Lesson 1 of 10
Module 1 · Understanding omics
Biology now produces more data than anyone can read. See how bioinformatics grew to meet that problem, what the field covers, the jobs people do in it, and how AI is changing the work.
This course follows the work of bioinformatics from the sample to the answer, across eight modules. It starts with the biology and the measurements, moves to where biological data is stored and how to find it, then to the command line and to comparing sequences. The later modules take raw sequencing reads to a count table, build genomes from reads, link genes to function and end with AI for biology. Module 1 sets the ground: before you can analyse data well, you need to know what each omics method measures and what its numbers mean.
From measurements to answers
Modern biology runs on machines that produce far more numbers than anyone can read. This module introduces each of them in turn, and this first lesson explains the field that makes sense of their output. A sequencer writes billions of letters, a mass spectrometer thousands of spectra, a spatial slide a measurement for every spot on a tissue section. Bioinformatics is the field that turns that output into answers. It builds and uses databases to store biological data, software to analyse it, and statistical methods to decide which patterns are real.
A short working definition: bioinformatics is the use of computers to store, find, compare and interpret biological data, most of all data about molecules. The word "informatics" means the science of handling information, and in biology that information is mostly sequences of DNA, RNA and protein, plus measurements of how much of each molecule is present.
Three things come together in almost every bioinformatics task: biological data, a question about it, and a method for answering that question with a computer. Take a simple example. A doctor sends a blood sample from a child with a suspected inherited disease, and the sequencer returns millions of short DNA reads. Bioinformatics lines those reads up against a reference human genome, lists the places where the child's DNA differs, looks each difference up in databases of known variants, and filters the list down to the few that could explain the illness. No single step is hard to describe, but each one needs software, a database and a decision about what counts as real.
Go deeper: where the word came from
The word "bioinformatics" was coined around 1970 by two Dutch researchers, Paulien Hogeweg and Ben Hesper. They meant the study of how living systems process information, at every level from molecules to ecosystems. Only later, as sequence databases grew, did the word narrow to its modern sense of computing on molecular data.
Much of the groundwork was laid by Margaret Dayhoff, a physical chemist in the USA. From 1965 she compiled the Atlas of Protein Sequence and Structure, a printed book of every protein sequence then known, and so ran what was in effect the first sequence database. She is widely credited with the one-letter amino-acid code you will use throughout this course, such as M for methionine. In the late 1970s her group also built the PAM matrices, tables that score how likely one amino acid is to replace another over evolution.
A short history
The field grew in steps, and each step answered a problem the step before had created. First there were too many sequences to compare by hand, then too many to store on paper, then whole genomes, then whole omics layers.
A table like the one below is a simplification. Many groups worked on each advance at the same time, and the dates mark when a method or resource became widely known rather than the first time anyone tried the idea.
| When | What happened | Why it mattered |
|---|---|---|
| 1950s–1960s | The first protein sequences are read, starting with insulin; printed collections of protein sequences appear | Sequences could now be compared, and similar sequences grouped into families |
| 1970 | The first general method for aligning two sequences; the word "bioinformatics" is first used, in a broader sense | Comparison became a precise calculation rather than a judgement by eye |
| 1970s–1980s | The Protein Data Bank opens (1971); public DNA sequence databases open in Europe, the USA and Japan | Shared, searchable archives that anyone could add to and use |
| 1985–1990 | Fast search programs, FASTA and then BLAST | A new sequence could be searched against a whole database in minutes, later seconds |
| 1995–2003 | The first genome of a free-living organism, a bacterium (1995); the Human Genome Project finishes (2003) | Assembly and gene-finding software became essential, not optional |
| Mid-2000s on | High-throughput sequencing; transcriptomics, epigenomics, single-cell and spatial methods | Data grew faster than computers did, so efficient tools and pipelines became central |
| 2020s | AI predicts protein structures (2020–21); a gap-free human genome (2022); language models trained on sequences | Computers now predict, not only store and compare |
How to read this table
Read it from top to bottom as a chain of cause and effect: the right-hand column of one row is usually the reason for the next row. Some terms need unpacking:
- To align two sequences is to line them up letter by letter, allowing gaps, so that matching positions sit in the same column. The number of matches and gaps gives a score of how similar the sequences are. Module 4 is about exactly this.
- The Protein Data Bank (PDB) stores the 3D shapes of proteins and other large molecules measured in the laboratory. It is one of the oldest biological databases still running, and you will use it in Module 2.
- BLAST (Basic Local Alignment Search Tool) finds the sequences in a database that are most similar to the one you give it. It is fast because it first looks for short exact matches and only then builds a full alignment around them. It is still one of the most used programs in biology.
- "Gap-free" means with no missing pieces. The genome published in 2003 still left out about 8% of the sequence: hard-to-read, repetitive DNA, around the centromeres for example. Long reads (How sequencing works) filled those gaps in 2022.
Two trends run through every row. The amount of data has grown by many orders of magnitude, and the cost of producing it has fallen: a human genome that cost billions of dollars to read the first time can now be sequenced for well under a thousand. The bottleneck has moved from producing data to analysing it.
This shift explains why the field's jobs exist. When reading a genome was the expensive part, most of a project's budget went on the laboratory. Now that sequencing is cheap, a growing share of the time and money goes on storage, computing and the people who can analyse the data and check it.
The key papers behind this table are: Needleman and Wunsch (1970, Journal of Molecular Biology) for aligning two sequences; Altschul and colleagues (1990, Journal of Molecular Biology) for BLAST; Fleischmann and colleagues (1995, Science) for the first genome of a free-living organism, the bacterium Haemophilus influenzae; and Nurk and colleagues (2022, Science) for the gap-free human genome from the Telomere-to-Telomere consortium.
"Finished" in 2003 meant the readable, gene-rich part of the genome. Even the 2022 sequence is one person's genome, from a cell line, so it is a reference, not a description of human variety.
What the field covers
Bioinformatics is broad, but most work falls into six areas. Many projects touch several at once.
- Sequences. Reading, assembling, aligning and comparing DNA, RNA and protein sequences; finding genes and variants. This is the oldest core of the field.
- Structures. Predicting and comparing the 3D shapes of proteins and RNA, and how they bind to each other or to drugs. A protein's shape largely decides what it does.
- Networks. Describing how molecules work together: which proteins touch, which genes switch others on, which enzymes form a metabolic pathway. A network here is a set of items (nodes) joined by links (edges) that stand for interactions.
- Databases. Designing, building and curating the archives that hold all of the above. Curation means experts checking records and adding reliable information to them by hand.
- Software. Writing the programs, libraries and pipelines that everyone else uses, most of them free and open source.
- Literature. Searching and mining millions of papers to find facts that no one person could read. Text mining means software that pulls facts, such as "gene X interacts with protein Y", out of written text.
The areas depend on each other. A newly read sequence is often understood by finding a similar protein whose shape is known, placing it in a pathway, and checking what papers say about it, all through public archives and free programs. That one task touches sequences, structures, networks, literature, databases and software. As you work through the course, it helps to notice which area each lesson belongs to.
Bioinformatics or computational biology?
The two names are often used as if they meant the same thing, and the line between them is blurry. When people do separate them, the distinction is about the main goal.
- Builds and applies tools, databases and pipelines
- Main goal: manage and analyse biological data well
- Example: a faster aligner, or a cleaner variant database
- Uses computing to answer a biological question
- Main goal: understand how a living system works
- Example: a model of how a gene network keeps a daily rhythm
In practice one person often does both in the same week. A useful test is to ask what the finished product is. If it is a tool or a resource that others will use, the work leans towards bioinformatics. If it is a new piece of biological understanding, perhaps from a mathematical model or a simulation, it leans towards computational biology. Job adverts use both words loosely, so read the description rather than the title.
Two short examples show the difference. Building a database that links known human variants to the diseases reported with them is bioinformatics: the product is a resource that others search. Writing a mathematical model of how a population of bacteria becomes resistant to an antibiotic, and testing it against data, is computational biology: the product is an insight into the bacteria. Both rely on the same foundations, and those foundations are what this course teaches.
What a bioinformatician does
"Bioinformatician" covers several quite different jobs. Four common ones are compared below.
- Runs and adapts pipelines on a study's data
- Chooses methods and statistics, and checks quality
- Explains results to biologists and clinicians
- Writes new methods and software
- Tests them, documents them and keeps them working
- Often a research software engineer
- Builds storage, databases and compute systems
- Runs pipelines at scale, often in the cloud
- Keeps data secure, organised and findable
- Runs validated pipelines on patient samples
- Works to laboratory quality standards
- Helps turn variants into a diagnostic report
A pipeline is a fixed series of programs that data passes through, each step taking the output of the one before: for example, quality control, then alignment, then counting. A research software engineer is a professional programmer who works inside research teams. A validated pipeline is one that has been formally tested against samples with known answers, and is not changed without testing again. That is what makes clinical work different: a hospital report about a patient must be reproducible and auditable, so every tool version and setting is fixed and recorded. In the UK, clinical bioinformatics is a recognised healthcare science specialism with its own NHS training route.
A typical week for an analyst shows how the roles overlap. On Monday a laboratory shares new sequencing files; the analyst checks their quality and finds that one sample produced far fewer reads than the rest. Midweek the pipeline runs, and the results are compared with what the biology predicts. By Friday there is a short report with plots, a note on the weak sample and a list of open questions. Much of the job is in these checks and conversations, not only in running software.
The skills behind the roles
- Biology. Knowing what the data measures, and what would be a surprising result. Module 1 is the start of this.
- Data and databases. Knowing where public data lives, and what the file formats and identifiers mean (Module 2).
- Computing. Working at the command line and on servers (Module 3), and programming, mostly in Python or R (from Module 4).
- Statistics. Deciding whether a difference is larger than chance would give, and avoiding traps such as finding patterns in noise when there are far more measurements than samples.
- Communication. Explaining a result, and its uncertainty, to people who will act on it.
Few people are strong in all five at once. Teams combine them, which is why the ability to talk across the biology–computing divide is often the most valued skill of all.
You do not need all five before you start. Most people come to the field from one side, biology or computing, and learn the other along the way. This course builds each skill in the module where it is first needed.
How to read this figure
The top row is the life of a typical project, from left to right. You measure something with an omics method, store and find data in databases, compute on it with software, compare sequences with each other and with a reference, interpret what the genes and proteins do, and learn patterns with AI models. The dashed lines drop to the modules that teach each stage.
The compare stage has three modules because sequences are still the centre of the field. Module 4 compares sequences with each other; Module 5 takes raw reads to a count table; Module 6 builds a genome from reads when there is no reference, then checks how complete and correct it is. Module 7 moves from lists of genes to pathways, networks, protein function and the integration of several omics layers. Module 8 shows where AI fits on top of everything before it: a model is only as good as the data, databases and processing behind it.
Worked example · why the field needs computers
reading one base per second: 93,000,000,000 s ÷ 31,500,000 s per year ≈ 2,950 years
Step by step. Coverage (How sequencing works) is how many reads, on average, overlap each base of the genome; 30× is a common target for a whole human genome. The genome has about 3.1 billion bases, so 30× means about 3.1 × 30 = 93 billion bases of sequence for one person. A year has about 60 × 60 × 24 × 365 = 31.5 million seconds. Dividing 93 billion by 31.5 million gives about 2,950 years for one person reading without rest.
What the answer means. A modest study of 100 patients would take a human reader nearly 300,000 years; software aligns the same reads in hours. But speed is not the only point. Software applies exactly the same rules to every read, records its settings, and can be rerun by someone else to check the result. The bioinformatician's job is less about reading data and more about choosing, checking and explaining the tools that do.
How AI is changing the work
Machine learning means methods that learn patterns from examples rather than following rules written by a person; AI (artificial intelligence) is the wider term. Bioinformatics has used machine learning for decades, to find genes and classify tumours for example. What is new in the 2020s is the scale and the kind of task.
- Prediction instead of measurement. AlphaFold predicts a protein's 3D shape from its sequence, often close to laboratory accuracy. Its public database holds predicted structures for more than 200 million proteins, and the 2024 Nobel Prize in Chemistry recognised protein structure prediction and design. Module 2 shows where to find these predictions, and Module 8 how such models work.
- Language models for sequences. A language model learns which letters tend to follow which by reading huge amounts of text. Trained on millions of protein or DNA sequences instead, the same idea can suggest which mutations a protein will tolerate.
- Assistants for the analyst. General AI chat tools can draft code, explain an error message or summarise papers. They save time, but they can also state wrong facts, invent references or write code that runs and gives a wrong answer. Everything they produce has to be checked, which needs the same skills this course teaches.
The roles above are shifting rather than disappearing. Less time goes on writing routine code, and more on deciding what to ask, checking the output against biology, and judging when a model is outside the data it learned from. A predicted structure, for example, comes with a confidence score for each part of the protein, and reading that score is part of the job.
A useful habit is to treat any AI output like the result of an unfamiliar laboratory test: plausible until checked. For code, that means running it on a small example whose answer you already know. For a fact or a reference, it means finding the original source. For a predicted structure, it means looking at the confidence score before drawing conclusions about any region.
Key point
Bioinformatics is the work of turning biological measurements into reliable answers with databases, software and statistics. The tools change every few years; the habits of checking data, recording methods and asking whether a result makes biological sense do not.
BLAST is more than 30 years old and still used every day, while many newer tools have come and gone. The same will be true of today's AI models. Learn what each kind of tool does and how to check it, and switching to a new tool becomes a small step.
Further reading · Module 1EMBL-EBI Training: free online courses on bioinformatics resources and methods, from first steps upwards. · AlphaFold Protein Structure Database FAQ: what the predictions are and how to read their confidence scores. · All reading for this course
Check your understanding
1. A team builds a simulation of how a gene network keeps a daily rhythm, to explain why the rhythm breaks in a disease. Which label fits the work best?
2. Why can a hospital's clinical bioinformatician not simply update a tool in the diagnostic pipeline to the newest version?
3. An AI chat tool writes a script that runs without errors and produces a neat table. What is the right next step?
Up next
Lesson 2
Bioinformatics in practice: the data flood, where it is used and a first analysis