Organs are built like walled towns.
Loading
How cancer works, drawn as one journey in nine chapters and 56 stages: from the defences a tumour has to breach, through the machinery it hijacks to copy itself, to the ways it feeds, hides, spreads and outlasts treatment. Every stage has its own page with the diagram, the molecular players, the medicines that act there, the escape routes, the tests, the open questions and the evidence.
Each glyph is a stage. Hover for a sentence, tap to open it; the chapter names jump to the chapter below.
Read it top to bottom as a story, or open any stage. On each stage page the diagram lights up when you pick a product, and every node opens the target, term or pathway behind it. Resistance: how tumours escape each drug class continues chapter nine, and the pathway index lists every diagram alone.
Cancer is not the default. A body runs about 37 trillion cells and copies its DNA billions of times a day; almost none of it turns malignant because layer upon layer of defence stands in the way. Every one of these layers has to be breached before a tumour exists, and every one is a place a treatment can help.
Organs are built like walled towns.
The immune system removes abnormal cells all the time.
Before a cell copies or divides, sensors check the DNA.
Every cell carries a demolition kit.
A stressed cell can stop dividing for good but stay alive.
Two master brakes.
Normal cells stop dividing when they touch neighbours or feel a stiff, crowded tissue.
Chromosome ends shorten with every division, a built-in counter that retires cells after roughly 50 divisions.
Cancer is evolution inside a body. Damage writes mutations, a few of them give a cell an edge, its descendants compete, and epigenetic and environmental pressures shape which lineage wins. Viruses and chronic inflammation shortcut the process. Reading this history in a tumour's genome now guides treatment.
Every cause of DNA damage leaves its own fingerprint.
Of the thousands of mutations in a tumour, only a handful drive it: a stuck accelerator (oncogene, one hit) or cut brakes (tumour suppressor, both copies).
A tumour is a population, not a clone.
Cells can change behaviour without changing their DNA sequence, by rewriting the chemical tags that decide which genes are read.
Years before a tumour, whole patches of normal-looking tissue are already colonised by mutant clones (a smoker's airway, sun-exposed skin, Barrett's oesophagus).
Cancer cells use the same engine as normal cells, only stuck at full throttle. The cell cycle, DNA copying, cell division, growth-factor relays, gene transcription, protein synthesis and protein disposal are all ordinary machinery; most drugs in oncology work by jamming one of these parts harder in the cells that lean on it most.
Division runs on a clock of cyclins and CDKs firing in order.
Copying three billion letters exactly once means licensing thousands of start points and firing them in waves.
A scaffold of microtubules pulls one copy of each chromosome to each daughter, and a checkpoint holds the split until every chromosome is hooked on.
Antennas on the surface pair up when a signal lands and switch on the 'divide' relay: RAS to RAF to MEK to ERK.
The 'grow and survive' circuit: PI3K to AKT to mTOR, with PTEN as the off switch.
Cancers run a few genes at deafening volume from super-enhancers, and MYC is the master amplifier.
Genes are read into RNA, then RNA into protein.
Unwanted proteins are tagged with ubiquitin and fed into a shredder.
To survive the damage they generate and the treatments thrown at them, cancer cells rewire death and repair. They block apoptosis, lean on whichever repair crew they have left, eat themselves to survive famine, and stiffen their membranes against oxidative death. Each rewiring is a dependence.
A dam (the mitochondrial membrane) held by guards (BCL-2, MCL-1, BCL-XL) against demolition crews (BAX, BAK).
Double-strand breaks (HR versus end joining), copying errors (mismatch repair), and single damaged letters (base excision, PARP) each have their crew.
Under famine or drug stress a cell eats its own components to survive.
A form of death by rusting: iron-driven oxidation of membrane lipids.
A tumour is a construction site that never stops. It burns glucose fast and dirty, gorges on glutamine and fat, calls for new blood vessels when it runs short of oxygen, and starves the immune cells that share its table. Metabolism is how we image cancer (FDG PET) and increasingly how we starve it.
Cancer cells burn glucose into lactate even with oxygen around: inefficient but fast, and it supplies building blocks.
The tumour's second favourite food: it feeds the energy cycle, donates nitrogen for DNA letters, and makes antioxidants.
Dividing cells need membranes, and membranes are fat.
When oxygen runs low, HIF proteins switch on a survival programme: new vessels, more glucose uptake, escape.
A tumour cannot grow past a couple of millimetres without new blood vessels.
Tumour and immune cells eat from the same plate.
Every tumour that exists has already beaten the immune system once. It hides its antigens, raises checkpoint brakes, keeps T cells out, recruits myeloid bodyguards, soaks the neighbourhood in TGF-β, shields itself from complement and dodges NK cells. Immunotherapy works when it undoes the particular trick a tumour used.
Cells pin fragments of their proteins on MHC molecules like wanted posters.
T cells carry brakes so they do not attack the body.
Three immune weathers: inflamed (T cells inside), excluded (stuck at the edge), desert (none).
Tumours recruit macrophages and immature myeloid cells and re-train them as bodyguards that switch off T cells, build vessels, and obey 'don't eat me' signals such as CD47.
A growth factor that starts as a brake and becomes an accelerator: late in cancer it drives invasion, activates fibroblasts, and walls T cells out of the tumour.
A cascade of blood proteins that punches holes in cells flagged by antibodies and calls in inflammatory cells.
Natural killer cells hunt cells that lost their identity papers (MHC-I) or show stress flags.
Metastasis causes about nine in ten cancer deaths, and no approved drug targets it directly. Cells loosen their grip, cut a path, enter the blood, survive a brutal journey, land where the soil was prepared for them, often sleep for years, and grow in the organ whose welcome matches their programme. Each step is a bottleneck we could learn to hold.
A brick in a wall becomes a nomad: it lets go of its neighbours, packs pumps to spit out drugs, and puts on camouflage.
Grip the scaffolding (integrins), dissolve a path (MMPs), haul forward (myosin), often along tracks that fibroblasts cut first.
Getting into the blood and surviving there kills all but one cell in a thousand.
Before a single cancer cell arrives, the primary tumour sends parcels ahead: vesicles and hormones that recruit bone-marrow cells to a distant organ and turn it into fertile soil.
Disseminated cells can sleep for years, held quiet by their niche and watched by immune cells, then wake after inflammation, injury or ageing.
Breast cancer goes to bone, lung, liver and brain; prostate to bone; colon to liver.
The brain's vessels are sealed tight and fitted with pumps that eject most drugs.
A tumour is a corrupted organ: cancer cells plus the fibroblasts, matrix, vessels, nerves, microbes and immune cells they recruit, and the signals they send to the rest of the body. The ecosystem decides whether drugs and immune cells get in, and it is why the same mutation behaves differently in different tissues.
Tumours keep the body's repair cells in wound-healing mode forever.
Chaotic, leaky vessels create pockets of hypoxia, high pressure that collapses capillaries, and easy exits for cancer cells.
Tumours grow their own nerve supply and use it: adrenaline and acetylcholine signal growth, gliomas form synapses with neurons, and nerve invasion causes pain.
Gut bacteria shape whether immunotherapy works, bacteria inside tumours degrade chemotherapy and inflame tissue, and one strain's toxin leaves a mutational fingerprint.
Tumours send hormones (GDF-15, IL-6) that switch off appetite and melt muscle and fat.
Every cancer drug eventually meets resistance. Tumours change the part the drug binds, take side roads, switch engines, hide in sanctuaries, or pump the drug out. Some cells simply go quiet and wait. Understanding the route a tumour took decides the next move; the resistance atlas lists them class by class.
Five routes back when a pathway is blocked: mutate the target, make more of it, bypass, mutate downstream, or change identity.
A biopsy samples one place at one time; the tumour is many places changing over time.
Even when a drug wipes out 99% of a tumour, a few cells survive without any resistance mutation.
Cancer cells can install pumps that throw chemotherapy back out.
Under a drug that blocks its identity, a tumour can become a different kind of cell, most dramatically a small-cell neuroendocrine cancer that no longer needs the blocked signal.
The chapters above describe what cancer does. This chapter is about what cancer is: the schools of thought that have tried to explain it, from Boveri's chromosomes and the somatic mutation theory to bioelectric patterning, with what each got right, what it got wrong, and the treatments that came from it. The map of how the theories connect draws them as one diagram.
The standard account: cancer begins when a single body cell picks up mutations in the genes that control growth, and its descendants inherit them. It explains why carcinogens are mutagens, why cancer risk runs in some families and why drugs aimed at a mutated gene can work, but it cannot by itself explain why normal tissue full of the same mutations does not become cancer.
When whole cancer genomes were read, tumours turned out to carry thousands of mutations, of which only a handful drive growth; the rest are passengers that happened to be in the cell. The refinement made the somatic mutation theory precise and testable, and it is the basis of genomic profiling and of matching drugs to mutations.
A tumour is a population of cells that mutate, compete and are selected, exactly as species are, and treatment is one more selective pressure. Peter Nowell proposed this in 1976; it explains why tumours are mixtures of clones, why resistance to almost any single drug appears, and why some researchers now try to steer a tumour's evolution rather than eradicate it.
Rather than asking what starts cancer, Hanahan and Weinberg asked what every cancer ends up able to do: keep growing, ignore stop signals, avoid death, live forever, grow vessels, invade, rewire metabolism and hide from the immune system. Each update has absorbed a rival theory into the list, so the hallmarks read as the field's working synthesis rather than a theory of cause.
The idea that a tumour is organised like a tissue, with a small pool of stem-like cells that renew it and a bulk that cannot, so killing the bulk shrinks the tumour but the stem-like cells regrow it. Proved in leukaemia and real in some solid tumours, but the rigid hierarchy gave way to plasticity: ordinary tumour cells can slip back into the stem-like state, especially under treatment.
The proposal that cancer begins not with a mutation but with a reversible change in how genes are switched on and off in a stem or progenitor cell, which then makes later mutations more likely and more dangerous. It explains cancers with almost no mutations and why cells can switch state under treatment, and it produced the epigenetic drugs used in blood cancers.
The oldest theory of cancer, from Theodor Boveri in 1914: tumours arise from cells with the wrong number or arrangement of chromosomes. Peter Duesberg revived it in the 1990s as an alternative to gene mutations. Most solid tumours are indeed aneuploid and chromosome shattering can create several drivers at once, but chromosomal chaos is now read as an accelerator of evolution, not the sole cause.
Otto Warburg noticed a century ago that cancer cells ferment glucose even when oxygen is plentiful and concluded that damaged respiration causes cancer. The observation held and became the basis of PET scanning, but the causal claim did not: most cancers rewire metabolism because mutated signalling demands building blocks, and only a few metabolic enzymes are themselves cancer genes.
Carlos Sonnenschein and Ana Soto argue that cancer is a disease of tissue architecture, not of single cells: carcinogens disrupt the conversation between a tissue's supporting stroma and its lining cells, and disordered growth follows as in a wound or an embryo gone wrong. Mutations are consequences. Their rat experiments are real, but the theory has few followers and no drug of its own.
A tumour is not a lump of cancer cells but a tissue: fibroblasts, vessels and immune cells, recruited by the signals a wound uses and never told to stop. Virchow saw white cells in tumours in 1863; Dvorak called tumours 'wounds that do not heal' in 1986. It explains why chronic inflammation causes about a fifth of cancers and why aspirin, HPV and hepatitis vaccines and anti-angiogenic drugs work.
Stephen Paget asked in 1889 why breast cancer spread to some organs more than blood flow could explain, and answered that a travelling cancer cell (the seed) grows only where the organ (the soil) suits it. Ignored for most of a century, then confirmed: each cancer has favoured destinations, tumours prepare distant organs before cells arrive, and drugs that change the soil reduce bone metastases.
The immune system patrols for cells that have turned malignant and destroys most of them; the tumours we see are the ones that learned to hide. Ehrlich guessed this in 1909, Burnet and Thomas argued it in the 1950s, it was declared dead in the 1970s, and Robert Schreiber's mouse experiments revived it in 2001. Checkpoint inhibitors, which can cure some metastatic melanoma, are its vindication.
Sequencing of healthy skin, gullet and blood shows that by middle age they are patchworks of mutant clones, many carrying classic cancer mutations, yet cancer stays rare until old age. The ageing tissue view says the mutations are there early and it is the tissue that changes: ageing, damage and inflammation alter which clones win. Clonal haematopoiesis in the blood is the best-measured example.
Physicist Paul Davies and astrobiologist Charles Lineweaver proposed in 2011 that cancer is not a new invention by each tumour but the re-awakening of an ancient survival toolkit from the earliest multicellular life, about a billion years old, which is why every cancer behaves in the same few ways. Gene-age studies give it some support; whether it predicts anything a doctor can use is unproven.
Cancer cells feel their surroundings. A stiff, dense matrix or a compressed tissue is not just a symptom but a signal that pushes cells towards malignancy, and pressure inside tumours squeezes vessels shut so drugs and oxygen cannot get in. Bissell and Weaver reverted cancer cells to normal by blocking the matrix; Rakesh Jain made solid stress and vessel normalisation a treatment strategy.
Cells hold a voltage across their membranes, and tissues share these voltages as patterns that guide growth and regeneration. Michael Levin proposes that cancer is a breakdown of this pattern: tumour cells are depolarised, and in tadpoles restoring the voltage with light-controlled ion channels prevented and reversed tumours caused by mutant KRAS. Striking animal results; no human evidence yet.
src/data/mechanics-atlas.ts); every diagram, drug, target, technology and term is an object in the knowledge graph with its own page and sources. The stage pages resolve their players, medicines, escape routes, tests, questions and papers from the graph at build time. Nothing here is medical advice; see about and methodology.