Where the Field Is Heading | Visual AI Learning Guide

Think of the roadmap as starting from where we actually are: single-target drugs and treatments designed around the average patient. Two branches leave that starting point. The upper branch is combination therapy. A tumor can adapt around a drug that hits only one protein, so researchers want molecules that hit two targets at once, or pairs of drugs that work together. The number of possible pairs is in the millions, and that is precisely the kind of oversized search that pattern-finding tools are good at. The lower branch is personalized medicine. Two tumors that look the same under a microscope can be driven by different genetic changes, so instead of treating everyone with the same diagnosis identically, the aim is to read an individual tumor and pick the treatment it most resembles. Notice that both branches leave the same starting point for the same reason: the old approach was too narrow, and both new directions widen the search beyond what a person could work through by hand.

0:00 / 0:00

Auto-play

Two directions are drawing the most attention, and both come from a limitation of the current approach.

The first is combination therapy. Most targeted cancer drugs act on a single protein, and tumors often adapt around a single point of attack, which is one reason responses can be temporary. AI is being used to search for molecules that hit two targets at once, or to predict which pairs of existing drugs are likely to work together. The search space here is enormous — the number of possible drug pairs runs into the millions — and that scale is exactly the kind of problem pattern-finding tools handle well.

The second is personalized medicine. Cancer is not one disease, and two tumors that look identical under a microscope can be driven by different genetic changes. AI is used to read a patient's tumor data and suggest which existing treatment that particular tumor most resembles, rather than treating everyone with the same diagnosis the same way. This is a shift from the average patient to the individual patient, and it depends on having enough data from enough kinds of patients to make the comparison meaningful.

Both directions share a common thread: they use AI to handle a search space too large for a person to work through by hand, whether that space is combinations of molecules or combinations of patient and treatment.