Contents
03 — Genomes, chromosomes and chromatin Part 1 — Molecular foundations

03 — Genomes, chromosomes and chromatin

Before this: Ch 01 · Ch 02 · Time: ~40 min

What you'll be able to do

The core idea

A is a solution to three that pull against each other.

It must fit. Two metres of DNA per cell, a nucleus six micrometres across.

It must stay readable. Whatever compacts it has to come apart locally, on demand, at thousands of sites, within seconds.

It must segregate accurately. At division, each daughter gets exactly one copy of everything — which requires a physical handle to pull on, and ends that the cell can tell apart from breaks.

Everything in this chapter follows from those three. The packaging is hierarchical because no single mechanism satisfies all three at once. And because compacted DNA is unreadable, the state of the packaging becomes a control layer in its own right — the genome's own metadata, written on top of the sequence and inherited alongside it.


1. The packaging problem, quantified

Start with the numbers, because the standard framing ("2 m into 6 μm!") points at the wrong difficulty.

A human cell holds 6.2 × 10⁹ bp. B-form DNA rises 0.34 nm per base pair, so the contour length is

6.2 × 10⁹ bp × 0.34 nm/bp = 2.1 × 10⁹ nm = 2.1 m

Now check whether volume is the . Treat DNA as a solid cylinder 2 nm across:

V_DNA  = π (1 nm)² × 2.1 m         ≈ 6.6 × 10⁻¹⁸ m³
V_nuc  = (4/3)π (3 μm)³            ≈ 1.1 × 10⁻¹⁶ m³
ratio  ≈ 6%

Six per cent. The DNA would rattle around in there. Volume is not the problem.

The problem is that a polymer does not sit still, and a long one occupies a region enormously larger than its own material. DNA has a of about 50 nm (~150 bp) — below that it is stiff, above it, floppy. For an ideal chain the end-to-end size goes as the Kuhn length b (twice the persistence length, so 100 nm) times the square root of the number of Kuhn segments. Take 1 alone, 248,956,422 bp in GRCh38:

contour L = 2.49 × 10⁸ bp × 0.34 nm    ≈ 85 mm
segments N = L / b = 8.5 × 10⁷ nm / 100 nm ≈ 8.5 × 10⁵
size R    ≈ b √N = 100 nm × 920         ≈ 92 μm

One chromosome, left to its own thermal devices, would sprawl across a sphere fifteen times the diameter of the nucleus — and that is the ideal-chain answer, which ignores the fact that the chain cannot pass through itself. Self-avoidance makes it worse.

So packaging is not about squeezing something bulky into a small box. It is about imposing a compact, non-random, actively maintained conformation on a polymer whose natural state is diffuse and tangled — and doing it without creating knots that would be lethal at .

The third constraint sets the extreme. At metaphase, chromosome 1 becomes a rod roughly 10 μm long. That is 85 mm of DNA in 10 μm of chromosome: about 8,500-fold linear compaction, achieved and then reversed once per cell cycle.

2. The nucleosome

The first level of packaging is a spool.

An octamer of histone proteins — two copies each of H2A, H2B, H3 and H4, assembled as one (H3·H4)₂ tetramer flanked by two H2A·H2B dimers — forms a squat disc about 11 nm across and 5.5 nm high. 147 bp of DNA wraps around it in ~1.65 left-handed superhelical turns. That unit is the core particle, and it is the most abundant protein–DNA complex on Earth.

   147 bp core                  linker
   (~1.65 turns)               ~20–80 bp
   ┌───────────┐              ┌───────────┐
───┤  octamer  ├──────────────┤  octamer  ├────── ...
   └───────────┘              └───────────┘
     2× H2A, H2B                 nucleosome repeat length
     2× H3,  H4                  ~185–200 bp in human somatic cells

Three structural facts do all the work:

Histones grip the backbone, not the bases. Contacts are made at ~14 points where the faces the octamer, largely by arginine side chains inserting into it, plus electrostatic attraction to the phosphates. Histones are 20–25% lysine and arginine — strongly basic, which is exactly what you predict for something that must neutralise a polyanion (Ch 01 §3). Because recognition is of the backbone, any sequence can be wrapped. The genome is uniformly packageable.

But sequence biases positioning. Wrapping requires bending, and bending is easier where AA/TT/TA dinucleotides recur with the ~10 bp helical period, so the compressible minor groove faces inward on each turn. This is a soft preference, not a code: nucleosome positions in vivo are set mostly by competition with other DNA-binding proteins, by ATP-dependent remodellers that slide octamers along, and by boundary effects — not by sequence alone.

The tails stick out. Each histone has an unstructured N-terminal tail protruding past the DNA gyres. These are chemically modifiable at dozens of residues, and those modifications are the substrate of Ch 23. Nothing about the tails is needed for wrapping. They exist to be written on.

H1 is not part of the octamer. It binds where DNA enters and exits, clamping an additional ~20 bp of linker (giving the ~166 bp "chromatosome"), and stabilises higher-order folding. It is present at roughly one copy per nucleosome, exchanges rapidly, and its depletion decompacts chromatin.

How constrained is all this? Histone H4 is 102 amino acids long and differs at just two positions between a pea and a cow. Over a billion years of divergence, essentially every residue has been held in place. When a protein is that invariant, every surface of it is doing something.

Counting. With a nucleosome repeat length of ~200 bp:

6.2 × 10⁹ bp / 200 bp ≈ 3.1 × 10⁷ nucleosomes per diploid cell
                      ≈ 2.5 × 10⁸ histone molecules

Thirty-one million spools, all of which must be duplicated during S in step with replication — a demand met by histone genes held in multi-copy clustered arrays and transcribed in a burst tied to S phase.

Compaction achieved: 200 bp of DNA is 68 nm of contour, and in the nucleosomal fibre it occupies roughly 11 nm of axial length. About 6- to 7-fold. Against a required 8,500-fold, the nucleosome is barely a down payment — but it is the level at which everything else is built, and it is already regulatory, because DNA on the octamer surface is occluded from most proteins that would read it. Measuring which DNA is not wrapped is the whole basis of (Ch 49).

3. Above the nucleosome: the fibre that isn't there

The nucleosomal fibre — "beads on a string", ~10 nm wide — is real and is seen readily in the electron microscope at low ionic strength.

The next level in every textbook diagram is the 30 nm fibre: coiling into a thicker helix, either as a one-start solenoid or a two-start zigzag, taking cumulative compaction to roughly 40-fold relative to naked DNA — a further ~6-fold beyond the nucleosome, not a further 40. It has beautiful supporting data. Purified nucleosome arrays plus magnesium plus H1 fold into 30 nm fibres in vitro; reconstituted 12-mers have been solved structurally; nuclei from chicken erythrocytes — transcriptionally silent cells with a specialised linker histone — show them.

What it does not have is evidence from inside a living cell. Cryo-electron microscopy of vitrified mitotic found no 30 nm periodicity. Small-angle X-ray scattering of nuclei found none. And ChromEMT — an EM method that stains DNA in situ and images it in three dimensions in human interphase and mitotic cells (Ou et al., Science, 2017) — found chromatin organised as disordered chains 5–24 nm in diameter, varying in local packing density, with no regular fibre at any level. The same study reproduced 30 nm fibres in chicken erythrocyte nuclei treated with magnesium, which is the tell: the structure forms under the conditions that produce it, and those conditions are not nuclear ones.

The classical hierarchy — 10 nm → 30 nm → 300 nm loops → 700 nm chromatid — is a diagram of a series of experiments, not a diagram of a nucleus. Only the first level survives in vivo. Everything above it is better described as an irregular, dynamic, locally variable polymer melt whose large-scale organisation comes from motor proteins and from self-association, not from a repeating geometric unit.

What is actually there, in ascending scale:

graph TD A["<b>naked DNA</b> · 2 nm<br/>contour 2.1 m per diploid cell"] B["<b>nucleosomal fibre</b> · ~10 nm<br/>147 bp per octamer · ~6× compaction<br/><i>established in vivo</i>"] X["<b>30 nm fibre</b><br/><i>in vitro only — not observed in situ</i>"] C["<b>irregular chains</b> · 5–24 nm<br/>variable density, no repeating unit"] D["<b>loops</b> · 10s–100s kb<br/>extruded by cohesin, anchored at CTCF sites"] E["<b>TADs and A/B compartments</b> · 0.1–10 Mb<br/>like associating with like"] F["<b>chromosome territories</b><br/>each chromosome occupies its own region"] G["<b>mitotic chromatid</b><br/>nested condensin loop array · ~8,500× linear"] A --> B B -.->|"Mg²⁺ + H1,<br/>purified arrays"| X B --> C --> D --> E --> F C -.->|"mitosis"| G

is worth naming now because it recurs: the complex grabs chromatin and reels it in from both sides, growing a loop until it hits a boundary — typically a protein bound in the correct orientation. It is a motor producing structure, continuously, at ATP cost. Condensin does the mitotic version, building nested loop arrays that turn an interphase tangle into a rod. Ch 50 develops all of it.

4. Chromosome anatomy and the karyotype

A linear needs exactly three kinds of functional element to persist through division: origins of replication (start copying here), a (a handle to pull on), and (ends that are not breaks). Supply those three to a piece of yeast DNA and it behaves as a chromosome; omit any one and it is lost.

Centromere. Visible as the primary constriction; the site where the kinetochore assembles and spindle microtubules attach (Ch 09). Human centromeres are built on alpha satellite DNA: a 171 bp monomer, monomers grouped into chromosome-specific higher-order repeats, and those tandemly reiterated into arrays spanning roughly 0.1–5 Mb.

But the sequence is neither necessary nor sufficient. Centromere identity is epigenetic, marked by in which histone H3 is replaced by the variant CENP-A. Neocentromeres form on ordinary single-copy sequence and function normally; conversely, some chromosomes carry a second alpha satellite array that is never used. The DNA is a preferred substrate, not the instruction.

Telomere. Tandem TTAGGG repeats, 5–15 kb in humans, ending in a single-stranded 3′ G-rich overhang that tucks back into the duplex to form a t-loop, coated by the shelterin protein complex. Two problems solved at once: the end-replication problem (Ch 04) and the recognition problem — without shelterin, a chromosome end is indistinguishable from a double-strand break and gets "repaired" by fusion to another chromosome (Ch 17).

Arms and morphology. The centromere divides the chromosome into a short p arm (petit) and a long q arm. Position of the centromere gives the classical morphology:

   metacentric              submetacentric            acrocentric
   (1, 3, 19, 20)           (2, 4–12, 16–18, X)       (13, 14, 15, 21, 22, Y)

      ┌────┐                     ┌────┐                    ┌┐  ← tiny p arm:
      │    │  p                  │    │  p                 ││    rDNA / NOR
      ├────┤ ← centromere        ├────┤                    ├┤
      │    │                     │    │                    ││
      │    │  q                  │    │                    ││
      └────┘                     │    │  q                 ││  q
                                 │    │                    ││
                                 └────┘                    └┘
      p ≈ q                      p < q                     p ≪ q

Humans have no truly telocentric chromosomes. The Y is acrocentric by centromere position but is not one of the NOR-bearing set — it carries no ribosomal RNA arrays. The five NOR-bearing acrocentrics matter disproportionately: their short arms carry the ribosomal RNA gene arrays (nucleolar organiser regions) embedded in near-identical repeat blocks shared between chromosomes. That shared sequence lets them recombine with each other, which is the origin of (Ch 20) — and it is exactly why those regions were absent from the until , which resolved the entire short arms of all five acrocentrics as part of the ~8% of the genome it added (verified-facts).

. The complete chromosome complement. Humans: 46 chromosomes — 22 pairs of autosomes plus one pair of , XX or XY. Written 46,XX / 46,XY; abnormalities append, so 21 is 47,XY,+21 and the Philadelphia chromosome is 46,XX,t(9;22)(q34;q11).

Chromosomes were numbered by size before anyone could measure them properly, and the numbering is slightly wrong: chromosome 21 (46.7 Mb in GRCh38) is smaller than chromosome 22 (50.8 Mb). Chromosome 1 is the largest at 249 Mb.

Band addresses. Stained chromosomes show reproducible banding patterns, and the resulting coordinate system predates sequencing but is still the language of clinical cytogenetics:

   17      q       21       .31
   │       │       │        │
   chr     arm     region   sub-band       →  17q21.31
                   + band

Bands are counted outward from the centromere. A modern genomic coordinate replaces this with a build and an offset, and the build is not optional: BRCA1 sits at 17q21.31, which in GRCh38 is chr17:43,044,295–43,170,245 on the minus strand — but in GRCh37 the same gene starts near chr17:41.2 Mb. Same gene, same band, coordinates differing by 1.85 Mb.

5. Euchromatin and heterochromatin

Cytologists noticed a century ago that some regions stay condensed through interphase while the rest decondenses. The distinction survived because it turned out to track function.

Euchromatin
Compaction in interphaseopencondensed
Gene contentgene-richgene-poor
active or activatablesilenced
Replication timingearly Slate S
Recombination ratehigherstrongly suppressed
GC contenthigherlower
Repeat contentlowerhigh — satellites, LINEs
MarksH3K4me3, H3K27ac, accessibleH3K9me3 + HP1, or H3K27me3 + Polycomb

Heterochromatin splits in two, and the split is the useful part:

Constitutive — heterochromatic in every cell type, always. Pericentromeric and telomeric satellite arrays, the Y long arm, the rDNA-flanking regions. This is structural: it keeps repeat arrays from recombining with each other, and it is required for function. Marked by H3K9me3 and bound by HP1, which self-associates and spreads — giving a mechanism that propagates along the and through cell division.

Facultative — heterochromatic in some cells, some of the time. The canonical example is the inactive X: in cells with two X chromosomes, one is silenced in early development and packaged into a compact body (the Barr body), and that choice is then maintained clonally for life. Developmental gene clusters silenced by Polycomb are the other major class. This is regulatory: it is how a cell locks in decisions about identity.

Treat the two-state picture as a first approximation. Genome-wide chromatin maps resolve chromatin into a dozen or more states, and the underlying variables — , modification, replication timing, compartment — are continuous and only partly correlated. "Euchromatin/heterochromatin" is a useful coarse-graining, not a type system.

6. What the human genome is actually made of

Size first. The haploid human genome is ~3.1 Gb; the complete T2T-CHM13v2.0 assembly totals 3,117,292,070 bp, of which ~8% is sequence that no prior reference contained.

The , from GENCODE Release 50 (verified-facts):

CategoryCount
Protein-coding genes19,442
Long non-coding RNA genes35,885
Small non-coding RNA genes7,608
14,702
Total annotated genes78,733
Total transcripts644,292

Two things fall straight out.

Protein-coding sequence is a rounding error. A typical human protein is a few hundred amino acids; take 400, which needs 1,200 bp of coding sequence:

19,442 genes × 1,200 bp ≈ 23 Mb ≈ 0.75% of 3.1 Gb

Add untranslated regions and alternative exons and the exonic fraction rises to a couple of per cent. The conventional "~1–2% codes for protein" is right, and the derivation shows why the figure is soft rather than exact. Meanwhile the units of those genes — included — span something like 40% of the genome. Genes are mostly not coding sequence; genomes are mostly not genes.

Non-coding genes outnumber coding genes about 3:1 (58,195 to 19,442). That single ratio is the cleanest available refutation of the framing that dominated popular genetics for fifty years.

Killing "junk DNA" — carefully

Ohno coined "" in 1972 for a defensible reason: population genetics limits how much sequence a species can maintain against , so most of a large genome cannot be under selection. The inference was sound. The word was not, because it collapsed "not protein-coding" into "not doing anything".

Against "it's all junk":

Against "it's all functional" — and this correction matters just as much:

Non-coding is not junk; transcribed is not functional; and the true functional fraction sits somewhere between 2% and 20% depending entirely on what you mean by the word. Ch 39 argues this out properly.

Genes are not evenly distributed

Average density is 19,442 / 3,100 Mb ≈ 6 protein-coding genes per Mb. The variance is enormous: 19 carries about 26 genes per Mb, four times the average, while chromosome 13 sits near 3 per Mb across a 38 Mb stretch of its interior.

Density tracks base composition. The genome averages ~41% GC, but that average is assembled from long, relatively homogeneous domains — isochores, in Bernardi's classification, running from L1 (<37% GC) to H3 (>53% GC) over hundreds of kilobases. The GC-rich end is where the genes are:

GC-poor (L) isochoresGC-rich (H) isochores
Gene densitylowhigh
Intron lengthlongshort
repeatAlu
Replication timinglateearly
Recombination ratelowhigh
Chromatinmore heterochromaticmore open

Why the domains exist is not "selection for thermal stability", the original guess. The leading explanation is mechanical: . When recombination resolves a mismatched heteroduplex, repair is biased toward G and C, so regions that recombine often GC-ward regardless of any consequence. It is a mutational bias masquerading as selection — a pattern worth learning to suspect early (Ch 33).

The practical consequence lands squarely in your lap: GC-extreme sequence amplifies and sequences badly. drops in GC-rich and and in AT-rich regions, which means the parts of the genome you most want to are the parts most likely to be undercovered (Ch 40, Ch 46).

7. Repetitive DNA

Repeats are the majority of the genome and the source of most of the difficulty in working with it. Two families, distinguished by arrangement rather than origin.

graph TD R["<b>repetitive DNA</b><br/>~50%+ of the human genome"] T["<b>tandem</b><br/>copies adjacent, head to tail"] I["<b>interspersed</b><br/>copies scattered — mostly TE relics"] R --> T R --> I T --> S["<b>satellite</b><br/>Mb arrays · centromeres,<br/>pericentromeric, Yq"] T --> M["<b>minisatellite / VNTR</b><br/>10–100 bp unit · 0.5–30 kb array"] T --> U["<b>microsatellite / STR</b><br/>1–6 bp unit · usually &lt;100 bp"] I --> L["<b>LINEs</b> ~21%<br/>L1 ~17% · 6 kb · autonomous"] I --> A["<b>SINEs</b> ~13%<br/><i>Alu</i> ~11% · 300 bp · needs L1"] I --> V["<b>LTR / ERV</b> ~8%<br/>retroviral relics"] I --> D["<b>DNA transposons</b> ~3%<br/>none active in humans"]
ClassUnitArrayCopiesWhere, and why you care
Satellite171 bp (alpha) and others0.1–5 Mb— and pericentromeric blocks. Unassemblable before long reads
Minisatellite (VNTR)10–100 bp0.5–30 kbthousands of lociHighly variable in copy number; the basis of 1980s DNA fingerprinting. are a special case
Microsatellite (STR)1–6 bptens of bp>10⁶ loci, ~3% of genomeMutate at 10⁻³–10⁻⁴ per locus per generation via polymerase slippage. Forensic identification, mapping, and repeat-expansion disease
—6 kb full length~500,000, ~17%The only autonomously active human transposon. A few dozen to ~100 copies per person remain retrotransposition-competent
Alu—~300 bp>10⁶, ~11%Non-autonomous — hijacks L1 machinery. Dense in GC-rich, gene-rich regions. Its abundance drives non-allelic

The class percentages come from different and are approximate; the pinned total for transposable-element-derived sequence is ~46% of the genome. Ch 19 covers the biology.

For programmers. Repeats are why genomics is algorithmically hard, and each class breaks a different tool. Interspersed repeats make short- ambiguous — a 150 bp read from an Alu has a million plausible origins, so the aligner emits it with mapping quality 0 and every downstream caller discards it (Ch 42). Tandem repeats break assembly: a cannot resolve a cycle longer than the read, so a megabase satellite array is a tangle no amount of short-read fixes (Ch 43). This is not a detail of implementation — it is why 8% of the genome had no reference sequence at all until reads got longer than the repeats.

8. The other genome: mitochondria

Mitochondria carry their own , a relic of the bacterium they descend from, and it breaks nearly every rule established above.

PropertyHuman mtDNA
Size16,569 bp, circular, no histones
Genes37 — 13 protein-coding, 22 tRNA, 2 rRNA
Non-coding~1.1 kb control region (D-loop); essentially no
Copy number~10³–10⁴ per somatic cell; ~10⁵–10⁶ in an oocyte
InheritanceMaternal — paternal mitochondria are actively destroyed after fertilisation
roughly an order of magnitude above the nuclear rate
slightly different: AGA/AGG = stop, ATA = Met, TGA = Trp (Ch 07)

The 13 proteins are all subunits of the oxidative-phosphorylation machinery. Everything else a mitochondrion needs — including the polymerase that copies mtDNA — is nuclear-encoded and imported. The organelle is genetically dependent; over two billion years most of its ancestral genes migrated to the nucleus.

High copy number is what makes mitochondrial genetics strange. A cell does not have two ; it has thousands of molecules, and they need not agree. Homoplasmy is all copies identical; is a mixture, quantified as a percentage. Three consequences:

A of a mitochondrial disease therefore shows transmission through females only, never through affected males, with wildly variable severity among siblings. It matches no Mendelian pattern, and recognising it is a standard exam question.

One bioinformatic trap: fragments of mtDNA have inserted into the nuclear genome repeatedly over evolution. These are near-identical to real mtDNA, so reads misalign between the two and produce phantom low-frequency heteroplasmy. Any pipeline calling mitochondrial variants must handle them explicitly (Ch 46).

9. Genome size and ploidy

Genome size across life spans five orders of magnitude, and the ordering will not be what you expect.

OrganismHaploid genomeProtein-coding genes
Escherichia coli K-124.6 Mb~4,400
Saccharomyces cerevisiae (yeast)12 Mb~6,000
Caenorhabditis elegans100 Mb~20,000
Arabidopsis thaliana135 Mb~27,000
Drosophila melanogaster180 Mb~14,000
Takifugu rubripes (pufferfish)~400 Mb~19,000
Homo sapiens3.1 Gb19,442
Allium cepa (onion)~16 Gb—
Triticum aestivum (bread wheat)~16 Gb, hexaploid~107,000 (high-confidence)
Tmesipteris oblanceolata (fork fern)160.45 Gb—

Read down that table with the "more complex organism, more DNA" assumption and it dies at the first plant. A pufferfish carries roughly the same gene count as you in an eighth of the DNA. An onion has five times your genome. A 10 cm fork fern from New Caledonia has fifty times it — the largest genome yet measured.

This is the : genome size correlates with neither organismal complexity nor gene number. ("C-value" is the DNA content of one haploid set; human 1C ≈ 3.1 Gb.) The resolution has two parts, transposable-element accumulation and whole-genome duplication, and Ch 39 works it through. Note also that the largest claimed genome you will find quoted — the amoeba Polychaos dubium at 670 Gb — comes from a 1960s dye-binding measurement that has never been reproduced and is generally regarded as unreliable. Genome-size records need method scrutiny.

is the number of complete sets:

TermSetsHuman example
Haploid (n)1Sperm and eggs — n = 23
(2n)2Somatic cells — 2n = 46
Triploid (3n)369,XXX — arises from dispermy; invariably lethal
Polyploid (>2n)manyRare in animals, routine in plants

Bread wheat is allohexaploid, 2n = 6x = 42: three complete diploid genomes from three ancestral grasses in one nucleus. Polyploidy is a major engine of plant evolution and near-absent in mammals, where the sex-determination and dosage-compensation machinery does not survive doubling.

Two refinements that prevent later confusion. First, "human = diploid" is an approximation: hepatocytes are frequently tetraploid or octoploid, cardiomyocytes polyploidise with age, megakaryocytes reach 64n, and roughly a third of solid tumours have undergone whole-genome doubling (Ch 56).

Second, ploidy and C-value are different axes. n counts chromosome sets; C counts DNA content. A diploid cell in G2 — after replication, before division — is still 2n but is 4C, because every chromosome now consists of two sister chromatids. Keeping those apart is the difference between following Ch 09 and being permanently lost in it.


Common misconceptions

What people thinkWhat's actually true
DNA has to be packaged because the nucleus is too small to hold itAs bare cylinder, DNA occupies ~6% of nuclear volume. The problem is polymer conformation: an unconstrained would sprawl across ~90 μm, fifteen nuclear diameters. Packaging fights entropy and entanglement, not volume
Chromatin folds 10 nm → 30 nm → loops → chromosomeOnly the 10 nm nucleosomal fibre is established in vivo. The 30 nm fibre forms from purified arrays with Mg²⁺ and H1; in situ imaging of human cells shows disordered 5–24 nm chains and no regular fibre. Large-scale order comes from and compartmentalisation
Chromosomes look like the X-shaped things in diagramsOnly for a couple of hours around mitosis. For most of the cell cycle a chromosome is a decondensed, territory-occupying tangle you cannot see individually without labelling it
The is defined by its DNA sequenceIt is defined epigenetically, by CENP-A . Neocentromeres form on ordinary sequence; inactive alpha satellite arrays exist. Alpha satellite is the usual substrate, not the instruction
Non-coding DNA is junkNon-coding genes outnumber coding ones ~3:1, and regulatory, structural and RNA-gene sequence is unambiguously functional. But the opposite over-correction is also wrong: ~46% is degraded transposon relic and only ~5–10% shows detectable
Bigger, more complex organisms have bigger genomesThe . An onion has 5× the human genome; a pufferfish has ⅛ of it with the same gene count; the record holder is a 10 cm fern at 160 Gb. Size tracks transposon load and duplication history, not complexity
You inherit mitochondria from both parents like everything elsemtDNA is maternally inherited; paternal mitochondria are actively degraded after fertilisation. Vanishingly rare exceptions have been reported and remain contested
A mitochondrial is either present or absentThere are thousands of mtDNA copies per cell, so a mutation has a load. Disease appears above a threshold (often 60–90%), and load varies between tissues, between siblings, and over a lifetime
Chromosome 22 is the smallest chromosomeChromosome 21 is (46.7 Mb vs 50.8 Mb, GRCh38). The numbering was assigned by eye in the 1960s and never corrected

Worked example: audit one diploid human cell

Everything in this chapter, in one accounting exercise. Start from 6.2 × 10⁹ bp of nuclear DNA.

1. Physical length.

6.2 × 10⁹ bp × 0.34 nm/bp = 2.1 m

2. . At a 200 bp repeat length:

6.2 × 10⁹ / 200 = 3.1 × 10⁷ nucleosomes
× 8 histones each  = 2.5 × 10⁸ core histone molecules

Every one of which must be synthesised and deposited during a single S .

3. DNA actually on a spool.

3.1 × 10⁷ × 147 bp = 4.6 × 10⁹ bp  →  4.6/6.2 = 74% wrapped

About three-quarters of the genome is on an octamer at any moment. The remainder is overwhelmingly linker — the short, constrained stretches between adjacent nucleosomes — and linker is not what assays report. Nucleosome-depleted regions at active and are a small subset of that remainder, on the order of 1–3% of the genome, and those are what and DNase-seq actually measure. Conflating "not on an octamer" with "accessible" overstates the accessible fraction by roughly an order of magnitude.

4. Protein-coding sequence.

19,442 genes × ~1,200 bp × 2 copies ≈ 47 Mb  →  47/6,200 = 0.75%

Under 1%. Call it 1–2% once UTRs and alternative exons are included.

5. Transposable-element-derived sequence.

0.46 × 6.2 × 10⁹ ≈ 2.9 × 10⁹ bp

Sixty times more transposon relic than protein-coding sequence. Any account of the genome that leads with genes is describing under 2% of it.

6. Alu elements alone.

0.11 × 6.2 × 10⁹ / 300 bp ≈ 2.3 × 10⁶ copies

Over two million near-identical 300 bp sequences scattered through the genome — roughly one every 3 kb. This is the number to hold in mind when an aligner reports mapping quality 0.

7. .

16,569 bp × ~5,000 copies ≈ 8.3 × 10⁷ bp  ≈ 1.3% of nuclear DNA content

Small, but not negligible — which is why 0.1–1% of reads in a standard whole-genome library are mitochondrial, and why mtDNA gets called at extraordinary depth for free.

8. Mitotic compaction, checked. 1: 249 Mb → 85 mm of DNA → a 10 μm metaphase chromatid.

85 × 10⁻³ m / 10 × 10⁻⁶ m ≈ 8,500-fold

The nucleosome contributes ~6-fold of that. The remaining ~1,400-fold is by condensin, and it happens in under an hour.

Connections