KAIST's ROOT names the loops that lock a cell, and two ways to unlock them

The number that decides whether this is usable at a bench is the size of the kernel ROOT returns, and the account does not give one. More than a thousand positive feedback loops are woven through the kind of intracellular network at issue, each one a molecule that activates others which reactivate it ^[4]. Against that, a kernel of five loops is an experiment somebody can run next quarter. A kernel of two hundred is a shorter list that is still not a target. The phys.org account of the work does not state how many loops the irreversibility kernel contains in any of the models ^[15].

The validation has the same shape. ROOT was run on differentiation and cancer-transition models and, in the team's account, correctly picked out causal circuits that matched known cell-fate determinants ^[9]. That is the right first test and also the softest one: the ground truth existed in the literature before the model ran, so what has been demonstrated is agreement with prior knowledge rather than prediction of something nobody had ^[1]. The interesting run is the one where ROOT names a circuit that was not already on the list and an experiment confirms it.

The two control strategies are not variants of each other, and the difference is operational. Resetting control returns the cell to its pre-change state while leaving the underlying irreversible property in place ^[6]. Since irreversibility means the change persists after the stimulus is gone ^[0], a cell reset this way is a cell that the same stimulus can lock again ^[2]. Reversing control removes the source of irreversibility so the cell can move between states ^[7], but the source is explicit that irreversibility is what makes normal differentiation hold ^[10], so freeing a cell to move is also giving up the mechanism that kept it where it was put ^[3].

Cho describes the achievement as identifying the causal circuits and building a technology to control them and restore cells to their previous condition ^[11]. Worth keeping the tense straight: the reported work is computational logic modelling analysed with systems biology methods ^[3], applied to models including some built from single-cell transcriptome data ^[8], and the account reports no animal or human experiment and no compound ^[12]. What ROOT produces is a shortlist expressed in the coordinates of a Boolean network. Turning a node in that network into something a molecule can bind is the step that is not in this paper, and it is the step that decides whether "reversion" stays a figure of speech.