At the time of writing, 15 years since the completion of the Human Genome Project in 2003, the sequence of the nuclear genome remains incomplete: the “finished sequence” is not finished! Very long stretches of tandem repeats offer a particularly difficult challenge to both DNA cloning (because of instability of DNA inserts) and to genome assembly (because of limitations in the lengths of insert DNA that can be sequenced).
Long arrays of tandem repeats are predominantly found in heterochromatin DNA, and so because of the difficulty in sequencing this DNA (and the general lack of interest in gene-lacking heterochromatin), the Human Genome Project was limited to sequencing just the euchromatin DNA (which accounts for just over 93% of the human genome). For technical reasons, too, no attempt was made to fully sequence the multi-megabase ribosomal DNA clusters (the arrays of tandem 45S rRNA repeats that specify the 28S, 5.8S, and 18S rRNAs). Because of various difficulties there are also significant gaps (missing sequence) within the euchromatin DNA sequence plus unplaced contigs.
Recall from Section 7.1 that the nuclear genome component of the human genome reference sequence – abbreviated as the nuclear reference sequence below – is not a single haploid genome from a single individual. That would have allowed relatively easy assembly of the genome sequence (the ideal assembly involves a golden path, a unique clone tiling path that can be reduced to one nonredundant haploid representation of the genome). Instead, the DNA clones contributing to the initial reference sequence came from blood-cell samples provided by a selected set of anonymous human donors, each with two different haploid genomes. As a result, the reference nuclear sequence is a mosaic with pieces of sequence from different individuals, each with two parental haplotypes (but is nevertheless dominated by the contribution from a single individual). And, of course, 24 chromosomal DNA molecules are represented in the nuclear reference sequence, instead of the 23 DNA molecules in haploid human cells.
The ideal nuclear reference sequence should be made up exclusively of clone contigs, ordered series of clones having overlapping DNA sequences without gaps. The reality, however, is that the reference sequence consists of multiple scaffolds, each consisting of an ordered and oriented set of contigs, but with some gaps between constituent contigs. Although a scaffold has gaps, however, there must be some evidence to support the order and orientation of its constituent contigs, with approximate estimates made for the gap size(s) within the scaffold.
The “finished” human genome sequence reported in 2003 had a few hundred gaps in the euchromatin DNA sequence, often occurring at regions prone to structural variation—large-scale changes involving deletions, duplications, insertions, or inversions. At subchromosomal regions, where the reference sequence is derived from DNA coming from different haplotypes, structural differences between the haplotypes may cause problems: one haplotype may have a sizeable deletion, for example, so that there is no sequence to form an overlap with a neighboring clone from a different haplotype. We now know that structural variation is much more common than first anticipated: the original human genome reference sequence was subsequently found to lack many sequences present in some humans because the source DNAs for genome sequencing had some large deletion alleles at several loci. About 1.4% of the genome sequence of James D Watson, for example, could not be matched to the current reference sequence at that time, hg18.
To address these difficulties, the Genome Reference Consortium (http://genomeref erence.org) was formed in 2007, and various improvements to the reference sequence have been made. Some gaps have been closed, and revisions are made periodically by adding scaffold sequence patches that are of two types:
(a) FIX patches, which are intended to correct errors in an assembly;
(b) NOVEL patches, which add alternative loci. Certain regions of the genome show highly significant and frequent structural variation, and so additional alternative sequences are included at some loci in order to have a more representative reference sequence. In certain subchromosomal regions, therefore, the reference sequence now has multiple alternative haplotypes, coming from additional human donors.
At the time of writing, the most recent human genome reference sequence is GRCh38. p12 (Genome Reference Consortium human assembly 38, patch release 12) which was released in December 2017 and can be downloaded at https://www.ncbi.nlm.nih.gov/ genome/guide/human/. As well as gaps between contigs within scaffolds, there are 349 gaps between scaffolds. Additionally, there remain two types of unassigned DNA sequence: unlocalized sequences (DNA sequences associated with a specific chromosome but whose order or orientation on that chromosome is unknown) and unplaced sequences (sequences found in the assembly but not associated with a specific chromosome). In the GRCh38.p12 reference sequence, the primary assembly—the collection of assembled chromosomal DNA sequences plus unlocalized and unplaced sequences—has a total length of 3,099,706,404 nucleotides including the length of all gaps (the cumulative length of gaps is 151,122,679 nucleotides).
The total GRCh38.p12 reference sequence has 3,257,319,537 nucleotides because in addition to the nearly 3.1Gb in the primary assembly sequence and the 16,569 nucleotides in mtDNA, it has just over 48 Mb of sequence representing 70 FIX patches and 109.5 Mb of multiple alternate reference loci. In the latter case, for example, variation in the major histocompatibility complex (MHC) is represented by eight different haplotypes, not just one. For some information on accessing alternate human sequence, see the Ensembl blog at www.ensembl.info/2011/05/20/accessing-non-reference-sequences-in-human/
The number of gaps is gradually being reduced, notably with the help of ultra long read nanopore sequencing (PMID 29431738). Limitations in DNA technology has meant that progress has been slow, but next-generation optical mapping techniques (such as Bionano Genomics’s Saphyr system) are designed to offer high-throughput, high resolution mapping of individual megabase-sized DNA molecules. Progress has also recently been made in assembling DNA sequences in heterochromatic regions of our genome (which were off limits in the Human Genome Project). Lack of human centromere sequence assemblies used to be simply acknowledged in the reference sequence as standard gaps of 3 Mb, but in 2014, the first centromere reference models were produced when two haploid human satellite arrays were modeled on the X and Y chromosomes (PMID 24501022), and reference centromere models have now been produced for each human chromosome. A significant recent advance has been the linear assembly of the centromere DNA on the Y chromosome (PMID 29553574).
Progress is also being made in developing sequence assemblies for the short arms of the acrocentric chromosomes (13, 14, 15, 21 and 22), including the nucleolar organizer regions (NORs) that contain tandem transcription units specifying 28S, 18S and 5.8S rRNA (ribosomal DNA repeats). Because the very similar NORs on the five acrocentric chromosomes associate to form a chromosomal structure around which a nucleolus can form, sequence exchanges are common between the ribosomal DNA repeats on different acrocentric chromosomes. To avoid the complication of interchromosomal sequence exchange, human–mouse hybrid cells with a single acrocentric human chromosome are used to prepare DNA for sequence assemblies. Thereafter, very long human DNA templates can be prepared using methods such as transformation-associated recombination in yeast, followed by long-read DNA sequencing of selected template DNAs (PMID 29788454).
