HumanCyc: direction

[GoogleCodeExporter]

Data source: HumanCyc
Date downloaded: 11 June 2013
Version, if available: 17.1
File name downloaded:
File location:

Data source issue:

Augustin's detailed inspection on the inferred binary interactions revealed 
another data conversion problem in the HumanCyc data. Below reaction of 
HumanCyc has no direction specified in the Conversion and in the Catalysis. And 
they are not in any Pathway or any PathwayStep.

Inline image 1
The reaction ID in PC is 
http://purl.org/pc2/4/BiochemicalReaction_6439e52f8312024fc6b6578968d64b91 
(normalized and merged it into PC db; original URI was 
http://biocyc.org/biopax/biopax-level3BiochemicalReaction127160)

Since there is no direction, ChiBE draws it from left-to-right, and binary 
interaction patterns use it as if it is left-to-right. This default behavior 
works for Reactome and NCI data. But we have a problem here.

When I go to the HumanCyc site and find this reaction (here), I see that it 
says:

β-nicotinamide D-ribonucleotide + diphosphate <=> 5-phospho-α-D-ribose 
1-diphosphate + nicotinamide + H+

The reaction direction shown, that is, A + B ↔ C + D versus C + D ↔ A + B, 
is in accordance with the Enzyme Commission system.

The reaction is favored in the opposite direction.

So they actually know that this reaction is favored in the opposite direction 
but they are not putting this information in the BioPAX output.

Is the data source aware of the issue?

Original issue reported on code.google.com by rod...@gmail.com on 14 Dec 2013 at 4:19

Attachments:

• temp3.png

[GoogleCodeExporter]

After digging a bit more, I just discovered that this is a bug on our side. 

When I check the HumanCyc owl in PC, I see there are 1194 
BiochemicalPathwayStep objects and all of them have a direction. However in the 
merged human model  there are no BiochemicalPathwayStep objects. That is how we 
are losing directions in PC. So there must be some bug in the integration.

Maybe we should write a test that checks if every object in the resource 
databases are also in the integrated model.

Original comment by ozgunba...@gmail.com on 14 Dec 2013 at 5:01

[GoogleCodeExporter]

This tunes out to be our data integration issue and partly due to the fact that 
HumanCyc Pathways all have NULL organism. 

The cPAth2/PC2 persistent BioPAX model does in fact contain all the 
BiochemicalPathwaySteps with directions set for the step Conversions, etc..

To fix:
- set homo sapiens organism in the HumanCyc Cleaner (and as them fix it in the 
next release);
- fix a Paxtools bug, in the path accessors (Conversion/stepProcessOf does not 
work);

Original comment by rod...@gmail.com on 16 Dec 2013 at 4:41

• Changed state: Accepted
• Added labels: Type-Defect
• Removed labels: Product-External-Data

[GoogleCodeExporter]

Pathway.organism property was set (to hs) in the PC2 db and downloads. 
Conversion/stepProcessOf is not fixed yet...

Original comment by rod...@gmail.com on 6 Jan 2014 at 7:58

• Changed state: Started

[IgorRodchenkov]

Seems done.