Volcano Plots for Dummies

Confused by volcano plots? You’re not alone! Here’s a simple guide to this powerful tool for analyzing cytometry datasets. A good use of volcano plots for biomarker scientists is to separate signal from noise. As a rule of thumb, the points on the “top right” and “top left” corners of a volcano (where it gets its name from), are the most interesting.

We’re going to use an easy to understand example: the impact of AI on apartment prices.

We’ll compare apartment prices in October 2022 (one month before ChatGPT’s launch) to May 2025, in 11 different cities, some real (like San Francisco, Seattle, Austin), and some imagined like Analogville and Papertown.

Our guess? That the AI boom increased prices in tech-heavy cities.

Let’s dive into the volcano plot straightaway:

‍

What do you notice? Well first, the X-axis is the effect size (% price change) before AI, and after. And the y-axis is a funny looking equation -log₁₀(p-value). More on that later. But, the short version is: the higher you are on that axis, the more significant the finding.

From a quick glance, seems like apartment prices have increased the most in tech-heavy hubs, like San Francisco, Seattle, and Austin. And prices are dramatically lower in places like Analogville and Papertown. Rural towns, like Greendale and Clearwater are around the 0 in the x-axis, so these seem largely unaffected.

This largely matches our intuition, cities that have a lot of exposure to AI have posted big gains in apartment rents.

Now we’re going to do a little magic on the x-axis.

‍

You’ll see the x-axis is now titled “log₂Fold Change.” What does that mean?

It sounds complicated, but all it means is:

• log₂(2) = +1 → Doubled (100% increase)
• log₂(0.5) = -1 → Halved (50% decrease)
• log₂(1) = 0 → No change

We do this transformation so you can see big fold changes (i.e. 10X or 100X) easily on a small chart. Probably less common (gulp) in renting apartments, but more common in cytometry analysis.

Now, let’s not forget about the y-axis.  That’s a similar log transformation, except this time of p-values. All that does is tell you that higher up in the chart, are more significant p-values (i.e. 0.05, 0.001, 0.0001, etc). This way you can easily figure out if the changes are signal vs noise.

For completeness, let’s look at the actual dataset.

‍

Wrapping up

What does that have to do with cytometry and translational science? Well, instead of a fold-change in prices for real estate, it’s not much of a leap of an imagination to consider analyzing the fold change in cellular population frequencies before and after a drug dose. Or, the fold change in marker expression from a baseline timepoint to an on-treatment timepoint.

If you’re interested in volcano plot functionality to analyze data from your clinical trial, check out the demo datasets at app.teiko.bio.

Dataset:

~440 apartments across 11 areas (35-45 each) - Tech hubs, Rural counties, and Manual towns.

Methodology
We use a volcano plot to identify statistically significant real estate price changes following ChatGPT's launch in November 2022. The volcano plot simultaneously evaluates both effect size (magnitude of price change) and statistical significance (reliability of the observed change).

Statistical Approach: One-sample t-tests compare each area's price changes against the null hypothesis of 0% change. Areas above the dashed line (p < 0.05) show statistically significant changes, while those below represent normal market variation indistinguishable from random noise.

Geographic Categories: Tech markets (expected to benefit from AI adoption), manual towns (paper-based economies potentially disrupted by automation), and rural counties (control group with minimal expected impact). The analysis separates possibly real AI-related effects from background market fluctuation.

‍