# Sequencing and Genomes

> Genome results depend on what each read records and where it fits. This lesson follows Sanger chain termination, short-read sequencing by synthesis and the two main long-read methods, then separates random sampling gaps from repeat ambiguity. Afterwards you can calculate the reads a genome needs, explain how long reads, with a nearly homozygous cell line, closed the last gaps in a human assembly, and interpret structural variants while accounting for the reference used.

Course: Genetics and Molecular Biology (Tools and genomes). Level: core university (third year). About 19 minutes.
Lesson page (link this one when citing): https://stemhop.com/lessons/sequencing-and-genomes/

Builds on: DNA Replication: Semiconservative Copying and Its Machinery · Recombinant DNA, Cloning and PCR · Mutation: Where Variation Comes From

## What you'll learn

- Explain how dideoxynucleotides turn base order into a ladder of fragment lengths, and how reversible terminators let short-read machines read one base per cycle instead.
- Calculate the reads needed for a stated mean coverage, and use the Poisson model to estimate how much of a genome goes unread.
- Explain why a repeat longer than a read breaks an assembly, and why long reads from PacBio and Oxford Nanopore instruments can resolve it.
- Describe the human reference from the 2001 drafts to T2T-CHM13 and the 2023 draft pangenome, and explain why a variant call depends on the reference used.

Two drafts of the human genome were published in February 2001. The first assembly in which every chromosome but Y ran end to end without a gap came in 2022. As the first film showed, faster machines alone could not close the gaps: a sequencer reports **reads**, the base sequences of single DNA fragments, and reporting a base is not locating it in a genome.

> **The one idea.** Genome results depend on what each read records and where it fits.
>
> *Scope:* sequencing, alignment, assembly and variant calling. Read evidence alone does not say what a variant means for a person.

## Reading DNA one base at a time: Sanger’s stopped copies

For the first film’s stopped copies, you supply a single-stranded template, a short **primer** (DNA that pairs with the template and gives the polymerase a 3’ end to extend), the four ordinary deoxynucleotides, and a small amount of **dideoxynucleotides** (ddNTPs). A ddNTP lacks the 3’ hydroxyl that the next nucleotide must bond to, the chemistry you met in [nucleotide metabolism](https://stemhop.com/lessons/nucleotide-metabolism/).

Frederick Sanger, Steve Nicklen and Alan Coulson published the method in 1977, using DNA from the phage phiX174. In the classical version, four tubes each held one kind of ddNTP, and the radioactive fragments ran in four lanes of a gel that separates pieces differing by a single nucleotide. Later versions give each terminator its own fluorescent dye, so one reaction holds all four, and a capillary sorts fragments by size while a detector reads each colour. Shortest to longest spells the **new** strand 5’ to 3’; the template is its reverse complement.

**Predict.** A detector reports the shortest fragments in order as A, C and G. What are the new strand and its template, each written 5’ to 3’?

> **Check your prediction.** The new strand is 5’-ACG-3’. Its antiparallel template is 3’-TGC-5’, or 5’-CGT-3’. Complementing without reversing gives the template backwards.

Sanger reads are accurate, but each reaction reads one region, commonly up to about 800 to 1,000 bases, and the first few dozen bases are often poor. The sequence in the 2001 drafts came mostly from this chemistry, run on many instruments at once.

## Reading millions at once, short and long

Short-read sequencing runs the first film’s add, image and reset cycle on hundreds of millions of fragments at once. Genomic DNA is broken into fragments a few hundred bases long and fitted with adapters. In the [Illumina](https://stemhop.com/methods/next-generation-sequencing/) method, which David Bentley and colleagues used to sequence a human genome in 2008, each fragment is copied in place on a glass flow cell into a **cluster** of identical copies. The pause comes from a **reversible terminator**, a removable block on the 3’ end of each dye-carrying nucleotide; cleaving off the dye and the block regenerates the 3’ hydroxyl for the next cycle.

*Figure (AI-generated illustration).* A removable dye and block let every cluster report one base, reset and begin the next sequencing cycle. Image: a cluster of identical DNA copies on a glass flow cell shown in three steps: a nucleotide with a dye and a terminator block is added, the cluster is imaged and its colour read, and the dye and block are cleaved off so the cycle can repeat.

Reading each fragment from both ends, a **paired-end** read, gives two short sequences and roughly how far apart they lie. Reads are short (35 bases from each end in the 2008 human genome; commonly 100 to 150 now), and the errors in the study of one such instrument were mostly substitutions.

In PacBio instruments the polymerase sits at the bottom of a tiny well, with the dye on the nucleotide’s terminal phosphate so it is released as the base joins; going round the circular template many times gives **HiFi** reads, in a 2019 study of one human genome averaging 13.5 kilobases at 99.8% accuracy, with leftover errors mostly small insertions or deletions in runs of one base. In Oxford Nanopore instruments the pore is a protein; a 2018 study of human DNA reported ultra-long libraries in which about half the sequence lay in reads of roughly 100 kilobases or longer. The trade-off: short reads give many accurate bases per run but cannot span long repeats; long reads span most of them, though the largest repeat arrays still need ultra-long reads and special assembly methods.

*Figure (AI-generated illustration).* Single-molecule methods preserve longer-range context: PacBio follows one polymerase round a circular template many times, and Oxford Nanopore reads the current as one strand passes through a pore. Image: on the left a single polymerase at the bottom of a tiny well copying a circular template, releasing a dye as each base joins; on the right a single strand threading through a protein pore in a membrane, with software decoding the current.

That trade-off sets up the second film: once the reads exist, can their combined evidence determine one genome?

**Film transcript: When Reads Become a Genome**

### Film transcript: When Reads Become a Genome

The first film turned molecular events into reads. How do millions of reads become one genome? Some overlap unique places. Others seem to fit several. The computer must separate missing evidence from evidence that cannot be placed.

Start with sampling. Coverage is the mean number of reads lying over each base. As more gold reads land at random, blue gaps shrink. The mean does not promise equal stacks everywhere. A position can still receive many reads, few reads, or none. Under ideal random sampling, the unread fraction falls exponentially as mean coverage rises. Adding reads therefore attacks a sampling problem. It makes an unseen position less likely, although real libraries are less even. Now the same red repeat block appears at two places along the blue ribbon. A short gold read from inside that block matches both places equally. Its bases may be accurate, yet its address is ambiguous.

More copies of that same short evidence do not create a unique address. Another pile of short reads is ready to land on the repeats. Will a deeper pile tell which flanking regions belong together? The pile grows, but both placements remain possible. More reads fill sampling gaps; reads that span a repeat place it. Now one long gold read crosses the red repeat and reaches blue unique sequence on both sides. That span links the repeat copy to its flanks. De novo assembly joins reads without a reference, first producing contigs or an assembly graph. Reference alignment asks where reads fit on a reference path or graph already chosen. Now the reference branches.

A read missing from one path can fit another. What counts as a reported difference changes with the reference. Long accurate reads and ultra-long spans closed difficult gaps in one human assembly. A pangenome then added alternative paths from many chromosome copies. Completeness asks whether sequence is missing. Representation asks whether one path captures human variation. Reference choice changes two stages. First, it changes whether a read has a close place to align. Then it changes which base or segment is labelled different. A variant call is therefore a comparison, not a claim about rarity, selection, or health. Genome results depend on what each read records and where it fits.

Was the region sampled, and was the evidence placed uniquely? Extra reads help the first. Reads that span the repeat help the second. A richer reference gives reads from missing sequence a place to land. Those are different limits, and they need different fixes.

From the film

- More reads fill sampling gaps; reads that span a repeat place it.
- Coverage is a mean, so random sampling can leave some bases unread.
- An internal read from an exact repeat may fit more than one place.
- A reference helps place reads but also shapes which differences are reported.
- Genome results depend on what each read records and where it fits.

## Coverage and assembly: two different limits

Reads **aligned** to a reference are placed on it; reads assembled **de novo**, without one, are joined by their overlaps, first into an assembly graph or **contigs**, continuous consensus sequences. **Scaffolds** place contigs in order and orientation, with estimated gaps between them.

**Coverage** (depth), the second film’s mean stack of reads over each base, is all the bases read, spread over the genome:

> **Equation card: How many times, on average, each base of a genome is read.**
>
>     C = N × L / G
>
> In words: Coverage is the number of reads times the read length, divided by the genome length.
>
> - C: the mean number of reads over each base, written as a fold (30-fold, 30x)
> - N: the number of reads
> - L: the read length (bases)
> - G: the haploid genome length (bases)
>
> Sanity check: Bases on top and bases below cancel, so coverage is a plain number. Double the reads or the read length and coverage doubles; sequence a genome twice as big with the same reads and it halves.
> Try one number Round numbers to keep the arithmetic easy (not from a real project): 1,000 reads of 100 bases on a 10,000-base piece of DNA. Total bases read: 1,000 × 100 = 100,000. Spread over 10,000 bases: 100,000 / 10,000 = 10. Each base is covered 10 times on average, 10-fold coverage.

**Worked problem.** Assume reads start at random positions, all reads are usable, and the haploid human genome is G = 3.1 x 109 bases (an estimate rounded for the calculation).

**Predict, then calculate.** How many 150-base reads give 30-fold mean coverage? First decide: should it rise or fall if each read were longer? Then multiply both sides by G (C x G = N x L) and divide by L.

> **N = C x G / L = 30 x 3.1 x 109 / 150 = 9.3 x 1010 / 150 = 6.2 x 108 reads**

That is 620 million reads, or 310 million read pairs of 2 x 150 bases, and 93 billion bases. Check: 6.2 x 108 x 150 = 9.3 x 1010, and dividing by 3.1 x 109 gives 30. Longer reads mean fewer of them: doubling L to 300 bases halves N to 3.1 x 108. (Going deeper works the powers of ten step by step; see also [powers](https://stemhop.com/math-toolkit/#powers).)

**Predict, then calculate.** At C = 5, what fraction of bases should be unread, and how many bases is that in this genome? The second film’s falling curve comes from the Poisson model: with random starts, the number of reads over a base follows a Poisson distribution with mean C. The fraction with zero reads is then:

> **Equation card: What fraction of the genome gets no read at all at a given mean coverage, if reads start at random.**
>
>     fraction unread = e^−C, where C is the mean coverage
>
> In words: The fraction of bases with no read is e to the minus C.
>
> - e^−C: the Poisson chance of zero reads at a base. e is about 2.718, and e^−C means 1 divided by e^C; for C = 3 that is 1 divided by e × e × e.
> - C: the mean coverage (fold)
>
> Sanity check: At C = 0, e^0 = 1: with no reads, every base is unread. Each extra fold of coverage multiplies the unread fraction by e^−1, about 0.37, so it falls fast but never reaches exactly zero.
> Try one number At C = 3: e^−3 is about 0.37 × 0.37 × 0.37, about 0.05 (0.0498 on a calculator). So at 3-fold coverage about 5% of the genome would get no read at all.

At three depths:

> **C = 1: e−1 = 0.37, so 37% unread**
>
> **C = 5: e−5 = 0.0067, so 0.0067 x 3.1 x 109 = about 21 million bases unread**
>
> **C = 30: e−30 is about 9 x 10−14, far less than one base**

So 30-fold is a mean, not a guarantee; Going deeper adds real libraries and the model’s source.

Put numbers on the second film’s repeat: two identical 6-kilobase copies. No 150-base read, and no pair from a 400-base fragment, reaches from unique sequence on one side across the repeat to its far edge. The assembler cannot determine which flanking sequences belong together, so it may collapse the copies or break the contig. More reads fill sampling gaps but add no new span; reads that span a repeat place it. A 15-kilobase read can hold one whole copy plus several kilobases of unique sequence on each side, pinning that copy in place.

That is the rule in short: ask what each read records, then where it fits. The exact-repeat model is deliberately simple: differences between copies, linked pairs and other long-range evidence can sometimes place repeats no single read spans. Some human-specific gene duplications described in [the human brain expansion lesson](https://stemhop.com/lessons/human-brain-expansion/) sit in long near-identical repeats, one reason they were hard to study.

*Interactive on the lesson page:* Coverage and repeats: will the reads assemble?

## The human reference: gaps, a finished genome and a pangenome

The 2001 papers, from the public International Human Genome Sequencing Consortium and from Celera Genomics, were drafts. The consortium’s “finished” sequence of 2004 covered about 99% of the **euchromatin** (the gene-rich, less tightly packed part of the genome) and still had 341 gaps, many associated with segmental duplications; centromeres and other heterochromatin were not part of its target. The same paper estimated 20,000 to 25,000 protein-coding genes, far fewer than the 2001 estimate. The current standard reference, GRCh38, descends from it: a mosaic of DNA clones from several people, not any one person’s genome.

Many remaining gaps were the worked example’s repeat problem at chromosome scale, in large centromeric arrays and near-identical segmental duplications. The Telomere-to-Telomere Consortium closed them using the CHM13 cell line, whose two copies of each chromosome are nearly identical, and a combination of PacBio HiFi and Oxford Nanopore ultra-long reads. Their 2022 assembly, T2T-CHM13, is 3.055 billion bases and gapless for every chromosome except Y, which CHM13 lacks. A complete Y from a different person, HG002, followed in 2023.

One complete genome still represents one person. In 2023 the Human Pangenome Reference Consortium released a draft **pangenome**, the second film’s branching reference: 47 people’s genomes, each with its two chromosome copies assembled separately, joined into a graph in which each person’s copy is a path. It added 119 million bases of variable euchromatic sequence missing from GRCh38.

## What varies between two genomes, and why the reference matters

Differences come in sizes. A **single-nucleotide variant** (SNV) changes one base. A small **indel** inserts or deletes a few bases. A **structural variant** is larger, by a common convention 50 bases or more, and includes deletions, duplications, inversions, insertions and translocations; when a deletion or duplication changes the number of copies of a segment, it is a **copy-number variant**. The chromosome-scale versions are in [chromosome changes](https://stemhop.com/lessons/chromosome-changes/), and how such changes arise is in [mutation](https://stemhop.com/lessons/mutation/).

In the 1000 Genomes Project’s 2015 analysis of 2,504 people from 26 populations, a typical genome differed from the reference at 4.1 to 5.0 million sites, and structural variants affected more bases than SNVs and small indels together. Each is a comparison with a reference. Where the reference carries a rare allele, most people look “variant” there; where a sequence is missing from the reference, reads from it have no correct place to align, so they go unmapped or land on similar sequence elsewhere. In the Human Pangenome Reference Consortium’s benchmarks, short-read workflows using the pangenome made 34% fewer small-variant errors and found 104% more structural variants per haplotype (2.04 times as many) than workflows using GRCh38; those figures describe the workflows tested, not every data set.

Short reads detect large variants indirectly, from **depth** (copy number changes the number of reads), **discordant pairs** (two ends of a fragment mapping farther apart than the fragment could be, or in an unexpected orientation) and **split reads** (one read aligning in two pieces across a junction).

### Predict and design

**Predict.** Short paired-end reads from a patient sample are aligned to the reference. The fragments are normally 400 ± 50 bases long. Depth is 30-fold across a chromosome segment but 15-fold across one 10-kilobase stretch within it. Forty read pairs map with their two ends 10.4 kilobases apart across that stretch, and several reads split, aligning partly to each edge. What does this pattern show, and how would you confirm it?

> **One way to do it.** Half depth means one copy instead of two: one chromosome has lost the stretch and the other has not, a heterozygous deletion. The discordant pairs give its size: ends that should lie about 400 bases apart map 10.4 kilobases apart, so about 10.4 - 0.4 = 10 kilobases are missing from the chromosome they came from. The split reads mark the two breakpoints to the base.
>
> Confirm it two ways. Design PCR primers just outside the two breakpoints. On the intact chromosome they lie about 10.4 kilobases apart, too far for a standard short PCR, but on the deleted chromosome they lie a few hundred bases apart, so a short product appears only from the deleted allele; sequence it to read the junction. Long reads across the region would show both alleles directly.
>
> Then ask whether the deletion is unusual at all. Check population variant collections and the pangenome haplotypes: if many people lack this stretch, the reference may carry a less common insertion, and the “deletion” is common. Finding a variant says nothing on its own about any effect on health.

## Going deeper

### The powers of ten in the coverage problem

The full arithmetic behind the read count is 30 x 3.1 = 93, so 30 x 3.1 x 109 = 93 x 109 = 9.3 x 1010. Then 9.3 / 150 = 0.062, and 0.062 x 1010 = 6.2 x 108 reads.

### Where the ideal coverage model breaks

Real libraries cover some regions, such as very GC-rich ones, less evenly than the Poisson model assumes. The treatment here follows Eric Lander and Michael Waterman’s 1988 analysis of random clones; the same reasoning applies to ideal random shotgun sampling.

---

From STEMhop (https://stemhop.com), free lessons by Christopher Deppmann. Lesson text © the authors. Practice questions: https://stemhop.com/lessons/sequencing-and-genomes/
