CMI Binding Simulator

CMI Binding Simulator

Experimental parameters

= calculated · click ○ to calculate that value instead

Equilibrium curve

–

Models to fit

Results

Experiment design notes

About this simulator

How it works

Each binding dataset is simulated from exact mass balance, using the true ligand Kd (and Hill coefficient, with Add Cooperativity on). Gaussian error is added to every replicate each time you change the experiment, the way a repeat experiment would. The selected models are then fitted to those points by least squares.

With Add Kinetics on, each ligand also gets rate constants, kon and koff. The association of protein and ligand is then calculated for each concentration, and each well is read at the end of the shared association time instead of at equilibrium. A short incubation will produce pre-steady-state results and under report binding at equilibrium. Ticking Dissociation turns this into a surface-bound experiment, such as SPR or BLI: after the association time the ligand is washed out, and each curve decays with koff for the dissociation time. The kinetic curves show the true binding time course without noise; only the equilibrium points carry measurement error. The results table adds, for each ligand, the time every concentration takes to reach 95% of equilibrium.

The equations are in The models, and every symbol and column is defined in Key terms.

What it assumes

  • Equilibrium, or with Add Kinetics a read at the end of the association time, from exact one-step 1:1 binding with instant mixing.
  • One kind of site, binding 1:1 unless Add Cooperativity sets a Hill coefficient (equilibrium only).
  • Full saturation at 1. Fit Bmax to treat the plateau as unknown.
  • Constant Gaussian noise with one SD and no outliers, so points can fall below 0 or above 1.
  • Known, fully active [P]T, unless you fit it in the quadratic model.
  • Depletion in solution, none on a surface. In solution both partners are drawn down as complex forms. Surface-bound (Dissociation ticked), the ligand flows past at a constant concentration and [P]T is not used.
  • One unit throughout. kon is per molar, so with kinetics the unit sets the time scale.

Key terms

Experimental parameters and plots

  • [P]T — total protein concentration, the same at every point.
  • [L]min, [L]max — first and last total ligand concentrations; points are log-spaced between them.
  • [L]T — total ligand added at a point: the x-axis of the equilibrium curve. With depletion, free ligand is less than [L]T.
  • dilution — the fold step between neighbouring points, d = ([L]max/[L]min)1/(points−1). Enter either it or [L]min; the other follows.
  • Points — the number of concentrations in the series, log-spaced from [L]min to [L]max.
  • Reps (replicates) — independent measurements at each concentration, each with its own noise. More replicates tighten the fit, roughly as 1/√n.
  • Noise SD — standard deviation of the Gaussian error added to each replicate, in fraction bound. One value is shared by every ligand.
  • Kd — the true dissociation constant the data were simulated from.
  • kon (M⁻¹s⁻¹), koff (s⁻¹) — association and dissociation rate constants, tied to the Kd by Kd = koff/kon. Enter one; the dot marks the one calculated from the other.
  • t½ — complex half-life, ln 2/koff.
  • Association time — how long protein and ligand are incubated before the read; in solution unless Dissociation is ticked. Shared by every ligand. The equilibrium curve is read at this time, so a short incubation shows up in the fit.
  • Dissociation time — with Dissociation ticked, the simulation becomes a surface binding experiment, such as surface plasmon resonance (SPR) or biolayer interferometry (BLI). The ligand flows over immobilized protein at a fixed concentration, so there is no depletion; after the association time it is replaced by buffer and the bound fraction decays as fend·e−kofft.
  • Fraction bound (f) — the share of protein carrying ligand: the y-axis of both curves, from 0 to 1.
  • mean ± SD — each plotted point is the mean of its replicates, with error bars at their standard deviation.
  • true equilibrium — with kinetics on, the faint curve showing the fraction bound at full equilibrium, for comparison with the read.
  • ft, ft/fend — in the kinetic-curve hover: the fraction bound at time t and, during dissociation, the fraction still bound relative to the start of the wash-out (0.5 at t½).

Results

  • f at [L]max — the true fraction bound at the top of the titration: how far up the curve the experiment is designed to reach.
  • Kd fit — the dissociation constant the model returns.
  • K fit — the Hill model's raw constant, in concentrationn; shown only while n is fitted, when its Kd is K1/n.
  • Hill n — the Hill coefficient: the value entered for each ligand (simulated), or the one the fit recovered (Hill n under Equilibrium fit).
  • [P]T fit — the protein concentration the quadratic model returns when it is fitted rather than fixed. Useful when the active concentration is uncertain; it is only well defined when [P]T is near or above Kd, where depletion shapes the curve, and shows as n.d. (not determined) when the data cannot pin it down. It is in the same unit as everything else.
  • Bmax — the plateau the fitted curve rises to. Hold at 1 for a normalised readout; fit it when the curve may not get there.
  • ± — asymptotic standard error: how tightly this dataset pins the value. A wrong model can fit precisely.
  • RMSE — root-mean-square residual over every replicate point. A correct model lands near the noise SD.
  • Steady-State Results
  • kobs (s⁻¹) — the rate of approach to equilibrium, kon·√((PT+LT+Kd)² − 4PTLT). Without depletion this is kon[L] + koff.
  • teq (95%) — time to reach 95% of the equilibrium fraction bound; ≈ ln 20 / kobs.
  • feq, fend — fraction of protein bound at equilibrium, and at the end of the association time.
  • f/feq (%) — how close a point got to equilibrium by the read (fend/feq in the Steady-State Results). Under 95%, shown in red, means that point is still rising when read.

Binding concepts

  • Cooperativity — how binding at one site changes binding at the next, measured by the Hill coefficient n. At n = 1 sites are independent (or there is only one); n > 1 is positive cooperativity, each binding event making the next easier; n < 1 is negative. n sets the steepness of the curve — 10% to 90% occupancy spans 811/n-fold in free ligand — and is not a count of sites or a stoichiometry.
  • Steady-state — the plateau the association reaches once binding and unbinding balance; for simple 1:1 binding it is the equilibrium the curve is meant to measure. Points read before they get there are pre-steady-state and under-report binding. In the results table, the Steady-state columns give the time to reach it (teq) and how close the top and bottom of the series are at the read (f/feq at [L]max and [L]min).
  • In solution, Surface-bound (SPR/BLI) — the two kinetic formats named beside Add Kinetics. In solution both partners are drawn down as complex forms; on a surface the ligand flows past at a constant concentration.
  • Specific binding, Quadratic 1:1, Hill model — the three models fitted to the equilibrium data; their equations are in The models.

The models

Equilibrium

Each describes one ligand binding one kind of site and gives the fraction bound as f = Bmax·θ, where θ is the occupancy the model predicts and Bmax is the plateau the signal reaches at saturation.

specific bindingf = Bmax LT ⁄ ( Kd + LT )

Hill model (Fit n)f = Bmax LTn ⁄ ( K + LTn ),  Kd = K1/n

quadratic 1:1f = Bmax [ (PT+LT+Kd) − √((PT+LT+Kd)² − 4PTLT) ] ⁄ 2PT

Specific binding and the Hill model assume free ligand equals what you pipetted in; the Hill model adds a slope n for cooperativity, and its Kd is the half-saturating concentration. The quadratic solves mass balance for both partners, so it needs [P]T and is strictly 1:1 — there is no Hill exponent to fit.

Kinetics

The same 1:1 binding, followed through time.

reactionP + L ⇌ PL,  Kd = koff ⁄ kon

rated[PL]/dt = kon[P][L] − koff[PL]

In solution, where C1 and C2 are the roots of the quadratic and C1 is the equilibrium complex (derived in Rate equations with depletion):

association[PL](t) = C1C2(1 − e−kobst) ⁄ (C2 − C1e−kobst)

observed ratekobs = kon√((PT+LT+Kd)² − 4PTLT)

Surface-bound (SPR/BLI), where ta is the association time:

associationf(t) = feq(1 − e−kobst),  kobs = kon[L] + koff

steady statefeq = [L] ⁄ ( Kd + [L] )

dissociationf(t) = f(ta)·e−koff(t − ta)

Rate equations with depletion

1 · The rate equation

Let x = [PL]. Every complex formed uses up one protein and one ligand, so the free concentrations fall as x rises.

free partners[P] = PT − x,  [L] = LT − x

ratedx/dt = kon(PT − x)(LT − x) − koffx

Factor out kon and use Kd = koff/kon:

rearrangeddx/dt = kon[ x² − (PT + LT + Kd)x + PTLT ]

2 · The quadratic and its two roots

The bracket is a quadratic in x, so it factors as (x − C1)(x − C2). With b = PT + LT + Kd:

smaller rootC1 = ( b − √(b² − 4PTLT) ) ⁄ 2

larger rootC2 = ( b + √(b² − 4PTLT) ) ⁄ 2

C1 is the equilibrium complex. It is where dx/dt = 0, the same [PL] the quadratic 1:1 model gives. C2 is a mathematical root only. It is larger than PT or LT, so the system never gets there, but it still shapes the curve. Two identities are useful: C1C2 = PTLT, and C2 − C1 = √(b² − 4PTLT).

3 · Solving it

dx/dt = kon(x − C1)(x − C2) separates and integrates exactly. Starting from x = 0 at t = 0:

association[PL](t) = C1C2(1 − e−kobst) ⁄ (C2 − C1e−kobst)

observed ratekobs = kon(C2 − C1) = kon√(b² − 4PTLT)

4 · Sanity checks

At t = 0, e0 = 1, so the numerator is 0 and [PL] = 0.

As t → ∞, the exponential goes to 0, so [PL] → C1C2/C2 = C1, the equilibrium complex.

With dilute protein (PT ≪ LT + Kd), C1 is tiny next to C2 ≈ LT + Kd, so the C1 term in the denominator drops out and the familiar single exponential remains:

dilute limit[PL](t) ≈ C1(1 − e−kobst),  kobs ≈ kon[L] + koff

Why depletion bends the curve

When protein is comparable to Kd or to the ligand, the C1 term in the denominator matters. As complex forms, free ligand runs down, so binding slows more than a single exponential predicts and kobs is no longer kon[L] + koff. The simulator computes C2 from the square root and then C1 = PTLT/C2, which avoids rounding error when C1 is very small.