Pathway view is missing edges for cases with dual positive and negative regulation

[cmungall]

consider this model http://noctua.geneontology.org/workbench/noctua-visual-pathway-editor/?model_id=gomodel%3A66b5638000001589

note there are two activities for EIF2S1

when these get rendered in the pathway view one path is truncated: https://www.alliancegenome.org/gene/HGNC:3255

Can also be seen here: https://amigo.geneontology.org/amigo/model/66b5638000001589

somehow the nodes are becoming disconnected

I find the source annotations a bit confusing (dual negative and positive from EIF2AK3 -> EIF2S1 - coming from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1220852/ and https://www.sciencedirect.com/science/article/pii/S0021925819877153?via%3Dihub, but the reason for the dual paths is the paradoxical increase in translation of ATF-4 even though the effect of the upstream phosphorylation is to reduce global transcription, see

in https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5016099/

An potential simpler GO-CAM representation would be

PERK/EIF2AK3 --[NEG]--> EIF2S1/eIF2α --[NEG]-->ATF4

• PERK (EIF2AK3) negatively regulates eIF2α by phosphorylating it at Ser51. This phosphorylation reduces the overall activity of eIF2α in promoting general protein synthesis, acting as a negative signal
• Phosphorylated eIF2α inhibits global protein synthesis by blocking the recycling of eIF2-GDP to eIF2-GTP, which prevents general translation initiation.
  • However, this negative effect on global protein synthesis indirectly leads to a positive regulation of ATF4 translation because ATF4 mRNA contains upstream open reading frames (uORFs) that are selectively translated when general protein synthesis is downregulated.

IMO some of this should be captured with free text on the edges

[cmungall]

note: the ndex view approximates the VPE one https://www.ndexbio.org/viewer/networks/6cee562d-85d2-11ef-af07-005056ae3c32