Level 4 ?

Benjamin Good

unread,

Jul 24, 2018, 9:42:11 AM7/24/18

to BioPAX discussion group

Is anyone actively working on extending BioPAX towards a level 4 release? I'm curious where those discussions might be hiding.

thanks

-Ben

Gary Bader

unread,

Jul 24, 2018, 12:54:14 PM7/24/18

to biopax-...@googlegroups.com

Hi Ben,
No one is working on BioPAX Level 4. There are a few things that the community wanted to put into Level 4, and a reasonable amount of discussion around these took place, but ultimately there was not enough energy to move those forward to implementation (I think the main reason was reduced funding for pathway database projects). As a result, BioPAX Level 3 has been stable and in general use since 2010.

Here is a list of requested features (I'm probably missing some items)

-refactoring the ontology to be more semantic web friendly. This involves removing classes from BioPAX that are covered by other ontologies and re-use the classes from those other ontologies instead. Data would be linked by URIs
-cell-cell interactions
-try to remove use of memberPhysicalEntity within physicalEntityParticipant - this likely requires refactoring and re-curation of parts of Reactome
-implement generic reactions and pathways, typically used in metabolic pathway representation and other places
-minor bugs and fixes

Why do you ask?

Best,
Gary

> --
> You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discus...@googlegroups.com.
> Visit this group at https://groups.google.com/group/biopax-discuss.
> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/b71bc434-8c06-43e8-9906-5fd2b9b58196%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

Benjamin Good

unread,

Jul 24, 2018, 5:27:13 PM7/24/18

to biopax-...@googlegroups.com, Chris Mungall

Hi Gary,

I'm asking because I've started working with the Gene Ontology folks on their new framework for representing pathways and pathway-like things as what they term Casual Activity Models or GO-CAMs (http://geneontology.org/gocam). We are considering the challenge of integrating several pathway databases using this representation (and seeking money to do so). Working through the problem, it feels a lot like what a BioPAX level 4 (if it came to exist) might end up looking like, especially with respect to the semantic web refactoring you mention above.

Happy to chat more about the project if you are interested,

-Ben

On Tue, Jul 24, 2018 at 9:54 AM, Gary Bader <gary....@utoronto.ca> wrote:

Hi Ben,
No one is working on BioPAX Level 4. There are a few things that the community wanted to put into Level 4, and a reasonable amount of discussion around these took place, but ultimately there was not enough energy to move those forward to implementation (I think the main reason was reduced funding for pathway database projects). As a result, BioPAX Level 3 has been stable and in general use since 2010.

Here is a list of requested features (I'm probably missing some items)

-refactoring the ontology to be more semantic web friendly. This involves removing classes from BioPAX that are covered by other ontologies and re-use the classes from those other ontologies instead. Data would be linked by URIs
-cell-cell interactions
-try to remove use of memberPhysicalEntity within physicalEntityParticipant - this likely requires refactoring and re-curation of parts of Reactome
-implement generic reactions and pathways, typically used in metabolic pathway representation and other places
-minor bugs and fixes

Why do you ask?

Best,
Gary

> On Jul 24, 2018, at 7:23 AM, Benjamin Good <ben.mcg...@gmail.com> wrote:
>
> Is anyone actively working on extending BioPAX towards a level 4 release? I'm curious where those discussions might be hiding.
>
> thanks
> -Ben
>
> --
> You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.

> Visit this group at https://groups.google.com/group/biopax-discuss.
> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/b71bc434-8c06-43e8-9906-5fd2b9b58196%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.

Visit this group at https://groups.google.com/group/biopax-discuss.

To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/95B681D3-B729-4C50-8019-035D8CFEB1EE%40utoronto.ca.

Gary Bader

unread,

Jul 25, 2018, 10:45:00 AM7/25/18

to biopax-...@googlegroups.com, Chris Mungall

Hi Ben,
Awesome - very interesting. The major difference between BioPAX and GO-CAM, as I understand, is that BioPAX focuses on the biochemical process paradigm (reactions) whereas GO-CAM focuses on activity flow representation, to use SBGN's vocabulary of the three types of pathway representation paradigm (the other one being "entity relationship"). We can convert process level to activity flow level (see some example rules for this in the Simple Interaction Format (SIF) section at http://www.pathwaycommons.org/pc2/formats ), but it is lossy and thus not easy to go the other way around in general. That said, some of the more general relationships that can be captured in BioPAX could be used to encode activity flow relationships or other relationship types and may be useful for mixing representations in the same model and converting between them.

If you want to integrate pathway databases and convert everything to CAM like structure, you should look at OmniPath (http://omnipathdb.org/ and see paper attached) as they have done the most comprehensive work in this space to date. If you've evaluated this resource, is there something you'd like that is missing from there?

If you're considering using BioPAX and associated technology for your projects, you should look at:
http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1003194
https://academic.oup.com/bioinformatics/article/30/1/139/235639 (this is what is used by Pathway Commons to convert process description to SIF in the example above).

Yes, happy to chat about all of this - just let me know.

Cheers,
Gary

nmeth.4077.pdf

Benjamin Good

unread,

Jul 26, 2018, 8:10:34 AM7/26/18

to biopax-...@googlegroups.com, Chris Mungall

Hi Gary,

Thanks for the references, very useful stuff here. And yes, you are spot on with the activity flow-like representation. On the technical side, so far I've been using a combination of PaxTools, SPARQL and the OWL-API to generate GO-CAM OWL models from Reactome BioPAX. See e.g. http://noctua-dev.berkeleybop.org/editor/graph/gomodel:-42045782 and more models at http://noctua-dev.berkeleybop.org (search for Reactome or the name of a Reactome pathway). A few slides on the project here: https://www.slideshare.net/goodb/pathways2go-converting-biopax-pathways-to-gocams. One of my nexts steps is to repeat that process for other resources in Pathway Commons. On first look, this is going to be challenging as I am learning that not all BioPAX models are created equal. Perhaps some of that Pattern module will help.

On a quick look at OmniPath, the lack of an option to download the data in any form is a big disappointment. It would take some time to learn the API well enough to see if it can answer the questions that the GO team is interested in.

-Ben

--
You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.
Visit this group at https://groups.google.com/group/biopax-discuss.

To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/EEAC19A3-5084-434B-A3D8-FEBFE0044106%40utoronto.ca.

For more options, visit https://groups.google.com/d/optout.

> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/CALNM3EofRGRpMuSNhxsoO9z8FiQXnDhtiboKfrZv0nWXGwOXpg%40mail.gmail.com.

> For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.
Visit this group at https://groups.google.com/group/biopax-discuss.

To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/EEAC19A3-5084-434B-A3D8-FEBFE0044106%40utoronto.ca.

Peter Karp

unread,

Jul 26, 2018, 12:04:17 PM7/26/18

to biopax-...@googlegroups.com, Benjamin Good, Chris Mungall

Hello Benjamin, Note that the link you provide to GO-CAMs is invalid
(geneontology.org/gocam), although I did find some info
on GO-CAMs via Google and at your link https://www.slideshare.net/goodb/pathways2go-converting-biopax-pathways-to-gocams

What I didn't see was a description of what limitations in BioPAX you are trying to overcome. What do you see as the key deficiencies in BioPAX? Is your goal to create another pathway exchange standard, or is your goal to create a framework for combining pathway information with GO annotations?

P

To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discus...@googlegroups.com.

Visit this group at https://groups.google.com/group/biopax-discuss.

To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/CALNM3Eoxzk6Tps7kVmf6LHP8B8nZBnffSXCziBRXKVd9j8focw%40mail.gmail.com.

Benjamin Good

unread,

Jul 27, 2018, 6:59:22 AM7/27/18

to Peter Karp, biopax-...@googlegroups.com, Chris Mungall, Suzanna Lewis, Paul Thomas

Hi Peter,

Sorry about the link. Its working for me now, just redirects to https://geneontology.cloud/home which is where work in progress is taking place. The GO-CAM project is quite new and not everything is exactly in production mode yet.. Perhaps early to be talking about it on a public mailing list, but I don't think GO folks mind being open about things. (If Paul and Chris don't mind, they could share a preprint of a full paper about to be sent out for review about the project.)

The GO-CAM project was created as a way to improve the approach to knowledge representation used by the GOC. It vastly extends what can be captured by GO curators without creating gigantic pre-composed terms in the ontology. It is not an attempt to create a new standard for pathway exchange. But, as it happens, GO-CAMs can also represent a lot of the information that currently resides in other existing pathway knowledge bases so we are exploring their use as an integrative representation. The products of this work may simply help GO curators access pathway knowledge as they do their curation or, perhaps, be used to build a large integrated knowledge base by combining other resources.

From the GO perspective, BioPAX alone is not sufficient primarily because of the way it handles the use of external ontologies - as Gary pointed out above as a known problem. The GO is fully embracing the OWL pattern and that means that the places in BioPAX that rely on recommendations that must be interpreted by people or custom software rather than hard logical constraints are problems for us. For pathway knowledge to be integrated with the GO it all needs to be represented using terms from the OBO ontologies and these terms need to be imported in the OWL sense (not referenced as strings..) to allow for description logic reasoning and all of the other general purpose tools of the semantic web to work.

While I don't think there was ever any intention to get into the BioPAX business, it does seem like more direct BioPAX/OBO integration might be a benefit for both communities. Would be pleased to know what you think about that.

I'm adding a few more of the leaders of GO-CAM (Suzi L., Paul T.) to the thread here in case they want to chime in further.

-Ben

Gary Bader

unread,

Jul 27, 2018, 10:12:35 AM7/27/18

to biopax-...@googlegroups.com, Peter Karp, Chris Mungall, Suzanna Lewis, Paul Thomas

Hi Ben,
I'm not so worried about the technical issue of using strings in a field vs. URIs to reference classes/instances. For example, it would be very easy to create a version of BioPAX that directly references classes and convert current BioPAX to that new BioPAX. We even planned for that in BioPAX level 3 and organized all the classes that you'd need to convert under the ControlledVocabulary class. The main reason we didn't actually do this for Level 3, is that not all of the needed ontologies were mature enough to rely on (like phenotypes for genetic interactions, and also cell and tissue type)
See: http://www.visualdataweb.de/webvowl/#iri=http://www.biopax.org/release/biopax-level3.owl

Two bigger issues to consider are:
1) Can GO CAM represent everything that BioPAX can represent? Sure, you can trivially represent everything in BioPAX as triples, which in terms of data structure would be similar to GO CAM, but my guess is that won't be very useful for querying or general consumption. For example, I suspect that detailed biochemical information captured in BioPAX would not be easy or useful to represent in GO CAM (happy to be proven wrong). We would need to see some BioPAX examples converted to GO CAM to better understand how much can match or not e.g. https://github.com/BioPAX/specification/tree/master/Level3/examples

2) independent users of ontologies can represent the same or similar types of data in different ways - this is a general problem, not just in BioPAX. In BioPAX, this usually relates to diversity of the representation styles of source database, usually caused by differences in how different groups represent generic concepts e.g. protein A in state A, protein A, the family that includes protein A. Some databases choose to curate at the state level, some at the protein family level - both are correct, but the BioPAX created from each of these will be different and BioPAX doesn't completely solve the semantic integration problem that results (at least it puts everything in the same language). Unfortunately, there are many unsolved issues here, including simple things like that there is no standard database of protein states (pathway databases have the most information about protein states). Also, biologists frequently don't care about these details and just want to see a gene level interaction network where all the detailed information is flattened.

Quick note about OmniPathDb - you should contact the authors of that paper, as they have worked with activity flow data for a long time. Maybe they can make the database available for batch download.

Thanks for initiating this discussion! It's important to figure these things out.

Cheers,
Gary

>> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discus...@googlegroups.com.

>> > > To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discus...@googlegroups.com.

>> > > Visit this group at https://groups.google.com/group/biopax-discuss.
>> > > To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/b71bc434-8c06-43e8-9906-5fd2b9b58196%40googlegroups.com.
>> > > For more options, visit https://groups.google.com/d/optout.
>> >
>> > --
>> > You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

>> > To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discus...@googlegroups.com.

>> > Visit this group at https://groups.google.com/group/biopax-discuss.
>> > To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/95B681D3-B729-4C50-8019-035D8CFEB1EE%40utoronto.ca.
>> > For more options, visit https://groups.google.com/d/optout.
>> >
>> >
>> > --
>> > You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

>> > To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discus...@googlegroups.com.

>> > Visit this group at https://groups.google.com/group/biopax-discuss.
>> > To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/CALNM3EofRGRpMuSNhxsoO9z8FiQXnDhtiboKfrZv0nWXGwOXpg%40mail.gmail.com.
>> > For more options, visit https://groups.google.com/d/optout.
>>
>> --
>> You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

>> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discus...@googlegroups.com.

>> Visit this group at https://groups.google.com/group/biopax-discuss.
>> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/EEAC19A3-5084-434B-A3D8-FEBFE0044106%40utoronto.ca.
>> For more options, visit https://groups.google.com/d/optout.
>>
>>
>> --
>> You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

>> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discus...@googlegroups.com.

>> Visit this group at https://groups.google.com/group/biopax-discuss.
>> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/CALNM3Eoxzk6Tps7kVmf6LHP8B8nZBnffSXCziBRXKVd9j8focw%40mail.gmail.com.
>> For more options, visit https://groups.google.com/d/optout.
>
>
>

> --
> You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discus...@googlegroups.com.

> Visit this group at https://groups.google.com/group/biopax-discuss.

> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/CALNM3Er4%2By6PVj2W6oH%2BwMuTNW9o_C5jXfdwY8MO9NMQhKRraA%40mail.gmail.com.

Gary Bader

unread,

Jul 30, 2018, 9:02:41 AM7/30/18

to Chris Mungall, biopax-...@googlegroups.com, Peter Karp, Suzanna Lewis, Paul Thomas

Hi Chris - thanks for the explanation. Sounds great. I suppose one still needs to figure out the mapping from BioPAX to GO-CAM to develop a converter, which it sounds like Ben is working on. Happy to have more discussion about that.

Also, I just noticed this paper today, from an imex consortium linked group:
https://academic.oup.com/bioinformatics/article/34/15/2684/4953371

"...we are actively working on defining a standard for causal interaction (manuscript in preparation), which will soon be formalized and integrated in SPV"

Gary

> On Jul 28, 2018, at 6:08 PM, Chris Mungall <cjmu...@lbl.gov> wrote:
>
> Hi Gary,
>
> Ben gave a great summary, I don't have much to add, but in answer to your questions:
>
> • No, we don't intend to try and represent everything BioPAX can represent. For example, representing the explicit cardinalities of all protein memberships in a pathway plus all associated states is out of scope for GO. Furthermore, we don't try and represent the inputs and outputs (plus stoichiometry) of reactions in a GO-CAM; instead we just use pre-composed ontology classes where available. The same goes for representing things like state change. E.g. rather than having an input of proteinX and an output of proteinX-in-phosporylated-state we use the GO class for kinase activity and only state the substrate.
>
> • We avoid some of the issues here due to aforementioned difference in scope and modeling. But there are always different ways to state things. We have regular curator meetings where we align on design patterns and try and document some of this, but it's never a solved problem.
>
> Regarding the URI vs string thing. I agree it would be easy to have a modified biopax with this transformation (it would be great to see that in some future biopax standard!). But the modeling difference is a little deeper. Basically, the triples in a GO-CAM correspond to the structure already used in the ontology which gives us a lot of value. For example we can use standard OWL reasoning to detect mistakes in the model, infer more specific GO assignments and to chain together multiple regulatory relationships.
>
> Another general point to make is that the GO-CAM datamodel provides a hopefully easier path from standard GO annotation to more expressive annotation; the model 'degenerates' to the classic simple GO association model naturally. For example, a curator can still express "gene product G has kinase activity" as a 'degenerate' standalone GO-CAM. They can then take one step further and say that this kinase activity is in the cytosol and is part of a kinase cascade; and then state the kinase cascade is part of cell proliferation, for a particular cell type. These would all be disconnected annotations using the standard GO datamodel. Then they can connect together the kinase activity of G to the kinase activity of G2 (which is not possible in the classic GO annotation datamodel).

Benjamin Good

unread,

Aug 2, 2018, 3:39:22 AM8/2/18

to biopax-...@googlegroups.com, Chris Mungall, Peter Karp, Suzanna Lewis, Paul Thomas

Hi Gary,

I'd be curious to hear your thoughts on the current automated mapping we have from Reactome to GO-CAM. Its certainly not complete, where do you think its important to focus on fleshing it out?

There is a Doc describing the process as well as an associated slide deck showing the simple rules used for the mappings. Both are open for comment.

You can browse (and if you sign up, edit) the generated models and download the corresponding OWL files from http://noctua-dev.berkeleybop.org . There should be one model for every human pathway in Reactome there, plus a bunch of other work in progress. Note that upper level pathways just show their children, not all of their children's reactions (avoiding redundancy and giant pathways).

-Ben

> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.

> Visit this group at https://groups.google.com/group/biopax-discuss.
> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/b71bc434-8c06-43e8-9906-5fd2b9b58196%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.

> Visit this group at https://groups.google.com/group/biopax-discuss.
> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/95B681D3-B729-4C50-8019-035D8CFEB1EE%40utoronto.ca.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.

> Visit this group at https://groups.google.com/group/biopax-discuss.
> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/CALNM3EofRGRpMuSNhxsoO9z8FiQXnDhtiboKfrZv0nWXGwOXpg%40mail.gmail.com.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.

> Visit this group at https://groups.google.com/group/biopax-discuss.
> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/EEAC19A3-5084-434B-A3D8-FEBFE0044106%40utoronto.ca.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.

> Visit this group at https://groups.google.com/group/biopax-discuss.
> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/CALNM3Eoxzk6Tps7kVmf6LHP8B8nZBnffSXCziBRXKVd9j8focw%40mail.gmail.com.
> For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.

> Visit this group at https://groups.google.com/group/biopax-discuss.
> To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/CALNM3Er4%2By6PVj2W6oH%2BwMuTNW9o_C5jXfdwY8MO9NMQhKRraA%40mail.gmail.com.
> For more options, visit https://groups.google.com/d/optout.
>

--
You received this message because you are subscribed to the Google Groups "BioPAX discussion group" group.

To unsubscribe from this group and stop receiving emails from it, send an email to biopax-discuss+unsubscribe@googlegroups.com.

Visit this group at https://groups.google.com/group/biopax-discuss.

To view this discussion on the web visit https://groups.google.com/d/msgid/biopax-discuss/57667D64-E013-43B9-8E5A-CD9DA1A1B39E%40utoronto.ca.

Gary Bader

unread,

Aug 2, 2018, 11:14:20 AM8/2/18

to biopax-...@googlegroups.com, Chris Mungall, Suzanna Lewis, Paul Thomas

Hi Ben - great. I will take a look. Emek Demir would also be a good person to review because he (and Ozgun in his group) developed the pathway commons SIF rules and pattern package for paxtools that they are implemented with.

Quick question - molecular complexes are part of GO cellular component. Should all complexes in reactome have a GO cellular component term? Are you using these?

Gary