A list of weird gene annotations or things that break bioinformatics assumptions
See also https://github.com/cmdcolin/oddbiology/ for more weird bio
Figures are reproduced from their original sources with attribution and keep their original licenses; the text of this repo is CC0
Evidence for a 1bp exon in Arabidopsis, with competing splicing models discussed
http://www.nature.com/articles/srep18087
Another 1bp exon is discussed here https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0177959
Microexons in general are an interesting topic and are "involved in important biological processes in brain development and human cancers" (ref https://www.cell.com/molecular-therapy-family/nucleic-acids/fulltext/S2162-2531(23)00013-6) yet are commonly misannotated (e.g. in plants https://www.nature.com/articles/s41467-022-28449-8)
See also cryptic splice sites and exons, poison exons
The phenomenon of recursive splicing can remove sequences progressively inside an intron, so there can exist "0bp exons" that are just the splice-site sequences pasted together.
"To identify potential zero nucleotide exon-type ratchet points, we parsed the RNA-Seq alignments to identify novel splice junctions where the reads mapped to an annotated 5' splice site and an unannotated 3' splice site, and the genomic sequence at the 3' splice site junction was AG/GT"
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4529404/
Aberrant recursive splicing may contribute to disease https://www.biorxiv.org/content/10.1101/2025.08.14.666599v1?med=mas
Satellite DNA study uncovers megabase scale introns https://www.biorxiv.org/content/early/2018/12/11/493254
An example in this paper kl-3 spans 4.3 million bp
In human, an example is Dystrophin. "Dystrophin is coded for by the DMD gene – the largest known human gene, covering 2.4 megabases (0.08% of the human genome) at locus Xp21. The primary transcript in muscle measures about 2,100 kilobases and takes 16 hours to transcribe; the mature mRNA measures 14.0 kilobases" https://en.wikipedia.org/wiki/Dystrophin
Transcribing introns this large is wasteful: the cell synthesizes megabases of RNA and then splices nearly all of it back out. Dystrophin's 16-hour transcription time also means a rapidly dividing cell cannot finish one transcript before the next division interrupts it https://pmc.ncbi.nlm.nih.gov/articles/PMC2754300/
Human TTN (titin) has ~364 exons, nearly double the runner-up NEB (nebulin) at ~184
https://www.ncbi.nlm.nih.gov/gene?Db=gene&Cmd=DetailsSearch&Term=7273
"A 2015 study suggests that the shortest known metazoan intron length is 30 base pairs (bp) belonging to the human MST1L gene (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4675715/). The shortest known introns belong to the heterotrich ciliates, such as Stentor coeruleus, in which most (> 95%) introns are 15 or 16 bp long (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5659724/)" https://en.wikipedia.org/wiki/Intron#Distribution
A novel splicing factor may be involved in small introns https://www.news-medical.net/news/20240215/Novel-splicing-mechanism-for-short-introns-discovered.aspx
An alga described in 2024 encodes PKZILLA-1, a 4.7 megadalton protein with 140 enzyme domains https://cen.acs.org/biological-chemistry/PKZILLA-proteins-smash-protein-size/102/web/2024/08
Human titin, the TTN product in muscle, is almost 4 megadaltons
The DMD gene above, despite being the largest known human gene (2.4 Mb), encodes a ~427 kDa protein — large, but nowhere near megadalton scale. A shorter isoform Dp71 (~71 kDa) is expressed in non-muscle tissues. https://pmc.ncbi.nlm.nih.gov/articles/PMC49288/
Inverted Alu pairs flanking an exon can fold into an RNA stem-loop hairpin that physically loops the exon out, causing skipping through RNA secondary structure alone rather than splicing factors. ~707 human exons are affected, including TBXT (linked to tail loss in hominoids).
https://academic.oup.com/nar/article/54/6/gkag196/8539533
Backsplicing circularizes RNAs, and alternative backsplicing yields different circles from the same gene
Figure from Dawoud et al. https://doi.org/10.1016/j.ncrna.2022.09.011
See https://academic.oup.com/nar/article/48/4/1779/5715065
"Dscam has 24 exons; exon 4 has 12 variants, exon 6 has 48 variants, exon 9 has 33 variants, and exon 17 has two variants. The combination of exons 4, 6, and 9 leads to 19,008 possible isoforms with different extracellular domains (due to differences in Ig2, Ig3 and Ig4). With two different transmembrane domains from exon 17, the total possible protein products could reach 38,016 isoforms"
Ref https://en.wikipedia.org/wiki/DSCAM https://www.wikigenes.org/e/gene/e/35652.html
Ref https://en.wikipedia.org/wiki/Translational_frameshift
https://www.sciencedirect.com/topics/neuroscience/ribosomal-frameshifting
SARS-CoV-2 uses ribosomal frameshifting and this video shows a 3D animation of the process, showing a 'pseudoknot' in the RNA contributes to it https://www.youtube.com/watch?v=gLcueW61QMU
Another lecture explaining frameshift in viruses https://youtu.be/b5BX5A3dGUQ?t=2980
In retroviruses like HIV, the gag and pol genes overlap in different reading frames. A -1 ribosomal frameshift at a "slippery sequence" between them produces the Gag-Pol fusion polyprotein at ~5% efficiency, while the other 95% of ribosomes terminate at the gag stop codon and produce only Gag. This ratio is critical — altering it is lethal to the virus. The Gag-Pol polyprotein is then cleaved by the viral protease (which is itself part of the polyprotein) to produce reverse transcriptase, integrase, and protease.
https://en.wikipedia.org/wiki/Gag-pol https://en.wikipedia.org/wiki/Pol_(HIV)
Figure from Atkins et al. (2016) showing the ribosome encountering a slippery sequence (X XXY YYZ) and downstream RNA stimulatory element (pseudoknot) that together promote -1 programmed ribosomal frameshifting https://pmc.ncbi.nlm.nih.gov/articles/PMC7618472/
"Ribosome hopping involves ribosomes skipping over large portions of an mRNA without translating them" Ref https://pubmed.ncbi.nlm.nih.gov/24711422/
The classic example is bacteriophage T4 gene 60, where the ribosome bypasses a 50-nucleotide coding gap — about half of ribosomes successfully make the hop https://pmc.ncbi.nlm.nih.gov/articles/PMC107096/
An IRES allows ribosomes to initiate translation at an internal position on an mRNA without scanning from the 5' cap. This enables cap-independent translation and is used by many viruses to hijack host ribosomes. Some cellular mRNAs also contain IRES elements, allowing them to be translated under stress conditions when cap-dependent translation is shut down.
https://en.wikipedia.org/wiki/Internal_ribosome_entry_site
"Stop codon suppression or translational readthrough occurs when in translation a stop codon is interpreted as a sense codon, that is, when a (standard) amino acid is 'encoded' by the stop codon. Mutated tRNAs can be the cause of readthrough, but also certain nucleotide motifs close to the stop codon. Translational readthrough is very common in viruses and bacteria, and has also been found as a gene regulatory principle in humans, yeasts, bacteria and drosophila. This kind of endogenous translational readthrough constitutes a variation of the genetic code, because a stop codon codes for an amino acid. In the case of human malate dehydrogenase, the stop codon is read through with a frequency of about 4%. The amino acid inserted at the stop codon depends on the identity of the stop codon itself: Gln, Tyr, and Lys have been found for the UAA and UAG codons, while Cys, Trp, and Arg for the UGA codon have been identified by mass spectrometry. Readthrough extent in mammals varies widely, and can broadly diversify the proteome and affect cancer progression."
https://en.wikipedia.org/wiki/Stop_codon#Translational_readthrough
Two amino acids are encoded by codons that normally stop translation. The "opal" (UGA) codon encodes selenocysteine https://en.wikipedia.org/wiki/Selenocysteine in all domains of life, humans included: a stem-loop in the mRNA called a SECIS element https://en.wikipedia.org/wiki/SECIS_element tells the ribosome to read that UGA through, and the products are selenoproteins, of which the human selenoproteome has 25 https://www.science.org/doi/10.1126/science.1083516
The "amber" (UAG) codon encodes pyrrolysine https://en.wikipedia.org/wiki/Pyrrolysine in some archaea and bacteria but not in humans, and needs only two genes to do it: pylT, an unusual tRNA with a CUA anticodon, and pylS, the synthetase that charges it.
Several other stop codon modifications are described here https://www.nature.com/articles/nrg3963
In mammalian apolipoprotein B, RNA editing introduces a premature stop codon rather than removing one. The default unedited mRNA encodes the full-length B100 isoform (550 kDa). In the intestine, C-to-U editing at nucleotide 6666 converts a glutamine codon (CAA) into a stop codon (UAA), producing the truncated B48 isoform (265 kDa). This is one of the best-characterized examples of RNA editing altering protein output.
https://pubmed.ncbi.nlm.nih.gov/8409768/
"Flexibility in the nuclear genetic code has been demonstrated in ciliates that reassign standard stop codons to amino acids...Surprisingly, in two of these species, we find efficient translation of all 64 codons as standard amino acids and recognition of either one or all three stop codons"
Translation termination is therefore "context dependent" — determined by surrounding sequence rather than just the 3-letter codon https://pubmed.ncbi.nlm.nih.gov/27426948/
See also how Ciliates perform programmed DNA elimination (see that section) and have some of the smallest known introns (see small introns section)
See also this Ensembl blog on annotating readthrough transcription which joins multiple genes http://www.ensembl.info/2019/02/11/annotating-readthrough-transcription-in-ensembl/
RNA-seq often makes a compelling case that two or more genes are joined by splicing
Tools like mikado https://academic.oup.com/gigascience/article/7/8/giy093/5057872 treat these as artifactual fusions or chimeras, sometimes traceable to tandem duplication, but they are very common in real data
The standard splice site is GT on the 5' end (donor) and AG on the 3' end (acceptor), processed by the major (U2-type) spliceosome. Non-GT-AG sites exist: GC-AG is the most common variant (~1% of introns) and is still handled by the major spliceosome. A separate minor (U12-type) spliceosome handles a distinct class of introns, originally identified by AT-AC termini but now known to also include GT-AG introns with different internal sequences.
https://en.wikipedia.org/wiki/Minor_spliceosome
Cryptic splice sites resemble splice sites but go unused, until a mutation disrupting the normal site or a shift in splicing factor levels activates them. Deep intronic variants are a major source and account for a large fraction of splice-disrupting events in genetic disease.
When two cryptic sites flanking a stretch of intron fire together, that stretch is included as a cryptic exon, usually carrying a premature termination codon that sends the transcript to nonsense-mediated decay. The degradation is what makes them hard to see. In ALS/FTD, TDP-43 normally represses inclusion of nonconserved cryptic exons, so when it is depleted from the nucleus they appear in neuronal mRNAs like STMN2 and UNC13A https://www.science.org/doi/10.1126/science.aab0983 https://www.nature.com/articles/s41586-022-04424-7
Review article https://academic.oup.com/nar/article/39/14/5837/1382796 and one on the cryptic exon side https://link.springer.com/article/10.1186/s13024-023-00608-5
Snaptron, from Ben Langmead's group, indexes huge amounts of public RNA-seq and surfaces many kinds of cryptic splicing http://snaptron.cs.jhu.edu/, and SpliceVault re-analyzes public RNA-seq to predict cryptic activation empirically https://www.nature.com/articles/s41588-022-01293-8
See also poison exons
Poison exons are conserved alternative exons that contain a premature termination codon (PTC). When included via alternative splicing, the PTC triggers NMD, degrading the transcript. Unlike cryptic exons which are aberrantly activated, poison exons are a deliberate gene regulation mechanism: cells fine-tune protein levels by adjusting the ratio of productive (poison-exon-skipped) vs. NMD-targeted (poison-exon-included) transcripts.
Splicing factors like SR proteins and hnRNPs autoregulate their own levels via poison exon inclusion. Therapeutically, ASOs that block poison exon inclusion can upregulate expression of disease-relevant genes.
Review https://www.sciencedirect.com/science/article/pii/S0165614725000021
Poison exons in neurodevelopment and disease https://pmc.ncbi.nlm.nih.gov/articles/PMC8042789/
Exitrons (exonic introns) are internal regions within protein-coding exons that can be spliced out like introns. When an exitron is retained it contributes to the protein sequence; when spliced out it causes an in-frame deletion or frameshift.
Exitron splicing is dramatically elevated in cancer: one study found 63% of human coding genes are affected in tumors vs. 17% in normal tissue, and they can generate tumor-specific neoepitopes https://www.cell.com/molecular-cell/fulltext/S1097-2765(21)00223-9
Original paper defining exitrons https://genome.cshlp.org/content/25/7/995.full
Repeats of the splice signal itself — NAGNAG at acceptors, GYNGYN at donors — shift the chosen splice site by a few nucleotides
"Another mechanism introducing small variations to protein isoforms is wobble splicing. Here, a GYN repeat at the donor splice site (5’ splice site; Y stands for C or T and N stands for A, C, G, or T) or an NAG repeat at the acceptor splice site (3’ splice site) leads to subtle length variations in the spliced transcripts and finally to alternative isoforms differing in few amino acids." ref https://onlinelibrary.wiley.com/doi/full/10.1002/bies.201900066?af=R
Intron retention (IR) leaves intron sequence unspliced in the mature RNA. It happens in both normal and disease conditions, and retained-intron transcripts often undergo nonsense-mediated decay.
https://en.wikipedia.org/wiki/Alternative_splicing
Review: https://pmc.ncbi.nlm.nih.gov/articles/PMC12385487/
Normally RNA is spliced by the spliceosome, a ribonucleoprotein complex where the catalysis is actually performed by its RNA components (snRNAs). There is also self-splicing RNA where the RNA catalyzes its own splicing without the spliceosome
Group I introns use a free guanosine as a cofactor to catalyze their own excision https://en.wikipedia.org/wiki/Group_I_catalytic_intron
Group II introns use a different mechanism (lariat formation, similar to spliceosomal introns) and are thought to be evolutionary ancestors of the spliceosome https://en.wikipedia.org/wiki/Group_II_intron
Archaea (and some other organisms) splice tRNA introns whose ends fold into a "bulge-helix-bulge" motif, cut by a tRNA splicing endonuclease rather than the spliceosome
From https://www.embopress.org/doi/full/10.1038/embor.2008.101
The figure above shows that the orange part is excised as an intron for the tRNA
A twintron is an intron-within-an-intron. The internal intron must be spliced first before the outer one can be recognized and spliced (if the internal one does not need to be spliced first, it is simply called a nested intron).
See https://en.wikipedia.org/wiki/Twintron
Figure from https://doi.org/10.1080/15476286.2015.1103427 showing twintron conformations with a) spliceosomal introns b) group I/II self-splicing introns and c) tRNA/bulge helix bulge type introns
Introns were first discovered in adenoviruses (1977), not eukaryotes — a finding that won Roberts and Sharp the Nobel Prize
https://en.wikipedia.org/wiki/Intron#Discovery_and_etymology (see also https://www.proquest.com/docview/303935681/)
Pieces of the mitochondrial genome can be inserted into nuclear chromosomes in eukaryotes. The human genome contains over 700 NUMTs. They confound mitochondrial studies because PCR primers for mtDNA can accidentally amplify NUMTs, and new insertions continue to occur — some are pathogenic.
https://en.wikipedia.org/wiki/Nuclear_mitochondrial_DNA_segment https://pmc.ncbi.nlm.nih.gov/articles/PMC3228813/
The 5' and 3' UTR (untranslated region) are parts of the mature mRNA that are not translated into protein — the 5' UTR is before the start codon, the 3' UTR is after the stop codon. Unlike introns, UTRs are retained in the final mRNA. They are often targets of miRNA binding which can lead to degradation.
This blog post by Ensembl shows how they annotate UTR and a 19kb 3' UTR in Grin2b http://www.ensembl.info/2018/08/17/ensembl-insights-how-are-utrs-annotated/
A poly-A signal (typically AAUAAA) directs cleavage of the pre-mRNA and addition of ~200 adenines that are not encoded in the genome https://en.wikipedia.org/wiki/Polyadenylation
A survey of poly-A using Oxford Nanopore found a transcript isoform with a 450bp poly-A tail (ENST00000581230), with intron retention being a possible correlate of longer poly-A tails https://www.biorxiv.org/content/early/2018/11/09/459529.article-info
Polyadenylation normally happens after the last exon, but a poly-A signal sitting inside an intron can be used instead — "intronic polyadenylation" (IpA). That intronic position becomes the end of the transcript and everything downstream, including the remaining exons, is lost, giving a shorter mRNA that often encodes a truncated or nonfunctional protein. The name misleads: nothing gets spliced away here, the intron is simply where transcription stops. The classic example is immunoglobulin M — naive B cells use a downstream poly-A site and make membrane-bound IgM, while activated B cells switch to an intronic site and make the secreted form, a key step in the immune response (https://pmc.ncbi.nlm.nih.gov/articles/PMC9715272/, https://pubmed.ncbi.nlm.nih.gov/9885564/).
Figure showing "intronic polyadenylation" (IpA) creating a different isoform from https://www.nature.com/articles/s41467-018-04112-z
In mammalian mitochondria, some mRNAs are polyadenylated immediately after a genomically encoded U — the poly-A tail supplies the two A's needed to complete a UAA stop codon. Without polyadenylation, these transcripts would lack a stop codon entirely (noted in https://www.ncbi.nlm.nih.gov/Taxonomy/Utils/wprintgc.cgi#SG2).
Circular chromosomes are routine for plasmids and prokaryotic genomes, but annotation formats use linear coordinates, so anything wrapping past the origin is awkward to represent.
While mitochondrial genomes are often assumed to be circular, some eukaryotes have linear mitochondrial DNA — including certain yeasts (Candida) and some plants and protists https://www.sciencedirect.com/science/article/abs/pii/S0168952503003044
Most genome viewers handle this poorly. GFF represents it by letting a feature end past the end of the sequence. Below the genome is 6407 bp long, but the CDS runs past that and sets Is_circular=true
##gff-version 3.2.1
# organism Enterobacteria phage f1
# Note Bacteriophage f1, complete genome.
J02448 GenBank region 1 6407 . + . ID=J02448;Name=J02448;Is_circular=true;
J02448 GenBank CDS 6006 7238 . + 0 ID=geneII;Name=II;Note=protein II;
GFF3 specification: https://github.com/the-sequence-ontology/specifications/blob/master/gff3.md
Genes can overlap, either on opposite strands (sense-antisense) or on the same strand in a different reading frame
https://en.wikipedia.org/wiki/Overlapping_gene
Some articles
- The novel EHEC gene asa overlaps the TEGT transporter gene in antisense and is regulated by NaCl and growth phase https://www.ncbi.nlm.nih.gov/m/pubmed/30552341/
- Overlapping genes in natural and engineered genomes https://www.nature.com/articles/s41576-021-00417-w
- Uncovering de novo gene birth in yeast using deep transcriptomics https://www.nature.com/articles/s41467-021-20911-3
Hepatitis B is an extreme example of overlapping genes — its ~3.2 kb genome packs 4 overlapping reading frames, with every nucleotide belonging to at least one gene https://pmc.ncbi.nlm.nih.gov/articles/PMC3126273/
Figure from https://commons.wikimedia.org/wiki/File:HBV_Genome.svg
Many genes produce transcripts from both strands of DNA at the same locus. Up to 70% of mammalian genes show evidence of antisense transcription. These natural antisense transcripts (NATs) can regulate the sense gene through various mechanisms https://www.nature.com/articles/nrm2738
The Xist/Tsix pair (see X chromosome inactivation section) is a well-known example — Tsix is an antisense transcript that represses Xist on the active X
Entirely distinct genes can reside within the introns of larger genes, transcribed independently and often from the opposite strand. The NF1 gene (~350 kb, 60 exons) has three genes inside its largest intron (intron 27b, ~60 kb), all on the opposite strand:
- OMGP — oligodendrocyte-myelin glycoprotein, inhibits neurite outgrowth
- EVI2B — transmembrane glycoprotein required for granulocytic differentiation
- EVI2A — putative transmembrane protein involved in hematopoietic specification
None of them carry neurofibromatosis-causing mutations, and all are regulated independently of NF1.
https://pmc.ncbi.nlm.nih.gov/articles/PMC359746/
Many mRNAs contain small open reading frames in their 5' UTR, upstream of the main protein-coding ORF. These uORFs are translated by ribosomes before they reach the main start codon, and can regulate translation of the downstream protein — often by reducing its translation efficiency, but sometimes by producing small functional peptides. About 50% of human mRNAs contain uORFs, yet they are routinely ignored by gene annotation pipelines.
https://en.wikipedia.org/wiki/Upstream_open_reading_frame
Chimeric genes (also called gene fusions) are formed when parts of two or more previously separate genes combine into a single new gene. This can happen through several mechanisms: retrotransposition of an mRNA into or near another gene, tandem duplication followed by fusion, unequal crossing-over between paralogs, transposon-mediated capture of gene fragments, or even readthrough transcription followed by retroposition. While many chimeric genes are nonfunctional, some acquire novel functions and are positively selected.
The gene Jingwei is a chimera (or fusion) of two genes, alcohol dehydrogenase and yellow emperor in Drosophila. Many chimeras are damaging but this has been selected for. Remarkably, two other chimeric genes (Adh-Finnegan and Adh-Twain) also independently derived from Adh via retroposition in different Drosophila lineages -- parallel chimeric gene evolution from the same parent gene http://www.pnas.org/content/101/46/16246 https://www.pnas.org/doi/10.1073/pnas.0503528102
Two Cytochrome P450 genes that don't confer any insecticide resistance on their own but a chimeric P450 does https://pubmed.ncbi.nlm.nih.gov/22949643/
SDIC in Drosophila melanogaster is a chimera of annexin X and cytoplasmic dynein intermediate chain that converted a cytoplasmic motor protein into a sperm-tail motor protein by replacing the original functional domain. Present in ~10 tandem copies on the X chromosome, it boosts sperm competition https://www.nature.com/articles/25126
Sphinx is a chimeric gene formed by retroposition of an ATP synthase gene that accumulated nonsense mutations and became a noncoding RNA. Despite losing all coding potential, it regulates male courtship behavior in Drosophila -- knockout causes increased male-male courtship https://www.pnas.org/doi/10.1073/pnas.072066399
TRIMCyp is a fusion of TRIM5 and cyclophilin A created when L1 retrotransposition inserted CypA cDNA into TRIM5. The fusion protein restricts HIV-1 infection. The same chimeric gene arose independently in both owl monkeys and pigtailed macaques via separate retrotransposition events -- convergent evolution at the gene-structure level https://www.pnas.org/doi/10.1073/pnas.0404640101
The antifreeze glycoprotein (AFGP) gene in Antarctic notothenioid fish evolved from a trypsinogen gene. The entire protein core was replaced by de novo amplification of a 9-nucleotide repeat from a tiny element at the boundary of trypsinogen's first intron. The resulting protein has zero sequence homology to its parent gene https://www.pnas.org/doi/10.1073/pnas.94.8.3811
Transposons can capture fragments of host genes during excision or replication and carry them to new genomic locations, creating chimeric genes. This "transduplication" occurs across multiple transposon superfamilies and is a major source of new genes, especially in plants.
Pack-MULEs in rice carry fragments from >1,000 genes, with ~23% fusing fragments from multiple loci into chimeric ORFs. Also found in maize and the dicot Lotus japonicus, suggesting this predates the monocot-dicot split https://www.nature.com/articles/nature02953
Helitrons (rolling-circle transposons) are the main gene-capturing transposons in animals, found in maize, bats, fungi, insects, and nematodes https://www.nature.com/articles/ncomms10716
Gene capture has been observed in real time with Pack-CACTA elements in Arabidopsis https://academic.oup.com/nar/article/47/3/1311/5198529
A broad survey found gene-capturing "Pack-TYPE" elements across all major DNA transposon superfamilies (MULEs, CACTAs, Harbingers, hATs) https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1010078
"About 70% of C. elegans mRNAs are trans-spliced to one of two 22 nucleotide spliced leaders. SL1 is used to trim off the 5' ends of pre-mRNAs and replace them with the SL1 sequence. This processing event is very closely related to cis-splicing, or intron removal."
The region that is spliced out is called an outron
http://www.wormbook.org/chapters/www_transsplicingoperons/transsplicingoperons.html
Figure from Lasda & Bhatt (2013) https://pmc.ncbi.nlm.nih.gov/articles/PMC3795323/
The leader sequence only affects the untranslated region of the mRNA, so it does not change the resulting protein.
This mechanism has evolved independently 10+ times across very different organisms. In trypanosomes, 100% of mRNAs get a spliced leader added — it is the only way to produce individual mRNAs from their polycistronic transcripts https://royalsocietypublishing.org/doi/10.1098/rsob.190072
In dinoflagellates, all mRNAs also get a spliced leader https://www.pnas.org/doi/10.1073/pnas.0700258104
The tunicate Oikopleura dioica is the only chordate known to use it https://pmc.ncbi.nlm.nih.gov/articles/PMC507004/
Copepod crustaceans use it too (discovered 2015), even though insects and most other arthropods do not https://www.nature.com/articles/srep17411
Also found in flatworms, rotifers, ctenophores, glass sponges, chaetognaths, euglenids, and cryptomonads https://academic.oup.com/mbe/article/27/3/684/1002183
The leader sequences are very similar within a phylum but totally different between phyla, suggesting many independent origins https://www.frontiersin.org/journals/genetics/articles/10.3389/fgene.2013.00199/full
The basal nematode Trichinella spiralis is an outlier — instead of the standard SL1, it has at least 15 diverse non-canonical spliced leaders https://pmc.ncbi.nlm.nih.gov/articles/PMC2271357/
Although prevalent in bacteria, operons are not common in eukaryotes. However, they are common in C. elegans specifically. "A characteristic feature of the worm genome is the existence of genes organized into operons. These polycistronic gene clusters contain two or more closely spaced genes, which are oriented in a head-to-tail direction. They are transcribed as a single polycistronic mRNA and separated into individual mRNAs by the process of trans-splicing"
http://www.wormbook.org/chapters/www_overviewgenestructure.2/genestructure.html
Another paper says "Once considered rare in eukaryotes, polycistronic mRNA expression has been identified in kinetoplastids and, more recently, green algae, red algae, and certain fungi. This study provides comprehensive evidence supporting the existence of polycistronic mRNA expression in the apicomplexan parasite Cryptosporidium parvum"
https://www.biorxiv.org/content/10.1101/2025.01.17.633476v1.full.pdf
In C. elegans, transcripts from opposite strands, eri-6 and eri-7, are joined in trans to make the eri-6/7 mRNA
https://pmc.ncbi.nlm.nih.gov/articles/PMC2756026/
An example from drosophila, C. elegans, and rat shows a gene with a 5' exon being shared between two genes
https://www.fasebj.org/doi/full/10.1096/fj.00-0313rev
An example here shows 5'UTR exons shared across different olfactory receptor genes ("Some OR genes share 5'UTR exons")
https://www.biorxiv.org/content/biorxiv/early/2019/09/19/774612.full.pdf
ANG (angiogenin/RNASE5) and RNASE4 share a non-coding first exon on chr14q11.2, then diverge via alternative splicing into separate coding sequences. Both are RNase A superfamily members; ANG is an angiogenic ribonuclease with ALS-associated mutations. The shared-exon arrangement is conserved in mouse.
https://academic.oup.com/nar/article/33/3/1077/1064173 (Dyer & Rosenberg 2005)
A possible adaptive horizontal gene transfer from bacteria to a eukaryote: the coffee berry borer beetle appears to have picked up a bacterial gene, and domesticating it meant gaining a poly-A tail and losing the Shine-Dalgarno sequence
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3306691/
also https://www.cell.com/cell/fulltext/S0092-8674(19)30097-2
Most organisms use the "standard genetic code", but alternative codes are common across the tree of life — especially in mitochondria, ciliates, and mycoplasma. The NCBI genetic code table lists these variants
https://www.ncbi.nlm.nih.gov/Taxonomy/Utils/wprintgc.cgi#SG31
One article explains how alternative genetic codes work https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6207430/
In vertebrate mitochondria, UGA codes for tryptophan instead of stop, and AGA/AGG code for stop instead of arginine — a direct reversal of the standard code https://www.ncbi.nlm.nih.gov/Taxonomy/Utils/wprintgc.cgi#SG2
The number of known alternative codes keeps growing — NCBI is currently up to table 33, and computational screening of 250,000+ genomes (including metagenome-assembled genomes) has found additional codes not yet officially adopted https://elifesciences.org/articles/71402
"The most common start codons for known Escherichia coli genes are AUG (83% of genes), GUG (14%) and UUG (3%)"
"Here, we systematically quantified translation initiation of green fluorescent protein (GFP) from all 64 codons and nanoluciferase from 12 codons on plasmids designed to interrogate a range of translation initiation conditions."
https://www.sciencedaily.com/releases/2017/02/170221080506.htm
Testing in eukaryotes has also revealed alternative starts being viable https://en.wikipedia.org/wiki/Start_codon#Eukaryotes
3-base codon system is assumed by many, but engineered tRNAs can decode 4-base codons with potential applications for using amino acids outside the 20 canonical ones
review https://elifesciences.org/articles/78869
evolving improved 4-base efficiency https://www.nature.com/articles/s41467-021-25948-y
The standard right-handed double helix is B-DNA, but A-DNA and left-handed Z-DNA also form. "There are also triple-stranded DNA forms and quadruplex forms such as the G-quadruplex and the i-motif." https://en.wikipedia.org/wiki/Nucleic_acid_double_helix https://en.wikipedia.org/wiki/Triple-stranded_DNA
Seen in trypanosomes. The kinetoplast DNA network contains two types of circles: maxicircles (~20-40 kb) encode mitochondrial genes, but many of their transcripts are "cryptogenes" — essentially unreadable without extensive RNA editing. The guide RNAs that direct this editing are encoded on thousands of minicircles (~1 kb each) in the same network. See also the RNA editing section.
Figure from https://www.sciencedirect.com/topics/agricultural-and-biological-sciences/kinetoplast-dna
https://pmc.ncbi.nlm.nih.gov/articles/PMC4835692/
Mitochondrial genomes are usually a single circular (or linear) molecule, but in dinoflagellates they are broken into multiple small linear chromosomes (~6-10 kb each). Only 3 protein-coding genes remain (cox1, cox3, cob), and even those are split into fragments that must be separately transcribed and pieced together https://academic.oup.com/mbe/article/24/7/1528/987488
Some cells — famously insect salivary glands — copy their chromosomes through repeated rounds of replication without cell division, giving giant banded chromosomes https://en.wikipedia.org/wiki/Polytene_chromosome
Figure source https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5768140/
"Polytene chromosomes are produced by endoreplication, in which chromosomal DNA undergoes repeated replication without cell division. Ten rounds of endoreplication produces 2^10 = 1,024 DNA strands, which when arranged alongside of each other produce distinctive banding patterns. Endoreplication occurs in cells of the larval salivary glands of many species of Diptera, and increases production of mRNA for Glue Protein that the larvae use to anchor themselves to the walls of (for example) culture vials." from https://www.mun.ca/biology/scarr/Polytene_Chromosomes.html
Endoreplication is not confined to salivary glands. Human liver and placenta carry polyploid cells in an otherwise diploid organism, which is called endopolyploidy https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4442802/
It also shows up as a genome assembly problem: a PAG talk on the vanilla bean genome reported heavy endoreplication driving very uneven sequencing coverage, so the team picked the tissue samples with the least of it https://plan.core-apps.com/pag_2023/abstract/e26dbeb1-df8f-4c57-a062-dcaf881b79f4
"While we commonly assume the genome to be largely identical across different cells of a multicellular organism, a number of species undergo a developmental regulated elimination process by which the genome in somatic cells is reduced, while the germline genome remains intact. This process, called Programmed DNA Elimination (PDE), affects a number of species including copepod crustaceans, lamprey fish, single-celled ciliates and nematode worms (though not C. elegans!)."
From ISMB2023 video "Deciphering developmentally programmed DNA elimination in Mesorhabditis nematodes" https://www.youtube.com/watch?v=2x6ElKeISRY
Ciliates are the most extreme case — spirotrich ciliates like Oxytricha can eliminate over 95% of their germline genome when building the somatic macronucleus, including virtually all transposons. Even in Tetrahymena and Paramecium, 20-30% of the germline genome is discarded https://pmc.ncbi.nlm.nih.gov/articles/PMC3839606/
Figure from Noto & Mochizuki (2017) https://pmc.ncbi.nlm.nih.gov/articles/PMC5666084/
See also the term "internal eliminated sequences" (IES)
Germline-restricted chromosomes (GRCs) are the same trick applied to a whole chromosome: accessory chromosomes that occur only in germ cells and are eliminated from somatic cells during embryo development, seen in several unrelated animal taxa and inherited in strikingly non-Mendelian ways https://pmc.ncbi.nlm.nih.gov/articles/PMC9508068/
PDE, GRCs, VDJ recombination and bacterial CRISPR spacer acquisition are all cases of organisms editing their own genomes, reviewed in "How and when organisms edit their own genomes" https://www.nature.com/articles/s41588-025-02230-1
Birds and most reptiles have a bimodal karyotype: a handful of large macrochromosomes plus 20-30 tiny microchromosomes. Chicken is 2n=78, roughly 10 macro and 28 micro pairs. Microchromosomes are gene dense and GC rich, and they recombine and evolve faster than the macrochromosomes. They look like the ancestral vertebrate state, with mammalian lineages having fused most of them into larger chromosomes.
https://www.pnas.org/doi/10.1073/pnas.2112494118
Short-read bird assemblies long left many microchromosomes fragmented or missing entirely, so a "complete" genome could quietly omit a chunk of the gene catalog.
Supernumerary chromosomes that some individuals of a species carry and others do not, known from thousands of plants, animals, and fungi. Most are heterochromatic and gene poor, and many encode drive mechanisms that push their own transmission above the Mendelian 50%.
Variable B counts inside a species are a practical trap for assembly: sequence present in one individual and absent in another inflates apparent heterozygosity and leaves regions collapsed or duplicated.
https://en.wikipedia.org/wiki/B_chromosome https://doi.org/10.1007/s00018-013-1437-7
Wikipedia's table of ploidy levels runs from haploid (male European fire ants) through the familiar triploid seedless watermelons and hexaploid wheat, out to tetratetracontaploid — forty-four sets — in black mulberry https://en.wikipedia.org/wiki/Polyploidy#Types
There are many chemical modifications that can happen to DNA, leading to an "extended alphabet" with functional changes.
The most common is 5mC, a methylated cytosine, usually in a CpG (a C followed by a G). Others include 6mA and oxidized derivatives of 5mC such as 5hmC
Many other modifications exist, see https://dnamod.hoffmanlab.org/
RNA editing changes the mRNA after transcription, so the sequenced RNA can disagree with the genome. A-to-I editing is common in some species and reads out as an A->G difference in RNA-seq with no corresponding genomic SNP.
Editing can also be conditional and tissue specific — see apolipoprotein B in the stop codons created by RNA editing section
Other editing also occurs https://en.wikipedia.org/wiki/RNA_editing
In trypanosome mitochondria, extensive U-insertion/deletion editing can more than double the length of some transcripts — the genomic sequence is essentially a compressed template that is "decoded" by editing. The instructions for editing come from guide RNAs encoded on minicircles in the kinetoplast DNA (see maxicircle architecture section). https://en.wikipedia.org/wiki/RNA_editing#Kinetoplastids
While the exon structure of most mRNAs follows the linear sequence of the transcribed DNA, there are a few cases where mature mRNAs contain exons in a non-linear order.
Al-Balool et al. (2011) validated several cases of PTES in human genes that are evolutionarily conserved, including MAN1A2, PHC3, TLE4, and CDK13: https://genome.cshlp.org/content/21/11/1788.short
The lncRNA VELUCT nearly escaped a lung cancer screen because its expression is so low that it is "below the detection limit in total RNA from NCI-H460 cells by RT-qPCR as well as RNA-Seq", however this study confirms it as a factor in experiments (2025 note: this article has basically no citations, unclear whether it actually is impactful)
https://www.ncbi.nlm.nih.gov/pubmed/28160600?dopt=Abstract
Note that X inactivation relies on relatively lowly expressed RNA also https://twitter.com/mitchguttman/status/1454256452990734336
X inactivation is a form of dosage compensation — equalizing X-linked gene expression between XX and XY individuals. In mammals, one X in each female cell is silenced early in development by Xist, a long non-coding RNA that spreads along the chromosome in cis, recruiting silencing complexes. An antisense transcript Tsix represses Xist on the active X.
Not all genes on the inactive X are silenced — about 15-25% of human X-linked genes "escape" inactivation and are expressed from both copies. This is relevant to bioinformatics because these escapee genes show biallelic expression in females where you might expect monoallelic.
https://en.wikipedia.org/wiki/X-inactivation
https://www.youtube.com/watch?v=y3ST0whbA4k (great series from iBiology on X chromosome inactivation)
See https://github.com/cmdcolin/oddbiology/ for calico cats, mosaicism, and how different organisms solve dosage compensation in completely different ways
tmRNA rescues bacterial ribosomes stalled on a truncated mRNA. One molecule carries both a tRNA-like and an mRNA-like domain: it enters the ribosome as a tRNA charged with alanine, and then the ribosome switches templates onto tmRNA's own short ORF, which tags the stuck protein for degradation and supplies the stop codon the broken message was missing. A single feature that is a tRNA and an mRNA at once is not what most annotation formats expect.
https://en.wikipedia.org/wiki/Transfer-messenger_RNA
Nearly every gene starts with a codon for methionine, but methionyl aminopeptidase often removes that methionine after translation https://en.m.wikipedia.org/wiki/Methionyl_aminopeptidase
Plenty of transcripts annotated as lncRNAs encode functional peptides under 100 amino acids, missed because annotation pipelines discard ORFs below an arbitrary length cutoff. MOTS-c is stranger still: it is encoded inside the mitochondrial 12S rRNA gene, translated in the mitochondrion, and acts in the nucleus to regulate the metabolic stress response.
https://doi.org/10.1016/j.cmet.2015.02.009 (MOTS-c) https://doi.org/10.1038/nrm.2017.58 (review)
An intein is a segment of a protein that catalyzes its own excision after translation, joining the flanking segments (exteins) together. Like introns for RNA, but at the protein level. https://en.wikipedia.org/wiki/Intein
See section here https://github.com/The-Sequence-Ontology/Specifications/blob/master/gff3.md#pathological-cases
Many viruses translate one long polyprotein and then cut it into pieces with a protease. Coronaviruses reach the second half of theirs by ribosomal frameshifting. Dengue and HIV use the same strategy
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6040172/ https://www.sciencedirect.com/science/article/abs/pii/S0959440X15000597
Some viruses use an alternative mechanism: 2A peptides cause the ribosome to skip forming a peptide bond during translation, effectively splitting the polyprotein without a protease. These are now widely used as a biotechnology tool for expressing multiple proteins from one mRNA https://pmc.ncbi.nlm.nih.gov/articles/PMC5438344/
Transposable elements make up ~45% of the human genome (https://www.nature.com/articles/nature01262) and up to 85% of some plant genomes like maize (https://www.science.org/doi/10.1126/science.1178534).
Or "How a quarter of the cow genome came from snakes" http://phenomena.nationalgeographic.com/2013/01/01/how-a-quarter-of-the-cow-genome-came-from-snakes/
Source http://www.pnas.org/content/110/3/1012.full
The host normally silences LINE1 elements, but LINE1 transcription turns out to matter for embryonic development.
https://www.ucsf.edu/news/2018/06/410781/not-junk-jumping-gene-critical-early-embryo
The RAG1/RAG2 recombinase behind V(D)J recombination (see Immunity) derives from an ancient Transib transposon, which makes the vertebrate adaptive immune system a domesticated transposon https://www.nature.com/articles/nature03564
CENPB, which binds centromeric DNA, derives from a pogo-like transposase https://www.nature.com/articles/384567a0
See also https://github.com/cmdcolin/oddbiology/ for syncytins and Arc (retroviral elements co-opted for placental development and neuronal communication)
VDJ recombination is a process of somatic DNA recombination in immune cells.
In bio 101 you learn: a gene is made of exons separated by introns, the cell transcribes the gene into RNA, introns are spliced out, and the exons are joined together in the mRNA. The DNA itself is never changed — every cell in your body has the same genome.
VDJ recombination breaks this rule. The genome contains arrays of V (variable), D (diversity), and J (joining) gene segments — many copies of each type, arranged in tandem. Instead of just transcribing and splicing RNA, immune cells physically cut the chromosomal DNA, join one V to one D to one J segment, and delete the intervening DNA. Which segments end up joined is essentially random. This is a permanent, irreversible change to the genome of that cell. The result is a brand-new exon that didn't exist in the germline genome. Only then does normal RNA splicing happen to join this newly assembled VDJ exon to the constant region exons downstream — as Janeway's Immunobiology puts it: "To make a complete immunoglobulin light-chain messenger RNA, the V-region exon is joined to the C-region sequence by RNA splicing after transcription."
Figure from https://commons.wikimedia.org/wiki/File:VDJ_recombination.png
This is why V, D, and J are called "gene segments" rather than "exons" — they are raw parts that must be assembled at the DNA level first, before the normal exon-intron splicing machinery can do its job.
Flanking "recombination signal sequences" guide the joining, which is deliberately imprecise — terminal transferase adds random nucleotides at the junctions. A few dozen V, roughly two dozen D, and a handful of J segments per locus therefore generate billions of different antibodies.
The rearrangement is called "somatic" because it happens in developing B and T cells rather than the germline, so it is never passed to offspring. Elsewhere in biology, recombination is a meiotic affair between homologous chromosomes. Here a dedicated RAG1/RAG2 complex, itself derived from an ancient transposon (see Transib above), cuts at the recombination signal sequences and leaves the broken ends to the cell's general-purpose NHEJ (non-homologous end joining) repair pathway — RAG supplies the targeted double-strand breaks and existing repair machinery does the rest. Only lymphocytes express RAG, which is why only they rearrange.
https://en.wikipedia.org/wiki/V(D)J_recombination https://www.ncbi.nlm.nih.gov/books/NBK27140/
The MHC on chr6 is the most polymorphic region of the human genome — diverse enough that hg38 ships extra alternative loci/contigs just to represent it
The "T2T" assembly of the maize Mo17 genome contains 235kb of nothing but TAG, called out in the abstract because it is that silly
https://www.nature.com/articles/s41588-023-01419-6
A tandem duplication is a stretch of DNA copied immediately next to itself. Where does it come from?
Mechanisms include
- replication slippage
- retrotransposition
- unequal crossing over (UCO)
- imperfect repair of double-strand breaks by nonhomologous end joining (NHEJ) (specifically generates 1-100bp range indels according to article)
Ref https://academic.oup.com/mbe/article/24/5/1190/1038942
Segmental duplications (SDs) are blocks of DNA (1 kb to hundreds of kb) present in two or more copies with >90% sequence identity. About 5-7% of the human genome is segmentally duplicated. They are a major headache for genome assembly — high identity causes collapsed or missing regions. T2T-CHM13 increased known SD content from 167 Mbp (5.4%) to 218 Mbp (7.0%), accounting for nearly one-third of newly assembled sequence.
https://www.science.org/doi/10.1126/science.abj6965
SDs serve as substrates for generating copy number variants (CNVs) via non-allelic homologous recombination (NAHR): meiotic recombination between paralogous copies instead of true alleles produces deletions, duplications, or inversions. Other CNV mechanisms include NHEJ, FoSTeS/MMBIR (replication-based template switching), and BFB cycles.
https://www.nature.com/articles/nrg2593
Medically important examples of NAHR between flanking SDs:
- 22q11.2 deletion (DiGeorge) — most common human microdeletion (~1:3,000) https://pmc.ncbi.nlm.nih.gov/articles/PMC1310636/
- CMT1A / HNPP at 17p12 — 1.5 Mb duplication (CMT1A) or reciprocal deletion (HNPP) affecting PMP22 dosage https://omim.org/entry/118220
- SMA — SMN1 loss in one of the most complex SD regions; most assemblies still don't correctly resolve SMN1/SMN2 https://pubmed.ncbi.nlm.nih.gov/15470363/
- Prader-Willi / Angelman at 15q11-q13 — same deletion, different syndrome depending on parent-of-origin (imprinting) https://pmc.ncbi.nlm.nih.gov/articles/PMC4449422/
Cancer genomes harbor catastrophic rearrangements beyond simple deletions and duplications:
- Chromothripsis — chromosome shattering and random reassembly in a single event, in >50% of some cancer types https://www.nature.com/articles/s41588-019-0576-7
- Chromoplexy — chains of balanced translocations across up to eight chromosomes, prevalent in prostate cancer (>60%) https://pmc.ncbi.nlm.nih.gov/articles/PMC3673705/
- Chromoanasynthesis — replication-based template switching producing localized duplications/triplications https://www.nature.com/articles/nm.2988
- Extrachromosomal DNA (ecDNA) — circular DNA (~1-5 Mb) replicating autonomously, driving oncogene amplification (EGFR, MYC) in ~17% of cancers. Lacks centromeres so segregates unequally https://www.nature.com/articles/s41576-022-00521-5
- Breakage-fusion-bridge (BFB) cycles — telomere loss causes sister chromatid fusion, anaphase bridge, and breakage, repeating to produce staircase amplification. Can trigger chromothripsis and ecDNA https://www.nature.com/articles/s41467-023-41259-w
- Kataegis — localized hypermutation clusters caused by APOBEC deaminases on single-stranded DNA at break sites https://elifesciences.org/articles/00534
- Tyfonas, Pyrgo, Rigma — novel complex SV classes: tyfonas (fold-back inversions, acral melanoma), pyrgo (tandem duplications, breast/ovarian), rigma (deletions at fragile sites, GI cancers) https://www.cell.com/cell/fulltext/S0092-8674(20)30997-1
Elephants carry many LIF copies, most of them pseudogenes. One of them, LIF6, is switched back on by p53 and kills damaged cells — a "zombie gene" that may help explain why such a large animal gets so little cancer
https://www.cell.com/cell-reports/fulltext/S2211-1247(18)31145-8
https://www.sciencealert.com/lif6-pseudogene-elephant-tumour-suppression-solution-petos-paradox
When a retrotransposon reverse-transcribes an mRNA and pastes the copy back into the genome, that copy has no introns and usually lands nowhere useful — a processed pseudogene. Some pick up a promoter and become functional retrogenes with an expression pattern of their own. GLUD2 is an intronless retrocopy of GLUD1 that landed on the X chromosome in the hominoid lineage and is now expressed in brain and testis. Paralogs, in other words, need not share an intron-exon structure.
https://doi.org/10.1038/nrg2487 https://doi.org/10.1038/ng1431 (GLUD2)
Some intron sequences boost expression much the way promoters do https://en.wikipedia.org/wiki/Intron-mediated_enhancement
The first intron of the UBQ10 gene in Arabidopsis exhibits IME, and "the sequences responsible for increasing mRNA accumulation are redundant and dispersed throughout the UBQ10 intron" http://www.plantcell.org/content/early/2017/04/03/tpc.17.00020.full.pdf+html
The peppered moth's industrial melanism allele is a transposon insertion in an intron of the cortex gene https://www.nature.com/articles/nature17951 (may not be strictly IME, I'm personally not sure)
Imprinted genes are expressed from only one parental allele. In humans, ~100-200 genes are imprinted. Classic examples: IGF2 (paternal only) and H19 (maternal only) share a regulatory region on chr11. The same chromosomal deletion at 15q11-q13 causes Prader-Willi syndrome if inherited from dad, or Angelman syndrome if from mom.
https://en.wikipedia.org/wiki/Genomic_imprinting
Wikipedia https://en.wikipedia.org/wiki/Promoter_(genetics)#Bidirectional_(mammalian)
"Bidirectional promoters are a common feature of mammalian genomes. About 11% of human genes are bidirectionally paired."
"The two genes are often functionally related, and modification of their shared promoter region allows them to be co-regulated and thus co-expressed"
See also 'divergently paired genes' https://academic.oup.com/gpb/article/23/4/qzaf058/8174973?login=false&utm_source=etoc&utm_campaign=gpb&utm_medium=email&nbd_source=campaigner&nbd=45398946253
TL;DR — every XX sample carries reads on chrY. NA12878, a karyotypically XX
female, has 884,578 of them against an XY sample's 6,118,277: 14% of a real Y
rather than zero, and 190,158 of hers survive a MAPQ >= 30 filter. They come
from sequence chrY shares with the rest of the genome — pseudoautosomal regions
byte-identical to chrX, the X-transposed region at ~99% identity to Xq21, and
beta satellite arrays. The measurement that does separate the two is depth
across the male-specific region: 0.48x the autosomal mean for XY against 0.015x
for XX.
Reads mapped to chrY are not evidence of a Y chromosome. NA12878 is a karyotypically XX female sequenced to ~35x in the 1000 Genomes 30x collection, aligned against an analysis-set GRCh38 that already hard-masks PAR on chrY.
Pick the wrong window and she is still indistinguishable from an XY sample. Give both samples a reference that keeps chrY PAR and their PAR1 depth lands on top of itself at one copy; inside a beta satellite array both pile up past 100x the autosomal mean. Only sequence unique to a real Y pulls them apart.
Three separate things put reads on chrY in an XX sample, and the ideogram shows where each of them sits.
The pseudoautosomal regions are the same sequence twice. PAR1 and PAR2 sit on both chrX and chrY, and in hg38 they are not merely similar, they are byte-identical — 2,771,479 bp of PAR1 and 329,513 bp of PAR2 compared, zero mismatching bases.
samtools faidx hg38.fa chrX:10001-2781479 | tail -n+2 | tr -d '\n' > parX
samtools faidx hg38.fa chrY:10001-2781479 | tail -n+2 | tr -d '\n' > parY
cmp parX parY # identicalNo aligner can split that. Pulling NA12878's PAR1 reads back out and remapping them to an hg38 that keeps both copies, bwa-mem puts 286,511 on chrX and 284,437 on chrY and gives 99.93% of them MAPQ 0. The XY sample splits the same way, 362,977 and 361,699 with 99.92% at MAPQ 0. Half of either sample's PAR reads land on chrY at full single-copy depth, so in PAR the two are the same picture.
Analysis-set references dodge this by hard-masking PAR on chrY. Uppercase the
UCSC hg38 chrY, N out chrY:10,001-2,781,479 and chrY:56,887,903-57,217,415, and
you reproduce the 1000 Genomes chrY byte for byte (md5
ce3e31103314a704255f3cd90369ecce). A pipeline pointed at a plain hg38.fa
gets no such protection.
The X-transposed region is ~99% identical to Xq21. Mapping hg38 chrY back onto chrX in 1 kb windows finds it as a single block, chrY:3,050,000-6,234,000 against chrX around 89.4 Mb, left by a human-specific transposition of Xq21 to Yp11.2. Masking PAR does nothing here.
Satellite arrays collect strays from the whole genome. 90% of NA12878's chrY
reads (797,492 of 884,578) sit in two blocks — chrY:10.0-12.5 Mb around the
centromere and chrY:56.67-56.89 Mb just proximal to PAR2 — both beta satellite
and (ATTCC)n arrays. The mates of those reads point at acrocentric satellite
contigs such as chr14_GL000225v1_random and chrUn_GL000216v2. A MAPQ filter
does not clean this up, because the array the reads came from is not in the
reference at all and bwa scores the surviving copy confidently: 96% of her
MAPQ >= 30 chrY reads are in these same two blocks.
What separates XX from XY is depth across the male-specific region, not the presence of chrY reads. Outside the satellite blocks NA12889 (XY) holds 0.48x the autosomal mean flat across the whole MSY, while NA12878 sits at 0.015x, and 0.0014x once low-quality alignments are dropped. Artifact coverage is spiky and local; a real Y is a continuous single-copy plateau.
https://academic.oup.com/gigascience/article/8/7/giz074/5530326
https://lh3.github.io/2017/11/13/which-human-reference-genome-to-use
A child can inherit both copies of a particular chromosome from one parent, instead of the "usual" one copy from mom, one from dad
"UPD arises usually from the failure of the two members of a chromosome pair to separate properly into two daughter cells during meiosis in the parent’s germline (nondisjunction). The resulting abnormal gametes contain either two copies of a chromosome (disomic) or no copy of that chromosome (nullisomic), instead of the normal single copy of each chromosome (haploid). This leads to a conception with either three copies of one chromosome (trisomy) or a single copy of a chromosome (monosomy). If a second event occurs by either the loss of one of the extra chromosomes in a trisomy or the duplication of the single chromosome in a monosomy, the karyotypically normal cell may have a growth advantage as compared to the aneuploid cells. UPD results primarily from one of these “rescue” events"
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3111049/
Older men can have a mosaic loss of the Y chromosome https://en.wikipedia.org/wiki/Mosaic_loss_of_chromosome_Y
https://www.karger.com/Article/FullText/508564 (found from https://www.biostars.org/p/9482437/)
may be associated with cardiac issues https://www.science.org/doi/10.1126/science.abn3100
The same happens with the X in older women https://www.cancer.gov/news-events/press-releases/2024/genetic-factors-predict-x-chromosome-loss
In organisms with normally linear chromosomes, circular or "ring" chromosomes can form from aberrant processes https://en.wikipedia.org/wiki/Ring_chromosome
Figure from the National Library of Medicine https://commons.wikimedia.org/wiki/File:NLM_ring_chromosome.jpg
There are also smaller fragments that can be circularized called "supernumerary small ring chromosomes" (sSRC) or their linear counterparts, "supernumerary small marker chromosomes" (sSMC) https://en.wikipedia.org/wiki/Small_supernumerary_marker_chromosome
Half the formats in bioinformatics count positions from 0 and half count from 1, and they also disagree about whether the end coordinate belongs to the feature.
- 0-based, half-open: BED, BAM, BCF, 2bit, and the htslib/pysam APIs
- 1-based, closed: GFF3, GTF, SAM, VCF, wig, and the locstrings people type into genome browsers
The same feature is chr1 100 200 in BED and chr1 101 200 in GFF. Half-open
arithmetic is the nicer one to compute with — length is just end - start, and
abutting features share a number — but write it out as though it were 1-based
and every start shifts by one, an error small enough to survive review and large
enough to change a codon. samtools hands you both conventions at once: the
region strings on its command line are 1-based and closed, while the C API
underneath is 0-based and half-open.
https://genome-blog.gi.ucsc.edu/blog/2016/12/12/the-ucsc-genome-browser-coordinate-counting-systems/
R makes it funnier. R indexes vectors from 1, so Bioconductor's GRanges is 1-based closed throughout and lines up with GFF rather than BED. rtracklayer's import() quietly adds 1 to every BED start and export() takes it back off again, so the conversion is invisible right until you combine an imported GRanges with coordinates you parsed yourself. Python, indexing from 0 and slicing half-open, agrees with BED instead — so the same logic ported between the two languages ends up off by one in whichever direction nobody checked.
https://bioconductor.org/packages/release/bioc/html/rtracklayer.html
The human genome as distributed by NCBI contains IUPAC ambiguity codes beyond ACGTN https://www.bioinformatics.org/sms/iupac.html
Here is the incidence of non-ACGTN IUPAC letters in the entire human genome GRCh38.p14 from https://ftp.ncbi.nlm.nih.gov/refseq/H_sapiens/annotation/GRCh38_latest/refseq_identifiers/GRCh38_latest_genomic.fna.gz (same for the "analysis set" files in https://ftp.ncbi.nlm.nih.gov/genomes/all/GCA/000/001/405/GCA_000001405.15_GRCh38/seqs_for_alignment_pipelines.ucsc_ids/)
{
'B' => 2,
'K' => 8,
'Y' => 36,
'M' => 8,
'R' => 29,
'W' => 15,
'S' => 5
};
Did you expect that in your bioinformatics software? Note that the mouse genome (GRCm38.p5) as far as I could tell does not contain any non-ACGT IUPAC letters
See count_fasta_letters.pl for a script to count this. The UCSC hg38.fa.gz does not have any non-ACGTN letters.
Due to how dbSNP is created (based on alignments), an rs ID can occur in multiple places on the genome https://www.biostars.org/p/2323/
In response to hg38 including a colon in sequence names, which conflicts with commonly used representation of a range as chr1:1-100 for example (note: SAMv1.pdf contains a regex to help resolve this), people analyzed meta-character frequencies in sequence names samtools/hts-specs#291
ENA
# 16927
* 1
, 231
- 122563947
. 521540419
/ 236951
\ 0
: 30181
; 72892
= 186611
@ 3713
| 949
Reference sequences
# 203
% 203
* 525
+ 1
, 496
- 154226
. 1826561
: 1577
= 26
_ 4961932
| 1098333
There is a proposal to make commas illegal in FASTA names, since the SAM/BAM supplementary alignment tag (SA) is itself comma separated
Chromosomes keep outgrowing the formats built to hold them. Wheat chromosomes average 806Mbp, past the 2^29-1 (~536Mbp) ceiling that the BAI/TBI binning strategy imposes, which is why the CSI index format exists.
The axolotl goes further: individual chromosomes of 3.14 Gbp https://genome.cshlp.org/content/29/2/317.long (2019), almost the size of the whole human genome, and past the 2^31-1 (~2.14Gbp) limit on chromosome length in BAM and CRAM, so bgzip/tabix SAM gets used instead (discussion samtools/hts-specs#655)
Bonus: I made a web tool to visualize BAI files and show how the binning index works https://cmdcolin.github.io/bam_index_visualizer/
Just some honorable mentions for largest genome
- Polychaos dubium (formerly Amoeba dubia) - ~600-1300Gbp (unsequenced, 1968 back of envelope measurement, needs confirmation) https://en.wikipedia.org/wiki/Polychaos_dubium (another ref https://bionumbers.hms.harvard.edu/bionumber.aspx?&id=117342)
- Dinoflagellates - up to 250Gbp (unsequenced, 1987 book referenced in this paper, needs confirmation, has weird chromosome "rod-like" structures) https://www.nature.com/articles/s41588-021-00841-y
- Tmesipteris oblanceolata (fork fern) - ~160Gb (unsequenced) https://www.nature.com/articles/d41586-024-01567-7
- Paris japonica (Japanese woodland herb) - ~149Gbp (unsequenced) https://en.wikipedia.org/wiki/Paris_japonica
- Tmesipteris obliqua (fern) - ~147Gbp (unsequenced) - https://en.wikipedia.org/wiki/Tmesipteris_obliqua
- South American lungfishes (Lepidosiren paradoxa) - ~91Gbp (sequenced) https://www.nature.com/articles/s41586-024-07830-1
- European mistletoe - ~90Gbp (sequenced) https://www.darwintreeoflife.org/news_item/2022-the-year-we-built-the-biggest-genome-in-britain-and-ireland/
- Antarctic krill - ~48Gbp (sequenced) https://www.cell.com/cell/pdf/S0092-8674(23)00107-1.pdf
- Neoceratodus forsteri (Australian lungfish) - ~43Gbp (sequenced) https://www.smithsonianmag.com/smart-news/australian-lungfish-has-biggest-genome-ever-sequenced-180976837/ https://www.ncbi.nlm.nih.gov/genome/?term=Neoceratodus+forsteri
- Ambystoma mexicanum (axolotl) - ~32Gbp (sequenced) https://en.wikipedia.org/wiki/Axolotl https://www.ncbi.nlm.nih.gov/genome/?term=axolotl
- Allium ursinum (wild garlic) - ~30Gbp https://en.wikipedia.org/wiki/Onion_Test
Inspired by twitter thread https://twitter.com/PetrovADmitri/status/1506824610360168455
Also see http://www.genomesize.com/statistics.php?stats=entire#stats_top
See also the plant C-value database, which is a measurement you will sometimes see instead of base pair length https://cvalues.science.kew.org/ ("C-value is the amount, in picograms, of DNA contained within a haploid nucleus")
On the other end of the scale, some obligate endosymbionts have extremely reduced genomes:
- Nasuia deltocephalinicola — ~112 kb, the smallest known cellular genome https://academic.oup.com/gbe/article/5/9/1675/555845
- Carsonella ruddii — ~160 kb https://www.science.org/doi/10.1126/science.1134196
- Mycoplasma genitalium — ~580 kb, used as the basis for "minimal genome" studies https://www.science.org/doi/10.1126/science.270.5235.397
Meanwhile, some giant viruses have genomes larger than these bacteria — Pandoraviruses reach ~2.5 Mb, blurring the line between viruses and cellular life https://www.science.org/doi/10.1126/science.1239181
The CG tag stores CIGAR strings longer than 64k operations, since BAM holds n_cigar_op in 16 bits. This concerns BAM only — CRAM stores the same information differently, via reference based compression.
As of Dec 2023 I extracted all the genes from a number of model organism databases here https://cmdcolin.github.io/genes/
Some highlights from earlier work
- Tinman - "In mutant or knockout organisms, the loss of tinman results in the lack of heart formation" https://en.wikipedia.org/wiki/Tinman_gene
- Sonic hedgehog (SHH) - named after the video game character; hedgehog mutants have 'spiky' fruit fly embryos https://en.wikipedia.org/wiki/Sonic_hedgehog
- Robotnikin - a small molecule that blocks Shh signaling, so not a gene at all, named for Dr. Robotnik, the villain of the Sonic the Hedgehog franchise https://pmc.ncbi.nlm.nih.gov/articles/PMC2770933/
- Heart of glass (heg) - a zebrafish gene with mutant phenotype "Individual heg myocardial cells are also thinner than wild-type" https://www.ncbi.nlm.nih.gov/pubmed/14680629
- Dracula (drc) - "we isolated a mutation, dracula (drc), which manifested as a light-dependent lysis of red blood cells" https://www.ncbi.nlm.nih.gov/pubmed/10985389 (now renamed https://zfin.org/ZDB-GENE-000928-1)
- Sleeping Beauty transposon system - https://en.wikipedia.org/wiki/Sleeping_Beauty_transposon_system
- Skywalker (sky) - https://www.ncbi.nlm.nih.gov/gene?Db=gene&Cmd=DetailsSearch&Term=35359
- TIME FOR COFFEE (TIC) - "We characterize the time for coffee (tic) mutant that disrupts circadian gating, photoperiodism, and multiple circadian rhythms, with differential effects among rhythms" https://www.ncbi.nlm.nih.gov/gene?Db=gene&Cmd=DetailsSearch&Term=821807
- WTF - "Some alleles of the wtf gene family can increase their chances of spreading by using poisons to kill other alleles, and antidotes to save themselves." - https://www.ebi.ac.uk/interpro/entry/IPR004982 https://www.sciencedaily.com/releases/2017/06/170620093209.htm
- Mothers against decapentaplegic - "it was found that a mutation in the gene in the mother repressed the gene decapentaplegic in the embryo. The phrase "Mothers against" was added as a humorous take-off" https://en.wikipedia.org/wiki/Mothers_against_decapentaplegic
- Saxophone (sax) - http://www.sdbonline.org/sites/fly/gene/saxophon.htm
- Beethoven (btv) - http://www.uniprot.org/uniprot/Q0E8P6
- Superman+kryptonite - https://en.wikipedia.org/wiki/Superman_(gene)
- Supervillin (SVIL) - https://www.uniprot.org/uniprot/O95425
- Wishful thinking (wit) - https://www.wikigenes.org/e/gene/e/44096.html
- Doublesex (dsx) - "The gene is expressed in both male and female flies and is subject to alternative splicing, producing the protein isoforms dsx_f in females and the longer dsx_m in males." https://en.wikipedia.org/wiki/Doublesex
- Fruitless (fru) - "Early work refers to the gene as fruity, an apparent pun on both the common name of D. melanogaster, the fruit fly, as well as a slang word for homosexual. As social attitudes towards homosexuality changed, fruity came to be regarded as offensive, or at best, not politically correct. Thus, the gene was re-dubbed fruitless, alluding to the lack of offspring produced by flies with the mutation.[10] However, despite the original name and a continuing history of misleading inferences by the popular media, fruitless mutants primarily show defects in male-female courtship, though certain mutants cause male-male or female-female courtship.[11]" https://en.wikipedia.org/wiki/Fruitless_(gene)
- Transformer (tra) - https://en.wikipedia.org/wiki/Transformer_(gene)
- Gypsy+Flamenco - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1206375/ also described in wiki https://en.wikipedia.org/wiki/Piwi-interacting_RNA#History_and_loci
- Jockey - http://flybase.org/reports/FBgn0015952.html
- Tigger - https://www.omim.org/entry/612972
- Nanog - named for Tír na nÓg, the land of the ever-young in Irish legend https://www.sciencedaily.com/releases/2003/06/030602024530.htm (source https://twitter.com/EpgntxEinstein/status/1057359656220348417)
- Jerky (jrk) - "A deficit in the Jerky protein in mice causes recurrent seizures" https://www.genecards.org/cgi-bin/carddisp.pl?gene=JRK
- Hippo (Hpo) - https://www.wikigenes.org/e/gene/e/37247.html
- Dishevelled (Dsh) - https://en.wikipedia.org/wiki/Dishevelled
- Glass bottom boat (gbb) - "fruit fly larvae with a faulty glass bottom boat gene are transparent" https://www.thenakedscientists.com/articles/interviews/gene-month-glass-bottom-boat http://www.sdbonline.org/sites/fly/dbzhnsky/60a-1.htm
- Makes caterpillars floppy (mcf) - https://www.pnas.org/content/99/16/10742 (source https://twitter.com/JUNIUS_64/status/1081007886560608256)
- Eyeless http://flybase.org/reports/FBgn0005558.html
- Straightjacket (stj) - http://flybase.org/reports/FBgn0261041.html
- Huluwa http://science.sciencemag.org/content/362/6417/eaat1045 ref https://twitter.com/zhouwanding/status/1065960714978897921
- frameshifts or pseudogene? - check sequence - https://www.ncbi.nlm.nih.gov/gene/?term=24562233%5Buid%5D
- Bad response to refrigeration (brr) https://twitter.com/hitenmadhani/status/1149471071675924481?s=20
- Mindbomb (mib1) - https://www.sdbonline.org/sites/fly/hjmuller/mindbomb1.htm
- β'COP http://flybase.org/reports/FBgn0025724.html (https://twitter.com/DarrenObbard/status/1260613447198412800)
- King-tubby https://www.uniprot.org/uniprot/B0XFQ9 see also tubby https://www.uniprot.org/uniprot/P50586
- fucK https://www.uniprot.org/uniprot/?query=fuck&sort=score
- scarecrow https://academic.oup.com/g3journal/article/15/5/jkaf055/8071394?login=false&utm_source=etoc&utm_campaign=g3journal&utm_medium=email
- Halloween genes (spook, disembodied, phantom, shadow, shade) https://en.wikipedia.org/wiki/Halloween_genes
- VANDAL21 https://www.arabidopsis.org/servlets/TairObject?type=transposon_family&id=139
- HotDog domain - superfamily of genes/proteins https://www.wikidata.org/wiki/Q24785143 https://www.ebi.ac.uk/interpro/entry/IPR029069
- Flower/fwe - https://flybase.org/reports/FBgn0261722.html
- Brahma https://www.sdbonline.org/sites/fly/polycomb/brahma.htm
- Pokemon gene - "The Pokémon Company threatened MSKCC with legal action in December 2005 for creating an association between cancer and the media franchise, and as a consequence MSKCC is now referring to it by its gene name Zbtb7" - Pokemon/pikachu/zubat (story https://bsky.app/profile/c0nc0rdance.bsky.social/post/3k6w3gwtell2j)
- Bring lots of money (blom7α) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2781463/ https://www.uniprot.org/uniprotkb/Q7Z7F0/entry
- MAGOH - Drosophila flies produce unfit progeny when they have mutations in their mago nashi (Japanese: 孫なし, Hepburn: mago nashi, lit. 'grandchildless') gene. The progeny have defects in germplasm assembly and germline development https://www.uniprot.org/uniprotkb/P61326/entry
- IGL@ - a locus containing many immunoglobulin genes, but why the @ sign? https://en.wikipedia.org/wiki/IGL@
- Spooky toxin - https://en.wikipedia.org/wiki/Ssm_spooky_toxin (https://twitter.com/depthsofwiki/status/1712555421918245242)
- Always early (aly) - http://flybase.org/reports/FBgn0004372.html
- Lonely guy (LOG) - https://onlinelibrary.wiley.com/doi/full/10.1111/pbi.13783
- PKZILLA (very large gene) - https://www-science-org.libproxy.berkeley.edu/doi/10.1126/science.ado3290
- Dachshund (dac) "plays a role in leg development" (in flies) https://en.wikipedia.org/wiki/Dachshund_(gene)
- Blanks ("Loss of Blanks causes complete male sterility") https://www.pnas.org/doi/10.1073/pnas.1009781108
- LUMP (and with a p-element insertion p-lump) https://pmc.ncbi.nlm.nih.gov/articles/PMC3166160/
- loquacious https://www.ncbi.nlm.nih.gov/gene?Db=gene&Cmd=DetailsSearch&Term=34751
- TOPLESS https://pmc.ncbi.nlm.nih.gov/articles/PMC2643930/
- hemingway https://www.ncbi.nlm.nih.gov/gene/42207
- RAVER2 https://www.ncbi.nlm.nih.gov/gene/?term=Homo+sapiens+RAVER2
- ARSE https://www.ncbi.nlm.nih.gov/gene/100627778
- XXX https://www.ncbi.nlm.nih.gov/gene/1260943
- LAZY https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2020.606241/full
- BABY BOOM https://pubmed.ncbi.nlm.nih.gov/30298388/
- fiz/fezzik - "embryos and adult flies completely lacking CG9509 are unusually large - hence renaming it after the gentle giant Fezzik" (from the movie The Princess Bride) https://flybase.org/reports/FBgn0030594.html https://www.thenakedscientists.com/articles/science-features/gene-month-fezzik
- Shavenbaby https://en.wikipedia.org/wiki/Shavenbaby see also polished rice (pri)
- DAD - defender against death https://www.ncbi.nlm.nih.gov/gene/1603
- Hang (hangover) - https://www.alliancegenome.org/gene/FB:FBgn0026575
- pickpocket (ppk) - https://flybase.org/reports/FBgn0020258.html
Sometimes it is not the gene, but the allele that is named
- Bad hair day http://www.informatics.jax.org/allele/MGI:3764934
- Samba, chacha, bossa nova http://www.informatics.jax.org/allele/MGI:3708457
- Yoda http://www.informatics.jax.org/allele/MGI:3797584
Ref https://twitter.com/hmdc_mgi/status/1242893531779391496
A long list of RNA types https://en.wikipedia.org/wiki/List_of_RNAs, and circular RNA in particular https://en.wikipedia.org/wiki/Circular_RNA
Tables of RNA modifications, whose links keep rotting: https://www.hindawi.com/journals/jna/2011/408053/tab1/ used to point at http://mods.rna.albany.edu/mods/, which is also dead, so try http://genesilico.pl/modomics/modifications
Maternal RNAs deposited in the egg stay active in the zygote and can have transgenerational effects https://en.wikipedia.org/wiki/Maternal_to_zygotic_transition
Interesting PDB entries, from the molstar repo https://github.com/molstar/molstar/blob/master/docs/docs/misc/interesting-pdb-entries.md
Great illustrations of interesting biology, including information about gene names https://twitter.com/vividbiology
Many of the stories behind fly gene nomenclature are collected at https://web.archive.org/web/20110716201703/http://www.flynome.com/cgi-bin/search?source=browse including the famous ForRentApartments dot com gene (just kidding but lol https://web.archive.org/web/20110716202150/http://www.flynome.com/cgi-bin/search?storyID=180)
Musing article: "What is in a (gene) name?" https://web.archive.org/web/20180731060319/https://blogs.plos.org/toothandclaw/2012/06/17/whats-in-a-gene-name/