Title | Content | Created By | Date Created | ID | Topic ID |
---|
Progress to date | This document is a very brief history of the project and lists at least a few of the tasks that I believe remain. | jwilcke | Thu Jul 07 12:12:04 Z 2011 | post4661 | topc3242 |
Re: Progress to date | Jeff, this seems like a good summary. I agree that this project group needs to be re-energized. I wasn't involved significantly in any of the prior work, but I have been and I am very interested in seeing this move forward and I would plan to participate, particularly in regard to microorganisms and infectious disease.
Rob | rhausam | Thu Jul 14 06:32:36 Z 2011 | post4718 | topc3242 |
Re: Progress to date | Hi all,
As I've discussed with Jeff, I'd like to propose that rather than removing the taxon reference (e.g.Genus, Family, etc) from organism FSNs, we instead use the text definition to specify that the meaning is "the population of all organisms comprising that taxon" (e.g. for Genus Brucella, "The population of bacteria comprising Genus Brucella." While reconciliation with OBO and other former ontologies may require the distinction between the singular intellectual construct of a taxon vs. the plural reality of organisms referred to in medical records, I think that distiction is not really important (and perhaps may appear esoteric) to users, as this is not a distinction they likely need/want (?) to make.
Inclusion of the taxon level term in FSNs was one of the changes various users had requested, and the pattern agreed on in the previous Organism Summit. While I believe absolutely in the necessity of clear, unambiguous FSNs, in this case there doesn't appear to be a compelling need to change the FSNs . Using the text definition could clarify the intended meaning for ontological purists while avoiding the appearance of FSN "churn."
In some respects, "IS A" represents a bit of a misnomer when used in the organism heirarchy, as the true nature of the relationships within the heirarchy is perhaps better expressed as "belongs to" or "member of" than "IS A [kind of]." This distinction may not warrant the creation and use of a different attribute to construct parent-child relationships within this single heirarchy--we have, in the past, tacitly agreed to ignore the difference, and to treat "IS A" relationships in the organism heirarchy as if they meant "Member of"--but it may be worth pointing out that the distinction does logically exist, and could potentially be represented by a different attribute. | plivesay | Fri Jul 15 19:18:19 Z 2011 | post4781 | topc3242 |
Re: Progress to date | I agree with the spirit of this, especially the admonition to avoid changing the FSN but instead use a text definition to clarify what it means.
I would propose "organism bearing <taxon designation>" as the definition, instead of "population etc".
Thus "Genus Escherichia (organism)" - organism is singular - means "An organism bearing the Genus designation (formally: the taxon quality) of Escherichia".
This does two things: (1) it avoids having to apologize for reinterpreting "IS A", since we aren't, and (2) it avoids apologizing for using a plural meaning when the FSN is singular. Double win. | kspackman | Fri Jul 15 22:14:30 Z 2011 | post4782 | topc3242 |
Re: Progress to date | I like "win-win" What step(s) do we take to get approval and add this to the workplan?
There are a few other "loose ends" that need attention before we begin any remodeling project. For example, we still need a means for coping with life-stages (e.g., larval, adult, worm, microfilarial, etc.). I expect we'll start separate threads for these. | jwilcke | Wed Jul 20 13:14:04 Z 2011 | post4801 | topc3242 |
Re: Progress to date | I have recently reviewed this model as well as the paper by Stefan Shulz and have a couple of questions and a possible revision to the proposed model.
My first question is whether there is actually a use case for the representation of non-taxonomic organisms (e.g. gram positive rod) in the organism hierarchy? If it is our desire to "clean up" the organism hierarchy to represent (primarily) Linnaean taxonomy, then I would submit that these are in almost every case findings about an organism. The nature of the finding is a characteristic of a subset of taxa that possess this particular property.
My second question is whether it is really necessary to represent the full taxonomic classification for an organism that possesses a Linnaean name. For example, is it really necessary to say:
Escherichia coli (organism) - bearer of - Species Escherichia coli (taxon quality) as opposed to:
Escherichia coli (organism) - bearer of - Species (taxon quality)
The taxonomic level of the organism appears to me to be the only needed quality. This would eliminate the need to reproduce the entire Linnaean structure in the quality hierarchy. Is there a use case where this is not true?
So my revision to the proposed model would involve the movement of non-taxonomic organisms to the findings hierarchy and using one of the existing defining attributes such as ASSOCIATED WITH or INTERPRETS (since this is usually a finding based on a procedure) to associate the proper organism taxon. Since I cannot conceive of any way to determine these non-taxonomic types of organisms except as finding values to some observable, I think it would clean up the organism hierarchy and move it towards more of a strict Linnaean hierarchy.
If I am missing something from the DL or computational aspect, please educate me. | jcase | Wed Aug 31 15:17:06 Z 2011 | post5185 | topc3242 |
Re: Progress to date | The arguements in favor of retaining organisms such as gram negative bacteria are essentially that: 1) That disorders such as "gram negative infection" and its subtypes require an etiology (etiologic agent now I think) for autoclassification to function correctly. SO if you post-coordinate a concept such as "Pneumonia caused by (say) Vibrio" it would autoclassify IF Vibrio can be identified as a gram negative bacteria and Gram negative pneumonia has a definition that includes an etiologic agent of gram negative bacteria, the Pneumonia caused by Vibrio concept autoclassifies as a gram negative pneumonia and as a gram negative disease. 2) (..and this part of my arguement may be rather pedantic) That a laboratory finding of "gram negative" is not laboratory finding about a patient, it is a laboratory finding about an organism. (The patient is not the bearer of gram-negativeness, the organism is...) I think that placing "gram negative organism" in the findings hierarchy is somewhat unsettling if not just plain wrong. It is not a "finding about a patient" as is implied by most of the rest of the content.
AS for the Non-linnaean subtypes of organisms, I would argue that "Salmonella serotypeX" is the name of an organism. Again, it is NOT a laboratory finding. IN fact, in the case of serotypes the laboratory test does not determine the serotype, the laboratory test determines the presence of antigens born by the organism. The list of antigens "found" then is interpreted as "organism named X." From my perspective, the bondary between species and serotype is artificial at best. I don't perceive any logical difference between interpreting antigen expression through Kauffman-White (for example) and interpreting aerophilicity, sugar fermentation et all through Bergey.
I don't think the purpose of the living organism hierarchy re-organization has ever been to banish non-linnaean concepts from LO. I think the purpose of the re-organization is produce an accurate Linnaean hierarchy and, where necessary, represent non-linnaean content logically and accurately. If we ban non-linnaean content from LO, we will need to find a considerable number of non-linnaean concepts a home. That's a political problem (and financing problem) that doesn't have an easy answer and I don't think its necessary.
I'm going to defer any reply to your taxon level comment to a later time. I'm laboring under the impression what we've proposed is consistent with Stephan Schulz work on the subject (I may need correcting in that regard) and believe that most of those assignments can be automated (so the workload is actually nominal). I believe we'd have to rethink the appropropriate attribute to attach "Species" to E. coli. I'm not absolutely certain that either approach is required for a functional LO hierarchy. | jwilcke | Wed Aug 31 16:36:44 Z 2011 | post5186 | topc3242 |
Re: Progress to date | Jeff,
Since I wrote my reply I had a long discussion with folks here at NLM and I have had an aha moment and understand much more clearly the implicaitons of the reorg as you proposed. Although I now understand and appreciate the benefits of the non-Linnaean organism in the hierarchy, I am still not so clear on the Taxon quality aspect.
As for the serotype, antigenic formulae, etc. I think from what I read that I agree with you. These subtypes are really just an extension of the Linnaean hierarchy without a specific name. Oneof the things I was thinking and would really like your thoughts is to create high level concpts that represent the major characteristics of organisms allowing the specific Linneaen type to classify under them, much as was done with procedures. For example, "Bacteria categorized by staining character", which would then have children such as Gram positive bacteria; Acid fast bacteria, etc.
By fully defining these characteristics, we would have a nice set of templates to assign to taxa. The difficult part would be to determine hwere to stop in defining these differnet characteristics. | jcase | Thu Sep 01 20:16:41 Z 2011 | post5188 | topc3242 |
Re: Progress to date | I've spent most of today reviewing the documentation for the project group and revising my microbiology textbook "Manual of Clinical Microbiology" looking for defining characteristics. I was wondering how you define a characteristic when it is strain dependant. Looking at characteristic reactions in Enterobacteriacea , lots of species show variable (10-89%) reactions in biochemical tests. Obviously modern software takes this into account when calculating which colonies belong to which species. But will we ever get close to a fully defined species? What if the characteristics of a species vary with geography...Snomed CT is meant to be a global terminology. What if characteristic antimicrobial sensitivity patterns are part of the defining criteria and these change over time under selective pressure of antimicrobial usage?
I'm thinking that we may not be able to add too many defining attributes - but any that we CAN add will be useful for aggregation and CDS.
Regards,
Michael Osborne | mosborne | Mon Oct 17 03:59:44 Z 2011 | post5541 | topc3242 |
Re: Progress to date | Sorry your message took so long to post. I was unaware of its existence. Clearly my understanding of Collabnet's behavior needs work.
Could you possibly find time to provide a detailed example (a test case if you will) that demonstrates the variability of which you speak. Up to this point, we've been thinking that strains are subtypes of species (or whatever class level) and have their own defining characteristics that separate them from other subtypes of that species. Fully defined organisms? I'm hoping for adequately defined. | jwilcke | Tue Mar 27 10:59:59 Z 2012 | post6417 | topc3242 |
Re: Progress to date | I think that "Bacteria classified by morphology" and related classes are navigational concepts. While they might enhance they are not necessary in order to make the organisms hierarchy FUNCTIONAL.
" | jwilcke | Tue Mar 27 11:05:55 Z 2012 | post6418 | topc3242 |
Re: Progress to date | Hi Jeff,
See my comments and your reply below (posted here as requested)
Michael:
At the species level there are a large number of tests with variable positive reactions. The attached pdf is from my old uni copy of "Manual of Clinical Microbiology - Fourth edition" which will illustrate my point. The chemical properties for defining organisms can only be _defining_ if it true for all instances of the classes. Many of the identification techniques use a fuzzy logic to calculate the probability of an isolate being in a certain species, because of the problem of chemical variablilty in isolates. (I'm sure this is seriously old news to you). The other issue is that some chemical characteristics are defining when Negative, and I'm not sure how you would include these in Snomed CT, the description logic does not cater for negation. Negation also looms when we talk about the family Enterobacteriaceae - golden rule is that all bacteria in this genus are oxidase negative.
For my example I'll choose Citrobacter freundii (column 5 of table 2) of enterobacteriacea.pdf:
There are 5 chemical characteristics that may be defining (100% positive), 23 that are variable and 5 that are always negative.
So discarding the 23 variable ones and 5 that can't be used because of the negation problem, that leaves 5 defining characteristics.
(Although very closely linked to the lab, I've been working in IT/Informatics for the last 10 years - before that I worked as a Medical scientist, so much of this might have changed or be irrelevant, especially when you take into account new technology like Maldi-tof and 16S RNA.)
Jeff's response:
I think we can overcome the problem of negation. You are certainly correct that SNOMED does not allow negation but it does allow assertion of a negative findings. Example from microbiology would be “Gram negative.” In truth, this is a finding that a bacteria does NOT retain a particular stain in its cell wall when washed with alcohol, but the assertion is that the organism is gram negative. By extension, I can also assert that the bacteria is “oxidase negative.” I assume this causes alignment issues with decision systems that anticipate negation and I assume that the day will come when SNOMED supports negation.
As for variably present, this is certainly not something I know how to do at this time (in SNOMED at least). I think perhaps, there are means by which certain sophisticated ontologies can manage probability.
On a practical note, I suspect that there are several person-years’ worth of work reviewing, correcting and defining the “upper hierarchy” before we reach the level of differentia among the species of enterobacteriaceae so there is time to work on these issues.
-Jeff | mosborne | Wed Apr 18 00:10:31 Z 2012 | post6533 | topc3242 |
Re: Progress to date | I can't resist a bit of nostalgia here - (and it truly isn't what it used to be, by the way) - and apologize in advance for inflicting this on you, but ... see pages 61-65, and appendix A pp 97-104, of the attached PhD dissertation by some obscure student from UI Urbana, circa 1988.
Bottom line: rules learned from the API 20 pattern dictionary could identify the species of Enterobacteriaceae with about 94% accuracy using just 10 tests (the API10). If you go to all 20 tests, you get a nominally correct identification.
But I wouldn't ever consider putting these (or similar) rules into SNOMED as defining characteristics. They are rules for operationally identifying the species, NOT rules that define what the species inherently is, ontologically. There's a big difference. | kspackman | Wed Apr 18 00:33:42 Z 2012 | post6534 | topc3242 |