Experiment: Only include chemicals that are causally connected in a model.

[ukemi]

Because the models are very noisy, we would like to try an experiment in which input and output chemicals are only included in a model if they are:

1. used in a reaction that is not in the middle of a causal chain (starting and ending reactions)
2. Are chemicals that are shared in the causal chain being outputs of an upstream reaction and inputs of a downstream reaction.
   Let's try making these primary inputs and outputs.

[ukemi]

I think this might be unsatisfying in the end. I think not including ATP as an output of glycolysis is a huge omission. We will have to see what happens with the experiment.

[deustp01]

I wonder if we are colliding with the limits of what a reasoning tool can do (at least, a reasoning tool that we know how to build), and so in the spirit of the 95% - 5% rule (95% of the problem can be solved with 5% of the total effort; the rest of the effort is needed to solve the last 5% of the problem), we make sure that it remains permissible to add and remove primary inputs and outputs manually and look for recurring patterns that can be used to improve the accuracy of the reasoning tool for a reasonable amount of effort.

There's also a biology issue here. Even when molecules like ATP/ADP or NAD(H) have a currency role, they are still required for the process to occur: a lack of NAD+ can arise under physiological conditions and stall glycolysis.