[ddxxdd]

At this time, I mean every hormone and protein flowing in the bloodstream, along with the state of every single cell.

Essentially, enough detail to get a list of symptoms and bodily sensations, and predict the outcome of any potential blood test or urinalysis. Or, alternatively, enough detail to take the results of a blood test/urinalysis, and predict the state of every single tissue in the body.

[bglazer]

I don’t mean to dissuade you, because simulating biology is a truly fascinating task. That said, what you’re describing would be a monumental task, not exactly a free time side project.

Consider protein folding. This occurs millions of times per second in the thirty trillion cells in the human body. We only just now built a machine learning model that can predict some of the conformations of individual proteins.

The crazy thing is that protein folding is the easiest problem in biology. Nice clean training data, static high resolution targets. Cell biology is not that. Cells are noisy and constantly changing, and its insanely difficult to even measure what they're doing. Like every time we try to determine how many different types of cells there are, we come up with a different, larger number than before. The reason for this is that all our measurements of the number of proteins, RNA, and other chemicals in cells are bad, like looking through a distorted broken lens.

Complicating all this is that biology is is insanely coupled across both spatial and temporal scales. Consider the KRAS protein. Its one of the most commonly mutated genes in cancer. The most common mutations in KRAS amount to about a dozen atoms being out of place in the protein, due to a single base change in DNA. This nanoscale change in a single cell propagates up to tumors that interfere with bulk physiology and result in death. Or, more specifically for the brain, consider huntington’s disease. Theres a section of DNA thats repeated in a gene called huntingtin. If you have less than 36 copies of the repeat, the you're fine. More than 36 results in a brain disease that typically takes 30-50 years to even manifest.

So, all that to say that simulating this is insanely hard. Like, you’re trying to write a simulation for a system where we dont know what it’s made of or how its individual parts work. Also, we just empirically know that changing the arrangement of a few atoms can result in massive changes that lay dormant for decades before they suddenly appear.

[heavyset_go]

This is one of those hard problems in computing and medicine. Accurately simulating a neuron, for example, or other cells is beyond our capabilities. We have a difficult time even simulating individual receptors themselves on cells, which includes accurately predicting and simulating ligand binding on proteins.

[PeterisP]

It's something way beyond our current capabilities.

There's some interesting work on "organs on a chip" e.g. arrays of microfluidic cell cultures that allows some experimentation on a combination of a few types of tissue modeling an organ or some aspect of it, but even that's already pushing the state of art and not fully mature at the moment; doing something similar for many (much less all or most) organs is something that doesn't work yet; and these "organs on a chip" are useful because pure simulation is even less mature and is unable to do what these cell cultures can.

The goal you state is interesting, relevant, and potentially achievable, but it's still some decades away.

[heavyset_go]

Tangential to these are organoids[1] that emulate organs at much smaller scales, which I thought was cool.

[1] https://en.wikipedia.org/wiki/Organoid