By David Stephen who looks at digital therapeutics in this article.

A priority innovation, using AI, in medicine is simply not drug discovery or protein structures, but a solution for the side-effects of medications. There are already several useful medications for different conditions, but some side-effects can be so cataclysmic that eradicating them would be a core leap of AI for humanity.

Why do medications have side-effects? If an answer is that targeting [say] a receptor could be therapeutic but would derail some other functions, then could a problem be the lack of an overall molecular configuration for functions?

Can Digital Therapeutics help?

Simply, say the functions of a particular segment of digestion has a molecular configuration of A1B2C3D4. Where the letters are respective molecules and the figures are their volumes. If, say, the receptors for molecule A is inhibited elsewhere, making it unavailable at the right volume for the digestive function, could it be possible to map the general configuration of molecules to functions and then prospect how to mitigate or eradicate side-effects?

While it is true that the basis of functions are much more complicated, sometimes involving enzymes, proteins, ions, and so forth, it is possible to seek a rough map of functions - by ions and molecules - from several empirical data in biology. This could begin with brain science for psychiatric medications, such that data of conditions, targets, side-effects are explored for patterns, with AI.

Why the brain? The brain is complex but can be functionally reductive. While there are several components of the brain for functions, neurons are correlated with most known functions - and experiences. But, neurons plus their electrical and chemical signals. Electrical signals are ions. Chemical signals are molecules. Neurons and their signals are not just responsible for memory, feelings and emotions, they mechanize regulation of internal senses [or bodily control]. They are also responsible for measures of those functions, or attributes like attention, subjectivity, intent and so forth, conceptually.

Since signals are far flexible than neurons [which are cells], and they [signals] transport [and are directly correlated with] functions, it can be theorized that signals are the basis for functions. So, it is possible to develop a conceptual model of how the mind works by ions and molecules. These could then be useful to develop the basis for several functions, with AI, placing therapies and potential side-effects.

Simply, to mitigate the side-effects of drugs [starting with psychiatry] using AI, it is possible to zero in [conceptually] on the ionic-molecular configurations that specify functions. Then to look at where some of those ions or molecules are present elsewhere in the body or in other brain functions, to predict what or where else might be affected, to determine how to increase the drug candidates for some conditions since configurations are already prognosticated, shaping far-reaching pre-clinical progress with AI.

This can later be extended to other conditions and other parts of the body. However, psychiatry is a medical opacity, needing biomarkers that could be initially computational, and medications with fewer or less intense side-effects.

There is a recent [July, 15, 2025] open question in The New Yorker, Can A.I. Find Cures for Untreatable Diseases - Using Drugs We Already Have?, stating that, "Doctors have long prescribed off-label medications, usually through trial and error or in clinical trials, but now A.I. appears poised to supercharge the practice. Earlier models needed examples of effective drugs to learn, so they were unlikely to identify promising candidates unless treatments already existed. But more advanced models can conduct what's called "zero-shot inference," nominating drug candidates for conditions without any known treatments. The tally of conceivable drug-disease combinations numbers in the tens of milli...