The Lineweaver-Burk Plot

What the Lineweaver-Burk plot does

Enzyme rate data is a curve. As you raise substrate concentration [S], the initial reaction velocity v climbs steeply, then bends over and creeps toward a ceiling, V_max. That ceiling is an asymptote — you approach it but never quite hit it — which makes it maddening to read off a graph by eye. The Michaelis constant K_m (the [S] at which v = ½V_max) is even harder to pin down, because it depends on knowing where the asymptote is.

The Michaelis-Menten equation describes that hyperbola:

    v = V_max · [S] / (K_m + [S])

The Lineweaver-Burk trick is one line of algebra: take the reciprocal of both sides. A hyperbola in v-vs-[S] space becomes a perfectly straight line in reciprocal space — and a straight line, unlike an asymptotic curve, can be extended with a ruler until it crosses the axes. Where it crosses tells you V_max and K_m directly, no guessing where an invisible ceiling sits.

The algebra, step by step

Start from Michaelis-Menten and invert. Watch the electron-free bookkeeping — this is pure algebra, but the payoff is a linear form:

  step 1:  v = V_max·[S] / (K_m + [S])

  step 2:  1/v = (K_m + [S]) / (V_max·[S])      ← flip both sides

  step 3:  1/v = K_m/(V_max·[S]) + [S]/(V_max·[S])   ← split the numerator

  step 4:  1/v = (K_m/V_max)·(1/[S]) + 1/V_max        ← simplify

              y   =    m       ·  x     +   b

Line four is the equation of a straight line, y = mx + b, with:

• y-axis = 1/v, x-axis = 1/[S]
• slope m = K_m/V_max
• y-intercept b = 1/V_max (set 1/[S] = 0, i.e. infinite substrate)
• x-intercept = −1/K_m (set 1/v = 0 and solve)

To find the x-intercept, put 1/v = 0: then (K_m/V_max)(1/[S]) = −1/V_max, so 1/[S] = −1/K_m. That intercept lands in negative territory — a physically impossible substrate concentration — which is your first clue that it comes from extrapolating the line beyond the real data, not from a measured point.

Reading the three constants off the line

Once you have the fitted line, three geometric features hand you everything:

1. Read V_max from the y-intercept. The line crosses the vertical axis at 1/V_max. Invert it. If the intercept is 0.02 (mM·s⁻¹)⁻¹, then V_max = 50 mM·s⁻¹. This is the cleanest number the plot gives you, because the y-intercept sits inside or near the real data range.
2. Read K_m from the x-intercept. The line crosses the horizontal axis at −1/K_m. If that intercept is −0.5 mM⁻¹, then K_m = 2 mM. Because this point is an extrapolation off the left end of the plot, it is the least reliable of the three.
3. Cross-check with the slope. Slope = K_m/V_max. Multiply slope × V_max and you should recover K_m; multiply slope × y-intercept-reciprocal and you get K_m again. If the two K_m values disagree wildly, your line is poorly fit.

Worked example: an enzyme with Km = 2 mM, Vmax = 50 mM/s

Suppose you assay an enzyme at several substrate concentrations and measure initial velocities:

[S] (mM)	v (mM/s)	1/[S] (mM⁻¹)	1/v (s/mM)
0.5	10.0	2.00	0.100
1.0	16.7	1.00	0.060
2.0	25.0	0.50	0.040
4.0	33.3	0.25	0.030
8.0	40.0	0.125	0.025

Plot the last two columns (1/v against 1/[S]) and fit a line. You get:

    1/v = 0.040·(1/[S]) + 0.020

    y-intercept = 0.020  →  V_max = 1/0.020 = 50 mM/s
    slope       = 0.040  →  K_m   = slope × V_max = 0.040 × 50 = 2 mM
    x-intercept = −b/m = −0.020/0.040 = −0.5 mM⁻¹  →  K_m = 1/0.5 = 2 mM  ✓

Sanity check against the raw table: at [S] = 2 mM the velocity is 25 mM/s, which is exactly half of V_max = 50 mM/s — and by definition K_m is the substrate concentration giving half-maximal velocity. The plot and the raw data agree: K_m = 2 mM.

Diagnosing inhibitor type — the plot's best modern use

The reason the Lineweaver-Burk plot survives in every biochemistry course is not fitting; it is pattern recognition for enzyme inhibitors. Run the assay with and without inhibitor, overlay the two lines, and the way they move tells you the mechanism instantly.

• Competitive inhibitor (binds the active site, competing with substrate): raises apparent K_m, leaves V_max unchanged. On the plot the lines share a common y-intercept and fan out with steeper slopes. Signature: lines pivot on the y-axis.
• Non-competitive inhibitor (binds an allosteric site, equally to E and ES): lowers V_max, leaves K_m unchanged. Lines share a common x-intercept and rotate about it. Signature: lines pivot on the x-axis.
• Uncompetitive inhibitor (binds only the ES complex): lowers both V_max and K_m by the same factor, so slope is unchanged. Lines are parallel, shifted upward. Signature: same slope, different intercepts.
• Mixed inhibitor: lines cross somewhere in the second quadrant (left of the y-axis, above the x-axis), neither on a clean axis.

These four fingerprints are why students still learn to draw the plot: the visual grammar (pivot on y, pivot on x, or parallel shift) is far easier to read than staring at two overlaid hyperbolas.

Lineweaver-Burk vs the other linearizations

Michaelis-Menten can be rearranged into several straight-line forms. They all give K_m and V_max, but they weight experimental error very differently:

	Lineweaver-Burk	Eadie-Hofstee	Hanes-Woolf
Axes plotted	1/v vs 1/[S]	v vs v/[S]	[S]/v vs [S]
Linear form	1/v = (Km/Vmax)(1/[S]) + 1/Vmax	v = −Km(v/[S]) + Vmax	[S]/v = (1/Vmax)[S] + Km/Vmax
Slope	Km/Vmax	−Km	1/Vmax
Vmax from	y-intercept (1/Vmax)	y-intercept (Vmax)	slope (1/Vmax)
Km from	x-intercept (extrapolated)	−slope	y-intercept × Vmax
Error distortion	Severe — low-[S] points amplified & high leverage	Moderate — v appears on both axes	Mild — most statistically balanced
Data spread	Clusters near origin, one point far out	Spread reasonably	Spread evenly
Inhibitor diagnosis	Best — clear pivot patterns	Usable	Less intuitive
Modern use	Teaching + inhibitor visual	Occasional fitting	Preferred linear fit

Why you shouldn't fit real data on it

The Lineweaver-Burk plot has a statistical flaw baked into the reciprocal transform, and it is worth understanding precisely:

• Error blow-up at low [S]. Measurement error in v is roughly constant in absolute terms. But 1/v is nonlinear: if v = 2 ± 0.2, then 1/v = 0.50 with an error of about 0.05 (10%); if v = 40 ± 0.2, then 1/v = 0.025 with an error of only about 0.0001 (0.5%). The slow (low-[S]) points, which are already the noisiest, get their errors magnified by the transform.
• High-leverage bad points. Those same low-[S] points land at large 1/[S] — far out on the right of the plot. A point far from the centroid of the data exerts outsized pull on an ordinary least-squares line (high leverage). So the transform simultaneously makes the worst points noisier and gives them the most influence.
• Unequal weighting. Naive linear regression assumes every point has equal error. After the reciprocal transform that assumption is badly violated, so an unweighted fit biases K_m and V_max. Weighted regression can partly fix it, but by then you have done more work than a direct nonlinear fit would take.
• The modern replacement. Fit the raw hyperbola v = V_max[S]/(K_m+[S]) directly by nonlinear least squares (any stats package, GraphPad, R's nls(), Python's scipy.optimize.curve_fit). This weights the raw errors correctly and gives unbiased constants. Reserve Lineweaver-Burk for the classroom and for the inhibitor-diagnosis picture.

Historical discovery

Hans Lineweaver and Dean Burk published "The Determination of Enzyme Dissociation Constants" in the Journal of the American Chemical Society in 1934, while both worked in the U.S. Department of Agriculture's Fertilizer Investigations Unit, Bureau of Chemistry and Soils, in Washington, D.C. The paper presented general graphical methods for testing velocity equations, and the double-reciprocal transform they introduced applied to any Michaelis-Menten enzyme.

The paper became one of the most cited in the history of biochemistry — for a simple reason: in 1934 there were no computers, no nonlinear curve fitting, no least-squares software. Extracting K_m and V_max from a curved plot meant guessing an asymptote by eye. The Lineweaver-Burk transform turned that guess into a straightedge exercise: plot the reciprocals, lay down a ruler, read two intercepts. For half a century it was the standard tool of the enzymologist. Only when desktop computers made nonlinear regression trivial (1980s onward) did its statistical weaknesses matter enough to demote it from a fitting tool to a teaching diagram — the role it still holds today.

Common pitfalls

• Confusing the intercepts. The y-intercept gives 1/V_max; the x-intercept gives −1/K_m. Students routinely swap them. Remember: set 1/[S] = 0 (infinite substrate) to isolate V_max on the y-axis.
• Forgetting to invert. The intercept is 1/V_max, not V_max itself. An intercept of 0.02 means V_max = 50, not 0.02.
• Trusting the far-left extrapolation. The x-intercept sits at a negative 1/[S] where no data exists. A tiny tilt in the line swings it a long way. Read K_m from slope × V_max as a cross-check, or better, from a Hanes-Woolf plot.
• Using it to publish kinetic constants. Reviewers now expect nonlinear regression on the raw hyperbola. A Lineweaver-Burk fit of K_m and V_max in a modern paper reads as dated and statistically naive.
• Misassigning inhibitor type from a noisy overlay. With scatter, competitive (pivot on y) and mixed (crossing left of the y-axis) can look similar. Confirm the pattern with a K_i replot (slope or intercept vs [I]) before naming the mechanism.