PFOCR "Copy to PathVisio" sometimes maps wrong genes

[khanspers]

Discussed in wikipathways/wikipathways-help#165

^{Originally posted by eweitz May 3, 2025}
While creating a pathway diagram, I opted to try the convenient "Copy to PathVisio" feature in PFOCR. It seems 17 of 18 gene nodes were correct.

One gene node was subtly but significantly incorrect. The pro-inflammatory "IL-23" was mapped to the anti-inflammatory "IL37".

To reproduce:

1. Go to https://pfocr.wikipathways.org/figures/PMC8901701__in-22-e7-g001.html
2. In "Gene mentions" header, click "Copy to PathVisio" icon
3. Open a text editor, paste
4. Find "IL37"
5. Note hit for: <DataNode TextLabel="IL37" GraphId="a1001" Type="GeneProduct"><Comment>OCR lexicon match: IL-23</Comment>
6. Go to https://en.wikipedia.org/wiki/Interleukin_37
7. Note IL-37 is an anti-inflammatory cytokine
8. Go to https://en.wikipedia.org/wiki/Interleukin_23
9. Note "IL-23 is an inflammatory cytokine"

Interestingly, "IL37" is not found in the "Gene mentions" HTML table that is visible at https://pfocr.wikipathways.org/figures/PMC8901701__in-22-e7-g001.html.

This issue seems like it could plausibly impact pathway diagrams made by PFOCR. I detected the problem when I could not find the "IL37" node in the source pathway image, then did a cursory search on "IL37" and "IL-23" to find their substantially different functions. For reference, the new pathway diagram I'm creating from this PFOCR entry is "Immunotherapies for psoriasis" (WP5537).

PFOCR is a fantastic tool, and its "Copy to PathVisio" functionality is overall quite useful! Hopefully refining its accuracy will speed up pathway creation even more.