[go] Fwd: [Obo-discuss] 3 forwarded messages on MF vs BP from obo-discuss
Chris Mungall
cjm at fruitfly.org
Mon Jan 14 22:20:46 PST 2008
[redirecting this part to GO, sorry if this confuses the thread, but
I want to ask a question of GO folks first]
Thanks for responding Jen.
I think Mike raises some interesting issues, many of which we've
raised internally. I have my ideas on how to answer Mike's question,
focusing on the signal transduction/transducer activity example. But
I'd like to get feedback, particularly from an annotation
perspective, before answering from an ontology-minded perspective.
It seems that we should have on the wiki as part of an annotation SOP
some rules/guidelines for when to annotate to "signal transducer
activity" vs "signal transduction" vs co-annotation of both. (which
would of course apply to analagous examples)
We also need to start making P<->F links here.
If it helps I have added an example query:
http://www.berkeleybop.org/goose
select from the popup "genes annotated to signal transducer activity
but not signal transduction" (and change the DB if you like - it's
set to MGI in the example)
you can also try swapping the terms around
as you can see there is quite a lot
Begin forwarded message:
> From: Mike Bada <mike.bada at UCHSC.edu>
> Date: January 14, 2008 6:21:44 PM PST
> To: jdeegan at ebi.ac.uk, obo-discuss at lists.sourceforge.net
> Subject: Re: [Obo-discuss] 3 forwarded messages on MF vs BP from
> obo-discuss
> Reply-To: obo-discuss at lists.sourceforge.net
>
> Thank you for your detailed reply, Jennifer. However, it seems
> there's
> not even consensus within the GO/OBO Consortium concerning the
> relationship between MFs and BPs. I originally asked whether the MFs
> were inherent functionalities of genes/gene products that get realized
> as processes or as "elementary" processes themselves. What you
> describe
> agrees with the documentation on the GO Web site and with most of
> the MF
> term names and their definitions, i.e., with the latter view, while
> Barry takes the former view.
>
> Again, we are most concerned with all of the parallel terms between
> the
> two subontologies, which makes application toward NLP, among other
> tasks, much more difficult. This is not limited to the large
> number of
> corresponding transporter activity (MF) / transport (BP) terms; some
> other examples include:
>
> signal transducer activity (MF) ; signal transduction (BP)
> translation regulator activity (MF) ; regulation of translation (BP)
> enzyme regulator activity (MF) ; regulation of catalytic activity (BP)
> enzyme activator activity (MF) ; positive regulation of enzyme
> activity (BP)
>
> According to your explanation, all of these pairs of terms seem to
> mean
> essentially the same thing except for the fact that each of the former
> is a single-step process and each of the latter a multi-step process.
> It may be relatively easy to divide a metabolic pathway into its
> component biochemical reactions, but what constitutes an elementary
> single step of a process is often ambiguous. Any process can be
> recursively divided into component processes: Even a biochemical
> reaction can be divided into component processes of the reaction
> mechanism. This ambiguous nature of what an "elementary" step is
> makes
> it very difficult to determine whether a process mentioned in, say, a
> journal article is a single-step or multi-step process (and thus a
> MF or
> a BP, respectively).
>
> So there are two views, both of which hamper usability of the
> subontologies applied in combination: As to the view that MFs are the
> inherent functionalities of genes/gene products that are realized in
> BPs, it is often ambiguous whether a natural-language mention
> refers to
> the inherent functionality or to the process itself. As to the view
> that MFs are the "single-step" processes that make up multi-step
> BPs, it
> is often ambiguous whether a natural-language mention is a "single-
> step"
> or "multi-step" process. Wouldn't it be easier and clearer if there
> were one merged ontology of processes of all granularities and no
> parallel terms such as the aforementioned ones?
>
> Cheers,
> Mike
>
>
> Jennifer Deegan (nee Clark) wrote:
>
>> Hi Mike,
>>
>> Midori forwarded this e-mail to me as I have been working on an
>> extensive overhaul of the transport terms in GO. I think that I
>> can help
>> clear up at least some of the confusion here, though it might take
>> a few
>> e-mails to get things completely clear.
>>
>> In GO, as you have mentioned, we have transport terms in the function
>> and process ontologies, and as you mention, they are really quite
>> similar. However, this is partly because many of these terms were
>> made
>> in the early days of GO when we were really not too clear
>> ourselves on
>> the distinction between function and process. Now that we are much
>> clearer about the distinction I have been working through with a
>> number
>> of expert biologists and improving the terms.
>>
>> Just to be clear, I will give a nice easy example of some function
>> and
>> process terms to get the distinction sorted out. We think of a
>> molecular
>> function as a single step reaction, whilst a process term describes a
>> multistep event. You might think of a process term as being made
>> up of
>> several functions happening one after the other. We currently make no
>> distinction between the actual execution of the function and the
>> potential execution of the function.
>>
>> Function terms
>> ==============
>> An example of a function in GO is when a secondary active transporter
>> molecule acts by changing
>> shape so as to propel a bound potassium ion across the membrane, in a
>> single step chemical reaction. We would consider this reaction to
>> be a
>> molecular function that belongs in the molecular function
>> ontology. The
>> term could be 'secondary active potassium transporter activity',
>> though
>> in fact have the term:
>> potassium ion uniporter activity
>> def: Catalysis of the active transport of a potassium ion across a
>> membrane by a mechanism involving conformational change, where energy
>> for active transport is derived from membrane potential if the
>> solute is
>> charged.
>> There is a picture here in case it helps anyone get the biology
>> clear:
>> http://www.ncbi.nlm.nih.gov/books/bv.fcgi?
>> highlight=carrier&rid=mboc4.figgrp.2000
>>
>> Process terms
>> =============
>> An example of a process that is nice and straightforward is the
>> set of
>> photochemical reactions that occur in photosynthesis.
>> http://www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=mboc4.figgrp.2589
>> This is currently included in GO as 'photosynthesis, light reaction'
>> def: The light reactions of photosynthesis, which take place in
>> photosystems II and I. Light energy is harvested and used to power
>> the
>> transfer of electrons among a series of electron donors and
>> acceptors.
>> The final electron acceptor is NADP+, which is reduced to NADPH.
>> NADPH
>> generated from light reactions is used in sugar synthesis in dark
>> reactions. Light reactions also generate a proton motive force across
>> the thylakoid membrane, and the proton gradient is used to synthesize
>> ATP. There are two chemical reactions involved in the light
>> reactions:
>> water oxidation in photosystem II, and NADP reduction in
>> photosystem I.
>>
>> The thing that I think is confusing you is that we have process terms
>> that look really a lot like the transport function terms. The
>> thing that
>> you need to know is that we have overhauled most of the function
>> terms
>> and given them proper definitions so that they are no longer carrying
>> the very non-specific definitions that you saw before. Most of
>> them now
>> have very clear defs describing single step reactions as shown above.
>> Process transport terms are defined as, for example:
>> The directed movement of potassium ions (K+) into, out of, within or
>> between cells. (This is clearly a multistep affair, if we could
>> think of
>> a good example, though I think all these terms need a good looking
>> at too.
>> For example, do we really need a potassium transport process term?
>> This
>> is on my list to look at soon. )
>>
>> In summary, the function terms are now better defined, to be
>> clearly a
>> single
>> step reaction, and the process terms have much more general
>> definitions
>> that
>> can cover a multistep reaction.
>>
>> The other thing that I should mention, is that I am currently working
>> with several transport experts to re-examine all of the transport
>> terms,
>> and to improve them to clearly represent the current state of the
>> biology.
>> I hope to improve the process terms considerably, and we still have a
>> lot of work to do on the more granular function terms. If you
>> would like
>> to be involved in this then I would be very pleased to include you.
>>
>> Here is a description of what we have done so far:
>> http://wiki.geneontology.org/index.php/Transporters
>> This is the documentation for the parts that we consider to be
>> rigorously logical, and up to date for the field, as it currently
>> stands:
>> http://wiki.geneontology.org/index.php/Docs
>>
>> The process and function ontology docs may also help:
>> http://www.geneontology.org/GO.process.guidelines.shtml
>> http://www.geneontology.org/GO.function.guidelines.shtml
>>
>> I hope this helps a bit. Please do write back to ask further
>> questions.
>> It is a tricky subject, and I have had to rattle this e-mail off
>> quite
>> quickly before diving into a meeting, ironically, with my transport
>> experts. Please could you also cc me, as I am not on the obo-
>> discuss list.
>>
>> Thanks, and best wishes,
>>
>> Jennifer
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> ---------- Forwarded message ----------
>> Date: Sat, 12 Jan 2008 20:13:23 -0700
>> From: Mike Bada <mike.bada at UCHSC.edu>
>> Reply-To: obo-discuss at lists.sourceforge.net
>> To: obo-discuss at lists.sourceforge.net
>> Subject: [Obo-discuss] biological processes & molecular functions
>>
>> Hi all,
>>
>> Like Colin Batchelor, who has recently posted to this list, we are
>> trying to extensively use OBOs in the annotation of journal articles
>> (i.e., NLP annotation, not annotation in the sense of that done by
>> the
>> various model-organism databases). And, like Colin, we are
>> experiencing
>> issues in trying to use these OBOs (although I won't speak for him!).
>>
>> One issue that has caused us trouble is the dichotomy between GO
>> biological processes and molecular functions, and this leads to a
>> reexamination of the nature of these concepts. My understanding from
>> that written by Barry Smith and others is that a biological
>> process is
>> an actual process and that a molecular function is a functionality
>> inherent in a gene/gene product that is realized in a corresponding
>> process. (Please correct me if I'm wrong.) This makes sense, I
>> think,
>> when examining the GO transport and transporter terms: The
>> transporter
>> terms in the MF subontology are of the form "Enables the transport
>> of *
>> into, out of, within or between cells", while the transport terms
>> in the
>> BP subontology are of the form "The directed movement of * into,
>> out of,
>> within or between cells". So drug transporter activity is the
>> functionality inherent in a given gene/gene product that is
>> realized in
>> a drug-transport process. I assume that this is the reason for so
>> many
>> parallel BP transport and MF transporter activity terms. But if
>> this is
>> the case, then why isn't there a corresponding BP term for each MF
>> term?
>>
>> Arguing on the other side, most of the MF terms are not defined as
>> inherent functionalities but rather sound like processes to me: Most
>> binding terms are defined as "Interacting selectively with *"; most
>> catalytic-activity terms are defined as "Catalysis of the
>> reaction: *";
>> many enzyme-regulator-activity terms are defined as "Modulates *";
>> etc.
>> Furthermore, on the GO Web site, molecular functions are described as
>> "activities, such as catalytic or binding activities, that occur
>> at the
>> molecular level", which to me sound like processes (at a molecular
>> level) rather than functionalities inherent in genes/gene products
>> that
>> are realized in processes.
>>
>> The specific issues we're having and how we deal with them depend
>> on the
>> nature of the relationship between BPs and MFs, so I won't go into
>> more
>> detail until this is clear. We're hoping someone from GO/OBO could
>> clear this up for us.
>>
>> Cheers,
>> Mike
>>
>> ---------------------------------------------------------------------
>> ----
>> Check out the new SourceForge.net Marketplace.
>> It's the best place to buy or sell services for
>> just about anything Open Source.
>> http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/
>> marketplace
>> _______________________________________________
>> Obo-discuss mailing list
>> Obo-discuss at lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/obo-discuss
>>
>>
>>
>> ---------- Forwarded message ----------
>> Date: Sat, 12 Jan 2008 23:24:15 -0400
>> From: "Smith, Barry" <phismith at buffalo.edu>
>> Reply-To: obo-discuss at lists.sourceforge.net
>> To: obo-discuss at lists.sourceforge.net, obo-
>> discuss at lists.sourceforge.net
>> Subject: Re: [Obo-discuss] biological processes & molecular functions
>>
>> At 11:13 PM 1/12/2008, Mike Bada wrote:
>>
>>
>>> Hi all,
>>>
>>> Like Colin Batchelor, who has recently posted to this list, we are
>>> trying to extensively use OBOs in the annotation of journal articles
>>> (i.e., NLP annotation, not annotation in the sense of that done
>>> by the
>>> various model-organism databases). And, like Colin, we are
>>> experiencing
>>> issues in trying to use these OBOs (although I won't speak for
>>> him!).
>>>
>>> One issue that has caused us trouble is the dichotomy between GO
>>> biological processes and molecular functions, and this leads to a
>>> reexamination of the nature of these concepts. My understanding
>>> from
>>> that written by Barry Smith and others is that a biological
>>> process is
>>> an actual process and that a molecular function is a functionality
>>> inherent in a gene/gene product that is realized in a corresponding
>>> process. (Please correct me if I'm wrong.)
>>>
>>>
>>
>> You are not wrong.
>>
>>
>>
>>> This makes sense, I think,
>>> when examining the GO transport and transporter terms: The
>>> transporter
>>> terms in the MF subontology are of the form "Enables the
>>> transport of *
>>> into, out of, within or between cells", while the transport terms
>>> in the
>>> BP subontology are of the form "The directed movement of * into,
>>> out of,
>>> within or between cells". So drug transporter activity is the
>>> functionality inherent in a given gene/gene product that is
>>> realized in
>>> a drug-transport process. I assume that this is the reason for
>>> so many
>>> parallel BP transport and MF transporter activity terms. But if
>>> this is
>>> the case, then why isn't there a corresponding BP term for each
>>> MF term?
>>>
>>>
>>
>> I believe that the GO consortium has committed itself to creating
>> more consistency here. There are steps, also, to address the issue of
>> functions at other (non-molecular) levels of granularity -- for
>> example cellular and organ-level functions, reflecting also a more
>> careful treatment of the granularities on the process side.
>>
>>
>>
>>
>>> Arguing on the other side, most of the MF terms are not defined as
>>> inherent functionalities but rather sound like processes to me:
>>> Most
>>> binding terms are defined as "Interacting selectively with *"; most
>>> catalytic-activity terms are defined as "Catalysis of the
>>> reaction: *";
>>> many enzyme-regulator-activity terms are defined as "Modulates
>>> *"; etc.
>>> Furthermore, on the GO Web site, molecular functions are
>>> described as
>>> "activities, such as catalytic or binding activities, that occur
>>> at the
>>> molecular level", which to me sound like processes (at a molecular
>>> level) rather than functionalities inherent in genes/gene
>>> products that
>>> are realized in processes.
>>>
>>>
>>
>> This is a big issue -- resulting from an unfortunate decision to
>> label all function terms with the suffix 'activity'.
>> When you read 'activity' in GO, try to thing 'function'.
>>
>>
>>
>>> The specific issues we're having and how we deal with them depend
>>> on the
>>> nature of the relationship between BPs and MFs, so I won't go
>>> into more
>>> detail until this is clear. We're hoping someone from GO/OBO could
>>> clear this up for us.
>>>
>>>
>> Hope the above helps, at least somewhat.
>> BS
>>
>>
>>
>>> Cheers,
>>> Mike
>>>
>>> --------------------------------------------------------------------
>>> -----
>>> Check out the new SourceForge.net Marketplace.
>>> It's the best place to buy or sell services for
>>> just about anything Open Source.
>>> http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/
>>> marketplace
>>> _______________________________________________
>>> Obo-discuss mailing list
>>> Obo-discuss at lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/obo-discuss
>>>
>>>
>>
>>
>>
>>
>> ---------------------------------------------------------------------
>> ----
>> Check out the new SourceForge.net Marketplace.
>> It's the best place to buy or sell services for
>> just about anything Open Source.
>> http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/
>> marketplace
>> _______________________________________________
>> Obo-discuss mailing list
>> Obo-discuss at lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/obo-discuss
>>
>>
>>
>> ---------- Forwarded message ----------
>> Date: Sun, 13 Jan 2008 18:21:45 -0700
>> From: Mike Bada <mike.bada at UCHSC.edu>
>> Reply-To: obo-discuss at lists.sourceforge.net
>> To: obo-discuss at lists.sourceforge.net
>> Subject: Re: [Obo-discuss] biological processes & molecular functions
>>
>> Smith, Barry wrote:
>>
>>
>>
>>> At 11:13 PM 1/12/2008, Mike Bada wrote:
>>>
>>>
>>>
>>>
>>>> Hi all,
>>>>
>>>> Like Colin Batchelor, who has recently posted to this list, we are
>>>> trying to extensively use OBOs in the annotation of journal
>>>> articles
>>>> (i.e., NLP annotation, not annotation in the sense of that done
>>>> by the
>>>> various model-organism databases). And, like Colin, we are
>>>> experiencing
>>>> issues in trying to use these OBOs (although I won't speak for
>>>> him!).
>>>>
>>>> One issue that has caused us trouble is the dichotomy between GO
>>>> biological processes and molecular functions, and this leads to a
>>>> reexamination of the nature of these concepts. My understanding
>>>> from
>>>> that written by Barry Smith and others is that a biological
>>>> process is
>>>> an actual process and that a molecular function is a functionality
>>>> inherent in a gene/gene product that is realized in a corresponding
>>>> process. (Please correct me if I'm wrong.)
>>>>
>>>>
>>>>
>>>>
>>> You are not wrong.
>>>
>>>
>>>
>>>
>>>
>>>> This makes sense, I think,
>>>> when examining the GO transport and transporter terms: The
>>>> transporter
>>>> terms in the MF subontology are of the form "Enables the
>>>> transport of *
>>>> into, out of, within or between cells", while the transport
>>>> terms in the
>>>> BP subontology are of the form "The directed movement of * into,
>>>> out of,
>>>> within or between cells". So drug transporter activity is the
>>>> functionality inherent in a given gene/gene product that is
>>>> realized in
>>>> a drug-transport process. I assume that this is the reason for
>>>> so many
>>>> parallel BP transport and MF transporter activity terms. But if
>>>> this is
>>>> the case, then why isn't there a corresponding BP term for each
>>>> MF term?
>>>>
>>>>
>>>>
>>>>
>>> I believe that the GO consortium has committed itself to creating
>>> more consistency here. There are steps, also, to address the
>>> issue of
>>> functions at other (non-molecular) levels of granularity -- for
>>> example cellular and organ-level functions, reflecting also a more
>>> careful treatment of the granularities on the process side.
>>>
>>>
>>>
>>>
>>
>> OK, it is as I thought. Does this mean that there will be many more
>> parallel BP and MF terms, even a one-to-one correspondence?
>>
>> The crux of our difficulties is that, when analyzing and
>> attempting to
>> annotate a typical biological assertion found in a journal article,
>> we're often finding it extremely difficult to differentiate whether
>> something refers to a BP or a MF. A simple example is "A
>> transports B
>> to the nucleus." Does this refer to the process "B transport" or
>> to the
>> corresponding inherent functionality "B transporter activity",
>> i.e., to
>> the actual process or to the ability to effect the process? (I'm
>> using transport/transporter activity examples throughout, but
>> there are
>> many others.) The use of the simple verb might suggest that it's a
>> process, but of course, this is a generic statement about the
>> ability of
>> A to transport B to the nucleus, not about an actual reporting of the
>> transport of a specific B to the nucleus by a specific A, so it
>> can be
>> seen as a MF as well. Frankly, I'd say that this is an ontological
>> distinction about which the large majority of biologists
>> understandably
>> wouldn't care, but it is something we have to consider because we're
>> interested in getting computational systems to be able to
>> automatically
>> recognize GO (and other OBO) terms in natural-language text, and the
>> distinctions between the two in text are often ambiguous to say the
>> least, which makes the task that much more difficult.
>>
>> This leads me to ask, perhaps radically: Do we need both BPs and
>> MFs?
>> I'm not denying an ontological distinction between BPs and MFs as
>> they're defined now; I'm just wondering if it might be a distinction
>> that is too subtle to be genuinely useful and too ambiguous to
>> differentiate in, for example, natural language. Even applied to GO
>> annotations, if a gene or gene product has an annotation that
>> indicates
>> that it has the functionality to effect a drug-transport process,
>> is the
>> addition of another annotation that indicates that it participates in
>> drug-transport processes useful? Wouldn't it be sufficient to
>> have only
>> processes or only functionalities in the GO? (I think processes are
>> more straightforward and thus more generally useful.) This would
>> eliminate a huge amount of ambiguity in the application of terms
>> among
>> these two subontologies, I think.
>>
>> Trepidatiously expecting a firestorm,
>> Mike
>>
>>
>>
>>>
>>>
>>>
>>>> Arguing on the other side, most of the MF terms are not defined as
>>>> inherent functionalities but rather sound like processes to me:
>>>> Most
>>>> binding terms are defined as "Interacting selectively with *"; most
>>>> catalytic-activity terms are defined as "Catalysis of the
>>>> reaction: *";
>>>> many enzyme-regulator-activity terms are defined as "Modulates
>>>> *"; etc.
>>>> Furthermore, on the GO Web site, molecular functions are
>>>> described as
>>>> "activities, such as catalytic or binding activities, that occur
>>>> at the
>>>> molecular level", which to me sound like processes (at a molecular
>>>> level) rather than functionalities inherent in genes/gene
>>>> products that
>>>> are realized in processes.
>>>>
>>>>
>>>>
>>>>
>>> This is a big issue -- resulting from an unfortunate decision to
>>> label all function terms with the suffix 'activity'.
>>> When you read 'activity' in GO, try to thing 'function'.
>>>
>>>
>>>
>>>
>>>
>>>> The specific issues we're having and how we deal with them
>>>> depend on the
>>>> nature of the relationship between BPs and MFs, so I won't go
>>>> into more
>>>> detail until this is clear. We're hoping someone from GO/OBO could
>>>> clear this up for us.
>>>>
>>>>
>>>>
>>>>
>>> Hope the above helps, at least somewhat.
>>> BS
>>>
>>>
>>>
>>>
>>>
>>>> Cheers,
>>>> Mike
>>>>
>>>> -------------------------------------------------------------------
>>>> ------
>>>> Check out the new SourceForge.net Marketplace.
>>>> It's the best place to buy or sell services for
>>>> just about anything Open Source.
>>>> http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/
>>>> marketplace
>>>> _______________________________________________
>>>> Obo-discuss mailing list
>>>> Obo-discuss at lists.sourceforge.net
>>>> https://lists.sourceforge.net/lists/listinfo/obo-discuss
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>> --------------------------------------------------------------------
>>> -----
>>> Check out the new SourceForge.net Marketplace.
>>> It's the best place to buy or sell services for
>>> just about anything Open Source.
>>> http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/
>>> marketplace
>>> _______________________________________________
>>> Obo-discuss mailing list
>>> Obo-discuss at lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/obo-discuss
>>>
>>>
>>>
>>>
>>>
>>
>>
>> ---------------------------------------------------------------------
>> ----
>> Check out the new SourceForge.net Marketplace.
>> It's the best place to buy or sell services for
>> just about anything Open Source.
>> http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/
>> marketplace
>> _______________________________________________
>> Obo-discuss mailing list
>> Obo-discuss at lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/obo-discuss
>>
>>
>>
>>
>
>
> ----------------------------------------------------------------------
> ---
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/
> marketplace
> _______________________________________________
> Obo-discuss mailing list
> Obo-discuss at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/obo-discuss
>
More information about the Go
mailing list