5.2 Venndir Case Studies
While many stylistic pieces were described through this text, putting them together in an artistic way is not always straightforward.
Further, while the concept of directional Venn diagrams may (or may not) sound valuable, it is often best evaluate using real world data.
Venndir Case Studies also serves as a Style Gallery, both to illustrate creative ways to customize a Venn diagram, and to illustrate examples of Venndir in action.
5.2.1 Seehawer Kmt2 Gene Venns
Context
Many genomics studies compare results across experimental conditions using Venn or Euler diagrams—for instance, to identify genes affected by a perturbation such as a treatment, toxin exposure, or disease onset. These studies often aim to uncover the genetic basis of a perturbation, in hopes of revealing ways to mitigate or prevent its effects.
There are thousands of studies that identify "genes affected", and many of them have associated direction of change. For example, "up" may be recognized as "up-regulated" or "increased function", and "down" may be recognized as "down-regulated" or "decreased function."
When two studies are compared, they often focus on the genes involved without regard to the direction of change. The prevailing assumption is that affecting the same genes implies affecting the same biological process. However, the very next question is whether the process is affected the same way.
Situation
For studies whose critical findings involve (1) identify important components, and (2) identify direction of those components, such as gain or loss, Venndir proposes two critical points.
- Overlap alone is not enough.
- Directionality matters.
Direction is important in science and medicine, where it often means the difference between disease and treatment.
Most scientific papers use Venn diagrams without directionality, although a subset compare each direction independently, thus ignoring the potential for discordance. It could be discordant, and they may never know.
To date, no Venn software tool indicates overlap and directionality together.
Example Data
The subject of this example is a scientific paper by Seehawer et al (Seehawer et al. 2024) that studies the role of two genes, Kmt2c and Kmt2d, on brain cancer metastasis, the migration property of cancer cells associated with poor clinical prognosis. They studied the effect of losing either gene, and in two very different cell types "168FARN" and "67NR".
Figure 5.7 shows the target figure, which compares the genes affected by loss of each target gene Kmt2c (left) and Kmt2d (right). Each panel compares the genes affected by the two cell types "168FARN" and "67NR". Up-regulated genes are compared on the top row, down-regulated genes are compared on the bottom row.
The authors intended to assess whether the effects of gene loss were similar in the two cell lines, which would imply similar molecular mechanisms. They concluded that the two cell lines, while being fundamentally different, shared the underlying mechanism because they shared some up-regulated genes, and some down-regulated genes.
This approach is fairly common, testing up-regulated and down-regulated genes independently. However, it does not test whether there are genes with discordant change in direction. They effectively only look for consistent direction, without looking at opposing direction.
Figure 5.7: Target Extended Figure 4h from Seehawer et al 2024.
Data for Extended Figure 4h were available via Supplemental Table 4. Tables were filtered for adjusted P-value 0.1, log2 fold change 0.6. Gene symbol and fold change sign were saved for use in Venndir.
Figure 5.8 shows the Venndir re-creation.
Only two values differ and by only by 1 gene each,
which suggests a minor "rounding error" discrepancy in
processing. See the 67NR-specific overlap in
"Kmt2c_KO Upregulated genes" and "Kmt2c_KO Downregulated genes".
(These differences are not cause for concern.)
Overall, Venndir reproduced the Seehawer figure (Seehawer et al. 2024).
## 168FARN Kmt2c.KO 168FARN Kmt2d.KO 67NR Kmt2c.KO 67NR Kmt2d.KO
## 904 763 6002 4759
Figure 5.8: Venndir re-creation of Seehawer Ext. Fig 4h.
However, the real question (the "next question" if you will), is whether there are shared genes with discordant direction.
Figure 5.9 explores the same data using the complete signed data, with up- and down-regulated genes together. The genes are shown: Kmt2c_KO (left) and Kmt2d_KO (right).
Figure 5.9: Venndir re-creation of Seehawer Ext. Fig 4h, using the complete set of up- and down-regulated genes together. The top row shows 'overlap', the second row shows 'agreement', and the third row shows 'concordance'.
Findings:
The first row shows regulated genes without regard to direction.
- For Kmt2c_KO (left) and Kmt2d_KO (right), the proportion of shared genes is notably higher than the Seehawer figure.
- The increase in shared genes suggests the increase is due to genes with discordant direction.
The second row displays the counts by 'agreement', using '=' for
agreement, and 'X' for disagreement.
- There are nearly as many shared genes that agree in direction as genes that disagree in direction.
The bottom row displays the counts summarized by 'concordance', using
arrows to indicate direction, or 'X' for discordant changes.
- There are more genes discordant in direction, than genes sharing either direction alone.
Conclusion:
Directionality provides important context for the interpretation of the Seehawer data (Seehawer et al. 2024). The comparison across "168FARN" and "67NR" was not critical to their main findings, however this particular conclusion, that the two cell lines exhibited consistent molecular mechanisms may warrant more careful review.