The graph below is called the Volcano Plot because it looks like a volcano is erupting.
This type of plot was originally used in the field of genetic analysis, and is used to simultaneously visualize the magnitude and significance of differences when comparing the expression levels of multiple genes between groups (e.g., gender or genotype).
However, it can also be used for purposes other than genetic analysis, such as efficiently screening important variables from among a large number of variables. For example, it is effective in situations where many parameters in process data are compared before and after an improvement, and processes with a large impact are extracted.
A volcano plot is a two-dimensional graph that displays statistical quantities such as the magnitude of difference and significance . JMP allows you to dig deeper into the data directly from the graph, making screening more efficient.
JMP 19 also adds new features related to volcano plots. In this article, we will introduce these features and the flow of data exploration using volcano plots in JMP.
What is a Volcano Plot? How to Create It in JMP
This section explains how to create a volcano plot using the JMP sample data set "Drosophila Aging.jmp."
This data is the result of an experiment on Drosophila, and the expression levels (logarithmically transformed to base 2) of 100 genes were measured for each combination of genotype (ORE, SAM), sex (FEMALE, MALE), age (WK1, WK6), and channel (Cy3, Cy5).
Using this data, we will compare the expression levels of 100 genes between genotypes, sexes, ages, and channels, and examine combinations with large differences (in this case, univariate analysis will be performed).
Even if you are not familiar with genetic data, think of it as an example of comparing many quantitative values between groups of people.
operate
Select [Analysis] > [Screening] > [Response Screening] , and specify a column of 100 genes (continuous scale) as [Y, Response Variable], and "Genotype," "Gender," "Age in Weeks," and "Channel" (nominal scale) as [X].
Furthermore, since the volcano plot compares mean differences, it is assumed that the specified Y variables are on a common scale, so you need to check [Y variables have a common scale] . Also, here we will check [Use FDR in volcano plot] and use p-values adjusted for multiplicity of the test (described below).
When the "Response Screening" report is displayed, click the "FDR Log-Worth By Difference" tab in the report to display the Volcano Plot.
In this example,
- Comparison of 100 gene expression levels between genotypes
- Comparison of 100 gene expression levels between sexes
- Comparison of 100 gene expression levels between age groups
- Compare the expression levels of 100 genes between channels
So a total of 400 comparisons were made, and the plot points on the graph correspond to those 400 results.
How to read a volcano plot
The horizontal axis of a volcano plot shows the magnitude of the difference, and the vertical axis shows the significance of the difference. If the difference is large, it may be due to chance alone caused by variability. Therefore, it is necessary to check the statistical significance as well.
Horizontal axis: Mean difference
The horizontal axis shows the difference in mean values between groups.The "Difference" values shown in the "Mean Difference" tab at the bottom of the report are used for the plot.
For example, the first line shows the expression level of log2in_GC6372, calculated as "mean MALE - mean FEMALE."
Therefore, if males have a higher average expression level, it will be displayed on the right, and if females have a higher average expression level, it will be displayed on the left.
Vertical axis: FDR log-worth
Since "Use FDR in Volcano Plot" is checked, the vertical axis plots FDR-adjusted log-worth. This value is calculated using the following steps (①, ②, ③).
①Calculate the p-value of the t-test
For each gene, a t-test p-value was calculated to compare the difference in gene expression between groups.
②Adjust the p-value using FDR for multiple comparisons
Because many comparisons are made, we calculate a p-value adjusted for multiplicity, which prevents significant differences from occurring by chance. FDR (False Discovery Rate) refers to the false positive rate, and is the expected rate at which results judged to be significant actually include those that are not significant (false positives).
➂Calculation of FDR log-worth
The FDR logarithmic value is the FDR adjusted p-value obtained in step 2, converted to -log10 (FDR adjusted p-value). This conversion causes smaller p-values to be displayed higher, making it easier to visualize and check significance.
The "FDR Log Worth" in the "Results Table" is the value calculated using steps ①, ②, and ③ above, and this value is plotted on the vertical axis.
As noted in the legend to the right of the volcano plot, red points indicate positive mean differences with an FDR p-value less than 0.01, and blue points indicate negative mean differences with an FDR p-value less than 0.01.
It's worth focusing on these colored dots.
Digging deeper into data with JMP
Looking at the volcano plot, the five red dots in the upper right corner are of interest. These are comparisons with large mean differences and high significance.
Selecting these points on the graph will also select the corresponding row in the Results Table and Means Table, allowing you to view more details.
To dig deeper, with the selection selected,Select "Bivariate Relationships of Selected Items" from the red triangle button to the left of the report "Response Screening."A "One-Way Analysis" report will be added to the Bivariate Relationships platform.
The figure below shows a report comparing the gene "log2in_CG6372" between the sexes. Looking at this graph, we can visually confirm that the expression level in males is much higher than in females.
Furthermore , the latest version, JMP 19, adds convenient features called "Select X" and "Select Y" to the right of the Volcano Plot. By selecting the X and Y variables to be analyzed here, the corresponding plot points and rows in the summary table are highlighted.
Bottom left: Click "Genotype." The plot points for the genotype comparison are highlighted, but most of the red and blue points are not selected.
Bottom right: Click on "Gender." The plot points for the gender comparison are highlighted, but many of the red and blue points are selected.
This reveals that there are many genes that differ significantly between the sexes.
You can also use "Select X" and "Select Y" simultaneously. The figure below shows an example where the first five genes are selected in "Select Y", and then "Genotype" is selected in "Select X" while holding down the Ctrl key. In this case, one red point is selected. If you move the mouse cursor close to that point, the hover label will indicate that the point is the result of comparing "log2in_TMS1" between genotypes.
summary
The Volcano Plot is a method for visually examining important factors by simultaneously visualizing the magnitude and significance of differences across multiple variables. In JMP, you can select important factors from the Volcano Plot and explore those factors in depth.
Finally, here is a video summarizing the JMP operations you have performed so far (no audio).
by Naohiro Masukawa (JMP Japan)
Naohiro Masukawa - JMP User Community
