[elcritch]

> One of these days, when I have enough free time, I want to develop a particle-level simulation of the human body, so that medical students, doctors, and curious laymen can see these processes in action on their personal computers.

Likely you wouldn't need/want a particle level simulation, but a kinetics level model. Just with kinetics you could draw simulations of "particle concentrations" or other fun things to show medical students, etc, but without the need to simulate the actual chemicals. Though eventually hopefully we could do that too! The AlphaFold breakthrough is a huge advance toward that IMHO.

> If I combine that with an affordable at-home blood test/diagnostic system, we will have a mechanism to bring affordable healthcare to every single person in the country. And perhaps a reliable method to find the true cause of a patient's depression, before prescribing them SSRI's or other psychiatric drugs.

Just getting all the kinetics together would be amazing. You could essentially do what you're talking about of calculating most likely causes (not just correlations). Though brain chemistry != entirety of depression or whatnot. Still just getting even the top 20 chemical cycles mapped from blood tests could be pretty awesome and help bring light to a huge range of conditions.

In my opinion, medicine currently is at a 1.0-1.5 order effects level. As in we can diagnose first order ailments pretty well. But second order effects? Not really, though I said 1.5 in that medical researchers and doctors are slowly figuring out some limited second order effects, like "X medicine with Y gene" or "X medicine with Y medicine" can have good or bad effects.

We're certainly not talking about 3rd order effects.. We have the math, we have the statistics, but it's just hard maths and stats.

[operator-name]

Sadly even a kinetics level simulation for anything remotely complex is far beyond what current compute is capable of. Even simplified models of basic biological processes are immensity complex, and still potentially incomplete. Just as neural networks do not even attempt to reach the level of molecules, (practical) models must necessarily exist at a much higher abstraction.

Although alphafold is a breakthrough, this is less optimisation, more chaos theory.

[ddxxdd]

I'm not going to try to predict the shape of a protein from a chain of amino acids. I'm not going to try to predict the behavior of a human cell from Schrodinger's equation.

My project's scope is more along the lines of:

1. Taking a look at Roche's list of all biochemical pathways in the human body[0].

2. Creating a 3D model of a human body, assuming that all the molecules in the body interact only in the way that the chart prescribes.

3. Create a database of symptoms and sensations in the body, and use the 3D model to determine all the possible ways that the biochemical pathways could go awry in order to create the ailment.

The way I envision it, the model is simply going after low-hanging fruit.

[0] http://biochemical-pathways.com/#/map/1