Preclinical models that do not predict people

Dormant cancer cells hide in bone marrow. A lab-built model of that hiding place would let us watch them sleep and wake, and test drugs on them.

Each batch of cells used in an experiment would carry a small digital record showing when it was authenticated, tested for contamination, and how many times it had been grown, attached to the published result.

Nearly all cancer drugs are found by killing fast-growing cells. Sleeping cells survive them. A screen designed around dormant cells would find a different class of drug.

Before any academic compound gets money for pre-trial studies, it would have to show activity in a standard panel of patient-derived tumour models run by an independent centre, so weak candidates are stopped early.

Almost all cancer deaths are caused by spread, yet very little spread tissue is ever studied. A network collecting donated tissue within hours of death would change that.

Failed laboratory experiments are rarely published, so other teams repeat them. A searchable place to deposit them would save years of duplicated work.

Before investing in a full trial, give a few patients a tiny dose of a new compound and use scans and blood tests to see whether it reaches the tumour and hits its target. Fund these small studies as a matter of routine.

Build a large, openly shared dataset of how tumour organoids respond to drug pairs, so that anyone can look up which combinations might work for which tumour type.

Train a model on millions of experiments where genes and drugs were altered, so it can predict the effect of a new combination without running the experiment.

Use existing cell-line and organoid data to score thousands of drug pairs, publish the ranking openly, and only test the top of the list in people.

Build a shared, openly available AI model that has learned how cancer cells respond to genetic and drug perturbations, so any lab can predict what a new drug or combination might do.

You cannot study a cancer without a laboratory model of it, and most rare cancers have none. A funded bank that makes and shares models would unlock research.

Map every state a cancer cell can be in, and how drugs and the surrounding tissue move it between states, into an open computational model anyone can query and improve.

For many rare cancers there is not a single laboratory model in the world, so no one can test drugs. A shared bank with free distribution would change that.

Most mouse studies of cancer drugs do not randomise animals or blind the people measuring tumours, which inflates results. Checking and publishing which institutions do it properly would change behaviour.

There are far more possible drug combinations than can ever be tried in patients. Test thousands on living samples of real tumours, then feed only the winners into adaptive trials.

Tag every cell in a patient's lab-grown tumour with a unique DNA label, give it a drug, and read the labels to see which cells survive. This predicts which resistant clone will emerge.

Drugs that reach brain tissue may still fail to reach the fluid where cancer spreads along the linings. A lab model of that second barrier would let us screen for drugs that cross it.

Cancer usually kills by spreading to bone, liver, lung or brain. Almost all laboratory models grow tumours under the skin instead, where the surroundings are nothing like those organs.

Labs testing new cancer compounds should always include a few well-known drugs as controls and report how those behaved, so results from different labs can be compared.

When doctors discover how a tumour escaped a drug, that finding usually stops at a paper. Recreating it in a model gives everyone a system to test the next drug against.

Drug tests normally use cells from the original tumour. Growing the rarer cells found in blood would test drugs against the cells that are actually travelling.

While patients are treated in a platform trial, their tumour cells grow in a dish and are tested against dozens of drug pairs. The pairs that win in the dish become the next arms.

Some patients cannot have their tumour biopsied safely. Cancer cells captured from a blood sample can sometimes be grown into a model instead.

Lab-grown mini-tumours usually contain only cancer cells. Adding the patient's own immune cells lets researchers test immunotherapy outside the body.

Lab-grown mini-tumours are already being sold to guide treatment, but the tests are not validated like other medical tests. They should be.

Most cancer drugs are tested in mice with no immune system, then given to people who have one. Mice carrying the same patient's immune cells and tumour would be a fairer test.

A rice-grain-sized implant can release small doses of many different drugs into separate spots of a tumour, then be removed so doctors can see which one worked in that person.

Simulating thousands of virtual patients on a computer can suggest which dose and schedule to test, so fewer real patients receive doses that are too high or too low.

After surgery, a tumour with its blood vessels can be connected to a pump and kept alive for hours or days, allowing drugs to be tested in genuinely human tissue.

Damage to the lungs, heart or liver is a common reason cancer drugs fail. Connected chips of human tissue may spot this earlier than animal tests.

Building a genetically engineered mouse for a specific cancer takes years. Editing genes directly in an adult mouse's organ can produce the same tumour in weeks.

Drug candidates are tested for shrinking tumours, almost never for stopping spread. A standard spread test would find anti-metastatic drugs we are throwing away.

A drug that works in one laboratory's mice often fails elsewhere. Running the key animal study across several independent laboratories first would catch this.

Instead of one lab's mouse study deciding whether a drug goes to patients, several labs run the same protocol independently, like a multi-centre clinical trial for mice.

Dogs get cancers that closely resemble human ones, with real immune systems and years of natural history. Treating them, with owner consent, can test drugs in a way mice cannot.

Labs usually use whichever tumour models they already have. A searchable index that finds the model closest to a specific patient's tumour would make experiments more relevant.

Clinical trials must be registered before they start so that failures cannot be hidden. Animal studies used to justify human trials should follow the same rule.

Many mouse experiments use so few animals that the results are unreliable, and papers show one 'representative' result out of several tries. Funders should require proper sample-size planning.

A troubling share of published cancer experiments use cell lines that are contaminated or mislabelled. Requiring a simple identity check before publication would stop this.

A large share of cancer research has been done on cells that were mislabelled or contaminated. A cheap DNA fingerprint test can prove identity; journals and funders should require it.

Instead of testing a drug in lab models first and hoping the results carry over, build the same models from trial participants and run both experiments in parallel to see how well the models predict.

No one keeps score of which laboratory models actually predicted what happened in patients. A public scoreboard would show which models to trust.

For each type of laboratory model, keep a public record of how often its predictions came true in patients, so that researchers know which models to trust for which question.

Robotic labs guided by AI that design experiments on tumour models, run them, read the results and design the next ones, around the clock, with every result published openly.

If every lab had access to the same set of well-characterised tumour models, results could be compared directly instead of each lab using its own private models.

Most cancer patients are older and have other illnesses, but nearly all animal experiments use young healthy mice. Results may not transfer.

A thin slice of a tumour, kept alive for a few days, still contains the immune cells and scaffolding that lab-grown cells lose. Drugs can be tested on it directly.

For very rare cancers there is often no genetic clue and no trial. Growing the patient's cells and testing drugs on them directly can suggest what to try.

Large drugs such as antibody-drug conjugates must cross vessel walls and travel through dense tissue. A chip with flowing channels and human tissue can measure how far they get.

Give patients a short course of one of several drug pairs in the gap before surgery and compare what happened inside the tumours. It is the fastest human test of whether a combination does anything.

Cell therapies are tested for purity and count, but not for whether they can actually kill that patient's tumour. Testing them against the patient's own mini-tumour would show this.

Build a computer model of each patient's cancer and body that simulates how different treatments would go, and prove in a proper trial that choosing treatment with the model helps.

Tumour cells injected into transparent fish embryos grow in days, so several drugs can be compared in about a week, fast enough to help a patient who cannot wait.