Direct comparison
Pearson vs. Manders: Colocalization Stats
Pearson's r measures intensity correlation; Manders' M1/M2 measure spatial overlap independent of intensity. When to use each, plus the thresholding caveat.
Written and maintained by CASRAI Editorial Board
Last updated
Ask CASRAI · included with Regulatory Radar
Ask about Pearson vs. Manders: Colocalization Stats
Ask CASRAI answers research-administration questions and cites the passages behind every claim — and says so when the corpus does not cover something, instead of guessing. It comes with a Regulatory Radar subscription at $29 a month, alongside the daily digest of regulatory changes and the dashboard of what changed.
150 questions a day, on this site, over the API, or inside your own tools through the CASRAI MCP server.
Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays free to read, with no account and no card.
How do Pearson's Correlation Coefficient (PCC), Manders' Coefficients (M1, M2) compare side by side?
The table below compares Pearson's Correlation Coefficient (PCC), Manders' Coefficients (M1, M2) across 8 procurement-relevant dimensions, from what it measures through what must be documented for reproducibility.
Side-by-side comparison
| Dimension | Pearson's Correlation Coefficient (PCC) | Manders' Coefficients (M1, M2) |
|---|---|---|
| What it measures | The linear relationship between the two channels' pixel intensity values across the region analyzed — a single value (r, from -1 to +1) describing how well intensity in one channel predicts intensity in the other. | The fraction of signal in one channel that overlaps spatially with signal in the other, calculated separately per channel: M1 is the fraction of channel A's total signal located in pixels where channel B is also present; M2 is the reverse. |
| Sensitivity to intensity | Directly sensitive to intensity covariation — if channel A goes up 2x wherever channel B goes up, that pushes r toward +1, regardless of the absolute intensities. | Deliberately intensity-independent. Two proteins can occupy the exact same pixels at wildly different absolute or relative expression levels and still return high M1/M2, because the calculation only asks whether signal is present above threshold, not how much. |
| Output and interpretation | One coefficient per image pair, from -1 (perfect inverse relationship) through 0 (no linear relationship) to +1 (perfect positive linear relationship). Values near 0 do not necessarily mean no spatial overlap — they mean no linear intensity relationship. | Two coefficients, M1 and M2, each from 0 to 1 (or reported as 0-100%). They are not symmetric: M1 and M2 can differ substantially for the same image pair if one protein has much more total signal than the other. |
| When it is the right choice | When you expect a genuinely proportional relationship between the two signals' intensities — for example, two subunits of the same complex that should be produced and degraded in step with each other. | When you care about spatial co-occurrence regardless of relative intensity — for example, two proteins expressed at very different absolute levels, where you want to know what fraction of a low-abundance protein's signal coincides with a high-abundance one. |
| Effect of a bright, non-overlapping outlier region | Can distort the coefficient substantially, since it is a whole-image linear-regression-style statistic sensitive to intensity outliers. | Less distorted by intensity outliers, since the calculation is pixel-presence-based rather than intensity-weighted, but still sensitive to how the background threshold is set in that region. |
| Thresholding dependency | Affected by background inclusion (unthresholded background pixels near zero in both channels can inflate an apparent positive correlation), but less mechanically dependent on the exact threshold value than Manders' coefficients are. | Directly and mechanically dependent on the threshold: M1/M2 are literally a count of above-threshold overlapping pixels divided by total above-threshold pixels, so shifting the threshold up or down changes the reported coefficient, sometimes substantially. |
| Common software implementation | Reported alongside Manders' coefficients by the same colocalization tools — Fiji/ImageJ's Coloc2 and the JACoP plugin both compute PCC as standard output. | Fiji/ImageJ's Coloc2 and JACoP plugins compute M1/M2 directly; both typically offer Costes' automatic thresholding as one method for setting the background cutoff objectively rather than by eye. |
| What must be documented for reproducibility | Whether background/ROI was excluded before calculating r, and over what region (whole image vs. a manually or automatically defined region of interest). | The exact thresholding method used (manual vs. Costes automatic vs. another method) and the resulting threshold values — since a different threshold on the identical image can produce a materially different M1/M2, this is the single most common source of irreproducible colocalization results in the literature. |
Common questions
Common questions about Pearson's Correlation Coefficient (PCC) vs Manders' Coefficients (M1, M2)
Can I report both Pearson's r and Manders' M1/M2 for the same image set?
+
Yes, and it is common practice to do so, since they answer different questions. A methods section that reports only one is giving readers half the picture — Pearson's r tells them whether intensities track together, and Manders' M1/M2 tell them what fraction of each channel's signal spatially overlaps with the other, independent of intensity.
Does a high Pearson's r mean the two proteins are colocalized?
+
Not by itself. A high r means intensity in the two channels correlates linearly across the image — it does not directly quantify what fraction of either signal spatially overlaps with the other. Two channels can have a strong linear intensity relationship in the regions where they do overlap while one channel also has substantial signal elsewhere with no counterpart in the other channel at all; Manders' coefficients are the statistic that speaks to that overlap fraction directly.
Why do M1 and M2 sometimes differ substantially for the same image?
+
Because they are not symmetric. M1 is the fraction of channel A's signal that overlaps with channel B; M2 is the fraction of channel B's signal that overlaps with channel A. If channel A has far more total above-threshold signal than channel B, a large share of A's signal can sit outside where B is present (low M1) while nearly all of B's smaller signal sits inside A's much larger footprint (high M2) — that asymmetry is expected, not an error.
Why does the choice of threshold matter so much for Manders' coefficients specifically?
+
Because M1/M2 are calculated as a ratio of above-threshold pixel counts, moving the threshold up or down directly changes both the numerator and denominator of that ratio. A looser threshold that admits more low-intensity background pixels as 'signal' can inflate apparent overlap, while a stricter threshold can understate it. This is why the thresholding method (manual, or an automatic method such as Costes' approach) and the resulting cutoff values should always be reported alongside the coefficients — omitting that detail is one of the most common reasons published colocalization results cannot be reproduced from the methods section alone.
Going deeper
