LSMA

LONDON SCHOOL OF MULTIOMICS & AI

Free course

Free course

0 of 73 lessons

Lesson 3 of 10

Module 1 · Understanding omics

The central dogma, as data

The central dogma, as data

The central dogma, as data

Every omics technology measures one layer of the flow of information inside a cell. Knowing which layer an assay reads is the first step to knowing what its results can and cannot tell you.

⏱ 11 min readModule 1 · Understanding omicsFree
Lesson at a glance6 steps · about 11 min
  1. One flow, several layers
  2. What each step means
  3. Reading the figure, box by box
  4. Every dataset ends up as counts
  5. What that means in practice
  6. How to use the table below

One flow, several layers

The central dogma describes how information moves in a cell: DNA is copied into RNA (transcription), and RNA is read to build proteins (translation). Proteins then do the cell's chemistry, producing and consuming small molecules called metabolites. Each step is a layer you can measure.

In Bioinformatics in practice: the data flood, where it is used and a first analysis you counted the G and C letters in a short DNA sequence. This lesson shows where that DNA sits in the flow of information, and what the other layers add. The same idea runs through the rest of this module: each later lesson takes one layer, explains how it is measured, and shows what its data looks like.

What each step means

  • Transcription. An enzyme called RNA polymerase reads one stretch of DNA, a gene, and builds a matching RNA copy. The DNA is not used up; the same gene can be copied hundreds of times, and the number of copies is how strongly the gene is "switched on".
  • Translation. A ribosome reads the RNA three letters at a time. Each group of three, a codon, names one amino acid, and the chain of amino acids folds into a protein.
  • Metabolism. Many proteins are enzymes. They turn one small molecule into another: glucose into lactate, for example, or ADP into ATP, the cell's energy currency. These small molecules are the metabolites.

A short example of the code. The RNA codon AUG names methionine and is also the usual start signal for translation, so most proteins begin with methionine. Three codons, UAA, UAG and UGA, name no amino acid and tell the ribosome to stop. Four letters taken three at a time give 64 codons for only 20 amino acids, so most amino acids have more than one codon. This is why some DNA changes alter the protein and others leave it exactly the same, a difference that matters when you read variants in later lessons.

When we say "a layer you can measure", we mean one kind of molecule, collected from a sample all at once: all the DNA, or all the RNA, or all the proteins. The word ending -ome means "the whole set": the genome is all the DNA, the transcriptome all the RNA, and so on. Adding -omics gives the name of the field that measures it.

A common slip is to treat the layers as separate experiments that happen to share a sample. They are steps in one process, so a change in one layer often leaves traces in the next. Reading them together is what lets an analyst move from "this gene is mutated" to "and the protein it makes is missing".

Information flows left to right. The further right you measure, the closer you are to what the cell is doing, and the faster the signal changes.

Reading the figure, box by box

  • DNA changes only when the cell acquires a mutation, so it is the most stable layer. That stability is useful: a mutation found in a tumour today was there yesterday and will be there tomorrow. The abbreviations under the box are the usual tests: WGS reads the whole genome, WES reads only the exome (the roughly 1–2% of the genome that codes for proteins), and a panel reads a chosen list of genes.
  • Access is the epigenome: chemical marks on the DNA and on the proteins it is wrapped around. One common mark is methylation, a small chemical tag on the DNA letter C; heavy methylation near the start of a gene usually keeps it silent. ATAC-seq finds the stretches of DNA that are unwrapped and open to be read. These marks can change over days to years, and they are copied when a cell divides.
  • RNA is made and broken down all the time. Many messages last only minutes to hours, so the transcriptome is a snapshot of what the cell is doing at the moment the sample was taken.
  • Protein levels follow RNA only loosely, because each protein is made and destroyed at its own rate. Proteins are measured with mass spectrometry, which weighs fragments of molecules very precisely and works out what they are.
  • Metabolites respond fastest of all, within seconds to minutes, for example after a meal or a drug. They are measured by mass spectrometry or by NMR, which identifies molecules from how their atoms behave in a strong magnet.

A useful way to hold this: the genome is what a cell could do, the epigenome what it is allowed to do, the transcriptome what it is trying to do, and the proteome and metabolome what it is actually doing.

That is why the roughly 200 cell types in your body share one genome. A neuron and a liver cell carry the same DNA, but different genes are open, transcribed and translated.

Take one example. The gene for albumin, the most abundant protein in blood, is present in every cell you own. In liver cells it is open and copied at a very high rate, so liver tissue is full of albumin RNA and albumin protein. In a neuron the same gene sits closed and silent. If you sequenced the DNA of both cells you would see no difference; if you measured their RNA, the difference would be enormous. Which layer you measure decides what you can see.

Go deeper: the dogma is not a one-way street

The arrows in the figure are the usual direction, not the only one. In 1970 two groups, led by Howard Temin and David Baltimore, found an enzyme in certain viruses that copies RNA back into DNA. It was named reverse transcriptase. Viruses such as HIV use it to write their RNA genome into the DNA of the cell they infect, and you will meet the same enzyme in How sequencing works, where labs use it to turn RNA into DNA before sequencing.

Francis Crick, who coined the phrase "central dogma", meant something narrower than DNA to RNA to protein. His point was that once information has reached a protein, it cannot flow back out into DNA or RNA. That part has held up. Many genes also stop at the RNA layer: their RNA does a job itself and is never translated.

Every dataset ends up as counts

Sequencing machines read millions of short fragments, called reads, and a computer maps each one back to a reference genome. Mass spectrometers measure fragments of proteins and metabolites. So almost every omics dataset becomes counts or intensities: how many reads cover a position or fall in a gene, or how strongly a molecule is detected. Much of omics analysis is about turning those numbers into fair comparisons.

Counts are a convenient common currency, but they lose information along the way. A count per gene says how many reads fell in that gene, not which version of the gene's RNA they came from; an intensity says how strongly a molecule was detected, which is not always the same as how much of it was present. Knowing what each number stands for is the first check on any dataset.

What that means in practice

  • A read is a short string of DNA letters, typically 100–150 letters long for the most common machines, such as ACGTTGCA…. The machine does not know where in the genome a read came from.
  • Mapping (also called alignment) is the computer's job of finding where each read matches the reference genome, a standard, previously assembled copy of the genome for that species. Once every read has a position, you can count how many land on each gene.
  • A count is not yet a result. Suppose gene A has 2,000 reads in sample 1 and 1,000 in sample 2. If sample 1 was sequenced twice as deeply, the gene has not changed at all. Long genes also collect more reads than short ones at the same activity. Adjusting for these effects is called normalisation, and it is what "fair comparisons" means. You will do it yourself in Module 5, Sequencing data.

One more practical point. A read that matches several places equally well, for example inside a repeated stretch of DNA, cannot be placed with confidence. Mapping software reports how sure it is of each placement, and many analyses leave such ambiguous reads out of the counts. This is one reason why repetitive regions of the genome are harder to study.

The central dogma, as data · LSMA free course Short reads are placed along a reference genome. Where more reads pile up, coverage is higher. Reads falling inside gene A and gene B are counted. 1 · READS, MAPPED TO THE REFERENCE Gene A Gene B reference genome 2 · COVERAGE ALONG THE GENOME Count per gene Gene A · 6 reads Gene B · 3 reads
Each coloured bar is one read placed where it matches the reference. Stacking them gives coverage (bottom); counting the reads inside each gene gives the numbers an RNA-seq analysis starts from. The grey read falls between genes and is not counted. Schematic, not to scale.

Worked example · sequencing depth

A genome is sequenced with 900 million reads of 150 bases. How many times, on average, is each base read? This is the mean coverage, or depth.

coverage = number of reads × read length ÷ genome size
= 900,000,000 × 150 ÷ 3,100,000,000
= about 44×

Step by step. First find how many DNA letters were read in total: 900,000,000 reads × 150 letters = 135,000,000,000 letters (135 billion). Then share them out over the genome: 135,000,000,000 ÷ 3,100,000,000 = 43.5, which rounds to 44. The "×" is read as "times": 44× means each position was read about 44 times.

Why it matters. Sequencing makes occasional errors, and a mutation may be present in only some of a tumour's cells. Reading each position many times lets you tell a real variant from a one-off mistake. About 30× is a common standard for a human whole genome; tumour samples are often sequenced more deeply. Remember that 44× is an average: some regions will be read far more often and some far less.

How to use the table below

Start from your question, not from the technology. Find the row whose "Question it answers" matches what you want to know, and the "Typical assays" column tells you what to ask a lab for. The last column tells you what kind of numbers will come back, which decides how the data will be analysed: variant calls are lists of differences from the reference, while counts and intensities need the normalisation described above.

For example, if the question is whether a tumour has lost a working copy of a DNA repair gene, start at the genome row. If the question is whether that gene is still being used, the transcriptome row fits better, and the answer arrives as counts per gene.

LayerQuestion it answersTypical assaysWhat the data looks like
GenomeWhich variants does this tumour carry?Whole genome, whole exome, gene panelsVariant calls
EpigenomeWhich genes are open, and which silenced?Methylation arrays, bisulfite sequencing, ATAC-seq, ChIP-seqMethylation levels, peaks
TranscriptomeWhich genes are in use, and how much?Bulk, single-cell and spatial RNA-seqCounts per gene
ProteomeWhich proteins are present and active?Mass spectrometry, antibody panelsIntensities
MetabolomeWhich chemistry is running?Mass spectrometry, NMRIntensities

Two terms in the table appear again later. A peak is a stretch of the genome where many reads pile up, for example where the DNA is open in ATAC-seq; the peak itself is the result. ChIP-seq finds the places in the genome where a chosen protein is bound, by pulling that protein out of the cell together with the DNA attached to it.

Notice also that one technology can serve several rows. Mass spectrometry measures both proteins and metabolites, and a sequencer reads DNA for the genome, DNA copied from RNA for the transcriptome, and chemically treated or selected DNA for the epigenome. What changes is how the sample is prepared before it reaches the machine, which is the subject of the next lesson, How sequencing works.

The coverage formula in the worked example goes back to Lander and Waterman (1988, Genomics), who modelled reads as landing at random positions along a genome. Their model also predicts the share of bases that no read covers at all, which shrinks very fast as average coverage rises.

Treat it as a best case. Real reads do not land at random: regions rich in G and C, repeats and some PCR-sensitive stretches are covered less well. That is why sequencing reports usually give, alongside the mean coverage, the share of the genome covered at least 10× or 20×. When you quote a coverage figure, say which of the two you mean.

Key point

Cancer starts with changes to DNA, but those changes matter because of what they do further along the flow. A mutation in a gene that is never transcribed may have no effect at all. That is the case for measuring more than one layer.

It also works the other way round: a gene can be switched off with no DNA change at all. In some bowel cancers, the DNA repair gene MLH1 is perfectly intact but heavily methylated, so it is silenced. Sequencing the DNA alone would miss this; a methylation test, or simply measuring MLH1 RNA or protein, shows it at once.

Further reading · Module 1Sources for this module and the rest of the course are gathered on one page. · All reading for this course

Check your understanding

1. Which layer stays most stable across a cell's lifetime?

2. 600 million reads of 150 bases from a 3.1 billion-base genome give about what coverage?

3. You want to know which genes a tumour is using right now. Which assay fits best?

Up next

Lesson 4

How sequencing works

Ask BioTutor

Create a free website with Framer, the website builder loved by startups, designers and agencies.