A major motivation of the Human Genome Project (HGP) was to define a human gene catalog. But many years after the completion of the project in 2003, we still do not know how many genes there are in the human genome! As described in Section 9.2, much of the uncertainty concerns identifying RNA genes, but even in the case of protein-coding genes, getting an accurate number has not been straightforward.
At the time of the HGP, attempts to build a human gene catalog focused on identifying protein-coding genes, which were widely assumed to make the overwhelming genetic contribution to dictating how our cells work. RNA genes, by contrast, were viewed primarily as making ubiquitous accessory molecules that assisted in a general way to guide the expression of protein-coding genes. And although some noncoding RNAs had long been known to be more specific gene regulators, they were widely regarded at the time as quirky exceptions.
When the first draft human genome sequences began to be studied, therefore, the initial analyses primarily focused on looking for protein-coding genes. At least, protein coding genes are comparatively easy to identify in a genome sequence: there is an open reading frame, the predicted protein sequence has frequently been quite highly con served during evolution, and the reading frame is normally as long, or almost as long, in closely related species.
Preliminary analyses of draft human genome sequences reported in 2001 suggested about 30,000 (protein-coding) genes. Most were gene predictions made without any supportive experimental evidence and heavily reliant on identification of sizeable open reading frames. The original gene number was recognized to be inflated when the genome sequences of other mammals were obtained: evolutionary comparisons failed to identify counterparts of many predicted human genes in the mouse and dog genomes. And in several cases, two or more predicted neighboring genes turned out to be components of a much larger gene.
Recent protein-coding gene estimates
Analyses of the human genome GRChp38.12 reference sequence (released in December 2017) suggest that there are around 19,900–20,500 human protein-coding genes (Table 1). The variation in the number exists because there remain some difficulties in how to define genes. For example, read-through transcription has been found to occur between well-characterized neighboring genes on the same DNA strand, leading to the concept of conjoined genes (Figure 1).
Table1. HEADLINE HUMAN GENE STATISTICS FROM THE GRCh38.p12 REFERENCE SEQUENCE (NOVEMBER 2017 FREEZE)
Fig1. An example of read-through transcription leading to the concept of a conjoined gene. The neighboring NME1 and NME2 genes on chromosome 17q21 are members of a family of genes making nucleoside diphosphate kinases. Red and blue boxes indicate exons incorporated in, respectively, NME1 and NME2 transcripts (connecting lines are introns and gray boxes represent coding sequences (CDS); the identification numbers for reference mRNAs in the RefSeq database are shown at left). In addition to transcripts from the parent genes, other transcripts are found that link exons of the two genes, including two transcripts with GenBank reference sequences DQ109675 and BC107894 (exons shown here as magenta boxes) and numerous EST (expressed sequence tag) sequences (not shown here). (Adapted from Prakash T et al. (2010) PLoS One 5:e13284; PMID 20967262.)
In the example in Figure 1, one might conceivably count two parental genes plus a conjoined gene. There are close to 500 examples of read-through transcription, and depending on whether conjoined genes are counted as additional genes or not, estimates of the number of human protein-coding genes can vary (see Table 1).
Further complications stem from how we classify genes. A few genes, for example, seem to make both coding transcripts (mRNA) and also transcripts that make a functional noncoding RNA. Additionally, genes originally classified as making long noncoding RNAs have subsequently been found to have very short, central coding DNA sequences that make micropeptides.
The total number of human protein-coding genes is likely to stabilize somewhere in the region of 20,000, but the number of nuclear protein-coding genes can never be exact, unlike the invariant 13 protein-coding genes in our mitochondrial genome. That is so because of structural variation in the nuclear genome: for many families of protein coding genes, the copy number varies between haploid genomes. Differences in gene number can be expected from one individual to the next, and between the maternal and paternal genomes inherited by a person.
