (again see the paper above for examples of some of the approaches we tried).

    So we have two different goals. 1. Getting coverage in usable but not necessarily complete sequence on many genomes. 2. Polishing, finisihing or completing genome sequences that will be valuable in a reference form. These goals are not an eitehr/or, rather both are needed and they are achieved on different time scales.

    The problem of large blocks of tandem repeats such as the ribosomal gene arrays or centromeric satellite sequences is, in my view, a completely separate, but interesting one. The truth is that although there was a period of examining the nature of these sequences by pulsed field gel electrophoresis and sequencing some years back, nobody has really come up with an approach that would be able to achieve long range sequence contigs that could resolve hundreds of kilobases of highly similar (mostly identical) tandem repeats. In fact one could argue that it is not really a sequencing problem, rather a mapping one. So to expect a finished reference sequence of these regions is is to be disappointed. However it is also true that the sequences are there to work on in the detritus that high throughput genome sequencing throws out (for instance the reads of the 1000 genomes project) so this would make a really ground breaking project for someone who doesn't fear the high likelihood of failure. i spent my first postdoc trying to come up with ways to quantitate human centromeric repeats in yeast artificial chromosomes without any great success!

    Finally I'd like to comment on the Cantor quote. This kind of observation is easy to make, and is true of almost all projects - think of the effort that is put into preparing the paper at the end of any project. However in all probability I think that the last 20-30% of the work for the human genome (lumped together as heterochromatin by the HGRC) will receive very little extra effort, because the world moves on, and as yet there has not been any overwhelming call that it is really important to know how many copies of such and such alphoid satellite repeat there is at the centromere of this chromosome. It may yet turn out to be important (perhaps in studies of fertility?) but it will be a hard nut to crack.

  5. Ian,

    Why were the pulsed field approaches to characterizing tandem repeat blocks abandoned? Has this been obsoleted by molecular combing? At least this would give some idea of the extent of human variation and meiotic instability rate at these loci. There are still some "finished" loci in GRGh37 that are going need to be revisited and corrected on this basis. See for example PMID 18992157, 18025267.

  6. I think people just got bored with the PFGE approach. It would be nice to go back now and do a survey from many individuals on some of these tandems (of the top of my head I think there were some comparisons between inidivduals). You have to remember, also, that your resolution with the sort of fragments that you would want to deal with for, say, centromeric alphoid sequences is not great.

    I'm not sure whether the size one can look at with molecular combing is suitable for Mb scale repeat blocks.

    It's right that there are finished sequences that need revisiting, since there are errors, plus significant other haplotypes.