Timing: the axis the pathway maps leave out

Every pathway map treats a link as a fact that either holds or does not. None of them record when. Yet rapamycin given once a week and the same rapamycin given every day are not the same intervention, and at least one link in this map does not exist at all under short exposure and appears after a long one.

This page separates two questions that get confused. For a study: how was the intervention delivered over time? For a link in the pathway: does the relationship itself depend on time? The first is a property of an experiment, the second a property of a claim.

How the interventions were delivered

A regimen is recorded only where a sentence in the source says so, and that sentence is stored next to it. Where the paper does not say, the record says not stated rather than guessing a plausible schedule.

A further 50 records sit outside that ladder: 7 where an intervention was given but the schedule is not stated, and 43 where the study administers nothing over time at all (reviews and syntheses). 340 of the 428 studies in the corpus have not been classified yet.

The withdrawal tested row counts studies whose main design is stopping. 1 more (HAL2026) included a recovery period after a different main regimen, so the stopping section below lists 3 studies in all.

Dose versus schedule

Almost every dose-response study asks how much. Far fewer ask how often while holding the amount fixed — and that is the question behind intermittent dosing protocols. Studies in this corpus that ran schedules against each other:

The rest of the intermittent arm tested a single spaced schedule without a continuous comparator, which answers whether the schedule works, not whether it works better.

What happens when you stop

A benefit that disappears on withdrawal is suppression, not repair. That distinction decides whether an intervention is a course of treatment or a commitment, and it is visible only in protocols that included stopping.

On the pathway map, 2 links carry the same finding:

Which links depend on time

Each link in the pathway map cites at least one study. Where those studies establish that the relationship changes with exposure, the link says so. Where no cited study in this corpus settles it, it says that too — and that is the majority.

5 of 121 links have any time dimension recorded. The other 116 are not thereby time-invariant — they are links where no time dependence is recorded in this Atlas. Treating an untested link as unconditional is the error this column exists to prevent.

How the signal itself moves

Everything above is time imposed from outside: how long and how often an intervention was given. There is a second kind of time, inside the cell. mTORC1 activity is not a fixed setting: it rises and falls without any change in the external stimulus, driven by the cell cycle and by the daily clock. Whether that pattern, and not only the average level, decides what the cell does is an open question — it has its own page.

How each study reads the signal

A study can carry more than one readout. Only studies that bear on signal dynamics or on the feedback loops below have been classified so far; a blank is not a snapshot, it is a study nobody has checked yet.

Feedback loops on the map

Walking the arrows of this map, 18 closed routes come back to where they started with a net inhibitory sign: when mTORC1 goes up, something it triggers eventually pushes it back down. They all return through one of 4 feedback arms. A loop like this is a structure that could produce pulses or oscillation. It is not proof that it does: that also needs a delay, a steep enough response and enough gain, and none of those can be read off a diagram.

“Followed in time” means at least one study linked to the arm reports a time series, a cell-cycle-resolved or a 24-hour measurement. It says nothing about whether the arm oscillates.

Why this is a separate axis

The clearest case in the corpus is the link between rapamycin and mTORC2. Rapamycin was described for years as an mTORC1-selective inhibitor. That is true of a short exposure and false of a long one: in some cell types prolonged treatment also disrupts mTORC2. The direction of the arrow does not change — its existence does.

A map without a time column cannot represent that. It has to either draw the link and overstate it, or leave it out and understate it. The same applies to every claim that a compound "works": the schedule is part of the claim, and dropping it is how a finding quietly becomes a general statement it was never entitled to be.

Method

Figures are recomputed on every build by build_timing_page.py from the study corpus and the pathway links. A regimen is only recorded when a sentence in the source states it, and that sentence is stored verbatim alongside the classification, so every entry on this page can be checked against the paper rather than taken on trust. Studies where the evidence did not settle the question were left blank rather than assigned a plausible value.

The schedule comparison above is derived, not curated: it lists studies whose intermittent arm ran alongside a daily one in the same experiment.

Data are CC BY 4.0 — see Data & Citation. The same corpus is measured from a different angle on the evidence audit.