Differentiation as Branching Flux

From One Precursor Flux to Several Lineages

A precursor population can supply more than one differentiated lineage. To represent this in a cell-population model, first determine the total rate at which precursor cells differentiate, then allocate that flow among the possible branches. The allocation must preserve the accounting meaning of a flow: if every differentiating precursor cell enters exactly one of the represented lineages, the lineage-specific influxes must sum to the total differentiation flux.

This is a statement about cell flow at the point of differentiation. It does not say that the resulting lineage populations will have the same sizes or grow at the same rates. Once cells enter a lineage, its population can change through its own inputs and outflows. The branch fractions determine the influxes, not the later cell counts.

The units make the relation transparent: \(k_D\) has units of inverse time, \(P\) has units of cells, and \(J_D\) and \(J_i\) have units of cells per time. A branch fraction has no units. In this formulation, differentiation is a first-order per-capita process in the precursor population. More generally, a model may specify \(J_D(t)\) directly without assuming a constant \(k_D\); the allocation equation \(J_i=f_iJ_D\) still applies.

Deriving the Flow-Accounting Rule

Suppose a small amount of precursor differentiation occurs over a short interval of duration \(\Delta t\). The total number of differentiating precursor cells is approximately \(J_D(t)\Delta t\). If fraction \(f_i(t)\) enters lineage \(i\), the number entering that lineage is approximately \(f_i(t)J_D(t)\Delta t\). Dividing by the interval gives the lineage influx rate.

For an exhaustive set of branches, summing the lineage influxes gives

This identity is the central accounting check. If the fractions sum to less than one, some differentiation destinations are unrepresented; the missing fraction can be assigned to an explicit unobserved or other branch. If they sum to more than one, the same total precursor flux has been counted more than once, unless the model deliberately represents a different process in which one precursor contributes cells to multiple destinations.

When the precursor has other sources or exits, its population balance must include them separately. For example, with precursor input \(I_P(t)\) and a non-differentiation outflow \(O_P(t)\), the balance is

The branch allocation redistributes \(J_D\) among destinations; it does not create extra precursor outflow. For each lineage, a population balance can likewise be written as its influx minus its own outflow. That balance is useful for distinguishing incoming differentiation flux from the observed change in lineage cell count.

Fractions and Rates That Change Over Time

Branch fractions need not be constant. A treatment, environmental signal, or changing precursor state might alter the allocation among lineages. At each time \(t\), the instantaneous fractions must still be nonnegative, and they must sum to one if the branches are exhaustive at that time. The influx into a branch then changes if either the total differentiation flux or that branch’s fraction changes.

For a time interval, the cumulative number of cells entering lineage \(i\) is the integral of its influx. The interval branch fraction is therefore a fraction of cumulative flow, not generally the arithmetic average of the instantaneous fractions. In particular, times with a larger total differentiation flux contribute more cells to the interval total.

If \(N_D(t_0,t_1)>0\) and all destinations are included, the interval fraction for lineage \(i\) is \(N_i(t_0,t_1)/N_D(t_0,t_1)\). This is a flux-weighted average of \(f_i(t)\). It equals a simple time average only under conditions such as a constant total differentiation flux over the interval.

Checking Flux Against Observed Lineage Counts

A lineage’s observed cell-count change is not automatically its differentiation influx. A population balance includes both influx and outflow. Over an interval, the increase in lineage cell count equals the cumulative influx minus the cumulative outflow. Thus, if outflow is known, the cumulative influx can be recovered by adding it back to the observed change. If outflow is unknown, a count trajectory alone generally cannot identify influx and branch fractions separately.

Assumptions, Alternatives, and Failure Modes

The branch-flux representation is appropriate when the modeled precursor flux has well-defined destinations and the fractions refer to how that flux is allocated. In the basic cell-count interpretation, each differentiating precursor contributes one cell to exactly one branch. If a transition involves cell division, cell fusion, or contributions to multiple lineages, the one-cell-to-one-destination accounting does not apply without modification. Those processes require explicit accounting of how many cells are produced or consumed per event.

An alternative is to assign a separate transition rate \(k_i\) from the precursor to each lineage and write \(J_i=k_iP\). For a set of exhaustive, mutually exclusive first-order branches, this is equivalent to the fraction formulation when \(k_D=\sum_i k_i\) and \(f_i=k_i/k_D\) for \(k_D>0\). The fraction formulation emphasizes how the total flux is partitioned; separate rates can be more convenient when each branch has a directly specified transition rate. In either parameterization, the total precursor outflow must agree with the sum of the represented branch flows.

A practical check is to compute both sides of the accounting identity using the same time basis and units. For an instantaneous model, compare rates such as cells/day. For an interval calculation, compare cumulative counts such as cells over five days. Do not add a rate to a cell count or compare a daily fraction with a cumulative fraction without defining the interval.

• Fractions do not sum to one. If the branches are intended to be exhaustive, investigate omitted destinations, inconsistent estimates, or a calculation error. If a destination is omitted deliberately, represent the residual fraction explicitly or state that the listed fractions are conditional on the represented branches.
• Influx is confused with net accumulation. A lineage can receive cells while its count falls if its outflow is larger. Use the population balance, not the count change alone, to infer influx.
• Fractions are averaged without weighting. When total differentiation changes over time, an unweighted average of instantaneous fractions need not match the interval fraction. Integrate \(f_i(t)J_D(t)\) and divide by the integrated total flux.
• Branch rates double-count precursor cells. If several independently specified rates imply a sum of outflows larger than the modeled total differentiation flux, the parameterization is inconsistent with the intended partition.

Branch fractions also have an identifiability limitation: observing lineage counts does not by itself reveal the allocation of precursor flux if lineage outflows are not known or otherwise modeled. Distinct combinations of influx and outflow can yield the same net count change. The flux equations make the accounting explicit, but they do not supply missing measurements or establish a biological mechanism.