LSMA

LONDON SCHOOL OF MULTIOMICS & AI

Free course

Free course

0 of 73 lessons

Lesson 2 of 10

Module 1 · Understanding omics

Bioinformatics in practice

Bioinformatics in practice

Bioinformatics in practice

Why cheap sequencing made bioinformatics essential, where it is used from hospitals to farms, and a first analysis you can do with a pencil, then see in Python and on the command line.

⏱ 10 min readModule 1 · Understanding omicsFree
Lesson at a glance5 steps · about 10 min
  1. Why the field became essential
  2. Where bioinformatics is used
  3. Your first analysis: GC content
  4. Where the field is heading
  5. How this course builds the skills

Why the field became essential

What bioinformatics is, and what a bioinformatician does described what bioinformatics is. This lesson shows why almost every biology project now needs it, where it is used, and what the work feels like, with a first small analysis you can do with a pencil.

The short answer to "why now?" is cost. Reading DNA has become so cheap that laboratories produce data far faster than people can look at it. The first human genome took 13 years and an international consortium. Today one sequencing machine can read dozens of human genomes in a couple of days.

Cost is not the only driver. The machines are also faster and more flexible, so a single laboratory can now run experiments that once needed a national centre, and public archives collect the results from thousands of such laboratories. Every new dataset is more useful when it can be compared with the ones already stored, and that comparison is itself bioinformatics work.

Bioinformatics in practice: the data flood, where it is used and a first analysis · LSMA free course A line chart with a logarithmic cost axis from 100 dollars to 100 million dollars. The cost of one human genome falls from about 100 million dollars in 2001 to about 10 million in 2007, then drops steeply to about 100 thousand in 2009, 10 thousand in 2011, 1.2 thousand in 2015 and about 500 dollars in 2022. A dashed grey line shows how cost would have fallen if it had only halved every two years; after 2007 the real cost falls far below it. 1 · COST OF ONE HUMAN GENOME (LOG SCALE) $100$1k$10k$100k$1M$10M$100M halving every 2 years $100M$20M$10M$100k$10k$1.2k$500 2001200420072009201120152022 high-throughput sequencing arrives
Figure 1. Approximate cost of sequencing one human genome, rounded from the US National Human Genome Research Institute (NHGRI). After about 2008 the cost fell far faster than computing power improved, so analysis became the bottleneck.

How to read this figure

The cost axis is logarithmic: each grid line is ten times the one below it, so a straight drop of one line means the price fell tenfold. Between 2001 and 2022 the cost fell by roughly 200,000 times. The dashed grey line shows what would have happened if sequencing had only kept pace with computers, whose power roughly doubled every two years (an observation known as Moore's law). From about 2008, when high-throughput sequencing machines arrived, the real cost fell far below that line.

That gap is the reason bioinformatics grew so quickly. When data is cheap and computing is not cheap enough, the slow, expensive step is no longer the experiment. It is the analysis: checking the data, comparing it with what is already known, and deciding what it means.

A common misreading is to take the falling line as a sign that genomics as a whole has become cheap. The figure counts only producing the sequence. Storing it safely for years, analysing it and, in medicine, interpreting each variant for a patient can cost more than the sequencing itself.

Worked example: how much data is one genome?

A human genome has about 3.1 billion letters. To read it reliably, each position is read about 30 times over (30× coverage), because every read has some errors.

3.1 billion bases × 30 = about 93 billion bases of sequence
stored as compressed text files: roughly 50–100 GB for one person

A study of 1,000 people therefore needs tens to hundreds of terabytes before any analysis starts. No spreadsheet can open files like these, which is why Module 3 teaches the command line.

The cost figures come from the NHGRI's long-running series DNA Sequencing Costs: Data (Wetterstrand, genome.gov). They count production costs at large NHGRI-funded sequencing centres: staff, reagents, machines and the first data-processing steps. They leave out interpretation, long-term storage and overheads, so they are not the price a patient or a small laboratory pays. When you cite them, give the page and the date you read it, because the series is updated. For the scale of the data problem, Stephens and colleagues (2015, PLOS Biology, "Big Data: Astronomical or Genomical?") compared genomics with astronomy and online video and predicted it would be among the largest producers of data by 2025.

Where bioinformatics is used

Any field that measures molecules now produces sequence or omics data, so bioinformatics turns up in places that once had little to do with computers. Six examples:

Medicine and rare disease
  • Sequencing a child's genome to find the one variant behind an undiagnosed condition
  • Choosing a drug dose from a patient's genotype (pharmacogenomics)
Cancer
  • Finding the mutations that drive a tumour
  • Matching patients to targeted drugs and tracking tumour DNA in blood
Infectious disease
  • Following a virus as it spreads and changes, as genomic surveillance did for SARS-CoV-2
  • Spotting antibiotic-resistance genes in bacteria
Drug discovery
  • Choosing which protein a new drug should target
  • Predicting protein shapes and how molecules bind to them
Agriculture and food
  • Breeding crops and livestock with the help of DNA markers
  • Tracing food-borne outbreaks back to their source
Ecology and forensics
  • Identifying species from traces of DNA in water or soil (environmental DNA)
  • Matching DNA profiles and studying ancestry and evolution

Two of these examples show how different the same skills can look. In rare disease, the analyst searches one child's genome for a single variant among millions of differences from the reference, so the hard part is filtering. In genomic surveillance of a virus, the analyst compares thousands of virus genomes from many patients to see how the virus is changing and spreading, so the hard part is comparison at scale. Both start by lining reads up against a reference.

The details differ, but the shape of the work is the same in every field. Figure 2 shows that shape.

Bioinformatics in practice: the data flood, where it is used and a first analysis · LSMA free course Six boxes run from left to right joined by arrows: question, sample, raw data, compute, results and answer. The compute and results boxes are highlighted and bracketed as the bioinformatics part of the project, between the laboratory and the answer. 2 · FROM A QUESTION TO AN ANSWER 01QuestionWhat do wewant to know?02SampleBlood,tumour,soil,leaf03Raw dataReads,spectra,images04ComputeClean,align,count,model05ResultsVariants,genes,groups06AnswerDiagnosis,target,decision where bioinformatics works
Figure 2. A typical project. The laboratory produces the raw data; bioinformatics turns it into results that someone can act on.

How to read this figure

Read the boxes from left to right. A project starts with a question and a sample, which a laboratory turns into raw data. The highlighted boxes are the bioinformatics part: computing on the data (cleaning, aligning to a reference, counting, modelling) to produce results such as a list of variants or genes. Only then can a person turn results into an answer: a diagnosis, a drug target or a decision. A weak step anywhere spoils everything after it, which is why bioinformaticians check data quality before anything else (Module 5).

Real projects rarely run straight through. A result that looks wrong often sends the work back a step or two: a sample is sequenced again, a setting is changed, or the question is sharpened. It also helps to involve the bioinformatician before the samples are collected, because decisions made at the sample stage, such as how many samples and which controls, limit what any analysis can find later.

Your first analysis: GC content

Here is a real bioinformatics measurement, small enough to do by hand. The GC content of a DNA sequence is the percentage of its letters that are G or C. It matters because G–C pairs are held together more strongly than A–T pairs, so GC-rich DNA is more stable; GC content also varies between species and affects how evenly a sequencer reads a region.

GC content turns up in everyday checks. Quality-control tools plot the GC content of every read in a run; a clean sample from one species usually gives a single smooth peak, and a second peak can be a sign of contamination from another organism. The GC content of whole genomes also differs widely between species, which is why it is one of the first numbers reported for a newly sequenced genome.

Worked example: count it by hand

Take this 18-letter sequence: GATTACAGCCGTAGGCAT

G + C = (5 G) + (4 C) = 9
GC content = 9 ÷ 18 × 100 = 50%

Check it yourself: underline every G and C, then count. You have just done the same calculation that genome tools run on billions of letters.

You do not need to install or run anything in this module. The two short programs below simply show what the same task looks like in the languages you will meet later. Read them now; you will run code like this from Module 3.

Python
seq = "GATTACAGCCGTAGGCAT"
gc = seq.count("G") + seq.count("C")   # count the G and C letters
print("Length:", len(seq))
print("GC count:", gc)
print("GC %:", round(100 * gc / len(seq), 2))
Length: 18 GC count: 9 GC %: 50.0
Command line
echo "GATTACAGCCGTAGGCAT" | grep -o "[GC]" | wc -l   # one G or C per line, then count the lines
9

Three different routes, one answer. The point is not the code but the habit: a biological question (how GC-rich is this DNA?), a precise definition (G plus C over all letters), and a method that gives the same answer for 18 letters or 3 billion. Real tools add one more decision you will meet again: genomes contain unknown bases written as N, and a careful analysis states whether they were left out of the total.

Go deeper: why a computer wins at scale

Counting 18 letters by hand takes under a minute. Counting the 3.1 billion letters of one human genome at one letter per second would take nearly a hundred years without a break. A short program does it in seconds, because it scans the letters in memory hundreds of millions of times faster than you can read.

Real tools go one step further. They read compressed files a piece at a time, a method called streaming, so even a 100 GB file never has to fit into the computer's memory at once. You will use the same idea on the command line in Module 3, where pipes pass data from one program to the next without saving it in between.

Where the field is heading

Four trends are changing what bioinformaticians do. Each one is taught later in this course.

  • Long reads. Sequencers that read tens of thousands of letters at once make complete, gap-free genomes routine (Module 6).
  • Single-cell and spatial data. Measuring thousands of individual cells, and where they sit in a tissue, multiplies the size and complexity of every dataset (Single-cell and spatial biology, and Module 8).
  • AI models. Models trained on sequences and structures now predict protein shapes and the effects of mutations, and they need careful, honest testing (Module 8).
  • Reproducible pipelines in the cloud. Workflow tools and shared computing let a whole analysis be rerun exactly, by anyone, from the raw files (Module 3).

None of these trends removes the basics. Long reads still need quality checks, single-cell data still needs careful statistics, and AI models are still trained on data from the same public databases. The foundations in the first modules are what let you judge each new method.

How this course builds the skills

The modules follow the path in Figure 2, from knowing what the data measures to analysing it and judging AI results.

In What bioinformatics is, and what a bioinformatician does you saw the same eight modules mapped onto the stages of the field; the table below adds what you will be able to do after each one, and which kind of coding it uses. Coding starts gently: Modules 1 and 2 need none, so you can build the biology and the vocabulary first. Module 3 then turns to the command line.

ModuleYou learn toCoding
1 · Understanding omicsSay what each omics layer measures and what it cannot tell youNone
2 · Biological data and databasesRead file formats, find any gene or protein in the main databases, follow it across them by its identifiers, and search the literatureNone
3 · The command lineWork with large files at the terminal: pipes, awk, scripts, conda, clusters and reproducible workflowsCommand line
4 · Comparing sequencesAlign and search sequences, build a tree and describe a variantPython
5 · Sequencing dataGo from raw reads to a count table, checking quality at each stepCommand line
6 · Genome assemblyAssemble a genome from short and long reads, and judge how complete and correct it isCommand line and Python
7 · From genes to functionFind and annotate genes, read domains and structures, test a gene list for pathways, and see how omics layers are combinedPython
8 · AI for biologyTurn sequences into features, build, evaluate and question machine learning models, and judge AI claimsPython

Key point

Bioinformatics became essential because producing biological data became cheap and analysing it did not. The work always follows the same path, from a question and a sample to data, computation and an answer, and it rests on the habit you used for GC content: define the measurement precisely, then let the computer apply it at scale.

Further reading · Module 1NHGRI: DNA Sequencing Costs: the data behind Figure 1, with notes on what the figures include. · EMBL-EBI Training: free online courses on bioinformatics resources and methods. · All reading for this course

Check your understanding

1. Why did bioinformatics become essential from about 2008?

2. What is the GC content of the sequence GGCCATAT?

3. In the path from a question to an answer, which steps are the bioinformatics part?

Up next

Lesson 3

The central dogma, as data

Ask BioTutor

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