I am looking for a relative simplistic pathway database that provides parseable information on which gene is upstream/downstream relative to another gene. Ideally, the datbase should cover human pathways with a decent coverage. It would be even better if there were different kinds of links (binding, positive regulation, negative regulation, PTM, transmport).
I know quite a few pathway databases, but either the required information is not stored or it is somewhere hidden deep in a cryptic XML file beyond my parsing capabilities. Here are some examples, to make the problem more clear:
1) Reactome stores lots of information, but its underlying model is not really based on gene->gene links but rather on states. For example, a certain sub-pathway may start with a protein node, which is connected to another node representing a complex containing this gene, connected to a third node representing an activated version of this complex, connected to an 'anonymous hub' with lots of incoming arrows from different nodes, connected to something that looks like a biological process. This might be a very suitable representation of how things really work in the cell, but it is nearly impossible to use the provided XML representations for reconstructing a simplified pathway such as A--> B---| C---> D
2) There are similar problems with KEGG, Biocarta, NCI/Nature Pathways etc.
3) PathVisio/Pathway Wiki looks like what I need - at first glance. Its graphical representation links genes rather than states, there are positive and negative arrows, the coverage is good, and the XML files are simple enough that even I can understand them. Unfortunately, on closer inspection most of the availabe pathways are not usable for my purpose. A few pathways are ok, they contain 'straight arrows' that connect gene nodes and these connections can be easily parsed from the pathvisio format files. Most other pathways have mainly 'dangling arrows' that look good in the graphics but are not connected to the nodes. Others have polygon-style arrows, where each segment is a different entity and no connection information is provided.
Any hint would be appreciated, be it a different database or a method for extracting the required information from the databases mentioned above.
this might be an option, although the prospect of manually editing a few hundred pathways is not so great. Also, WikiPathways often group protein to something called 'complex' in the XML file, although these proteins are certainly not subunits of a complex. Typically, they are collections of proteins that are all affecgted by the same input or otherwise belong together. Thus, it is ok to gorup them, but I find it questionable to use the 'complex' method for this grouping. Other pathway databases use this tag for 'real complexes'.
WikiPathways uses different tags for grouping proteins. If you know a case where the wrong tag is used, please let us know!
By the way, I don't expect that this problem really occurs for 100s of pathways. A dozen at most.
There really shouldn't be that many, especially for the Human pathway collection.