Laboratory test modeling using spreadsheet template
Bill Hogan
At the last call, I volunteered to try out the spreadsheet for filling
out template information to create representational units for
laboratory tests.
I uploaded my effort thus far to the IAO site:
A summary of the issues I encountered:
1. Derived units of measure: although the spreadsheet says code UO:
0000300 is mmol/L, I cannot find this code in any extant version of UO
available for download on the WWW. Similarly, I cannot find a code
for mg/dL. I realize there are more fundamental issues here, i.e.,
whether we want to create a code for every derived unit. My gestalt
though is that for clinical lab testing, this set is relatively finite
and would not lead to overbearing combinatorial explosion.
2. Blood urea nitrogen. First, despite its name, it is performed on
serum (historically, it used to be done on blood). But that's not the
problem. The issue is that the test measures the mass per unit volume
(of serum) of all nitrogen atoms that are part of some urea molecule.
Thus, this test is not as straightforward as the others, and does not
fit the template. The analyte is really: nitrogen and (some part-of
urea).
3. BUN/Creatinine ratio. How do we handle the ratio of two measured
values? Also, I don’t have a use case at this point, but it would not
surprise me if some tests measure the ratio directly, and thus you
have a measurement of a ratio as opposed to the ratio of two
measurements. For this test, however, we just need the ratio of two
measurements. Also, this test has no units of measure as they “cancel
out” (mg/dL / mg/dL = 1). Do we need to do anything special with
“unitless” measurements? Also, I assume the evaluant still makes
sense as serum, as we could report this ratio for, say, urine.
4. I wasted hours investigating the difference between total calcium
and ionized calcium. It seemed to me that any calcium atom would be
involved in an ionic (or mostly ionic, let’s not discuss the fact that
no bond is 100% ionic) bond, and thus the analyte was Ca(2+). Of
course, from my clinical training, I knew the relevance of the
difference for patient care, but not the difference at an atomic and
molecular level. Nevertheless, suffice it to say that ~50% of Ca
atoms are bound up so tightly in serum that they have no relevant
physiological effect in the metabolism of Ca. So then, is the
difference b/w total and ionized Ca testing the analyte? Or is it the
“space” or “compartment” in which the analyte is contained? At this
point I favor the latter, since I don’t think any Calcium atom in the
body has all its electrons entirely to itself. The ionized Ca test
seems to be measuring the concentration of Ca(2+) ions in solution,
where the albumin-binding of Ca removes a substantial portion of Ca
atoms out of solution. The issue then would be how to model the
“water-space-with-no-albumin” of blood, because that’s definitely not
the specimen, the specimen is serum, and then the methodology of the
test somehow isolates and thus measures just the portion of Ca not
bound up and thus free in solution.
Interestingly, there is a few mmol/L difference between total sodium
and ionized sodium. There’s lots of interesting literature on
converting Na ion-activity measurement (the most common method today
of measuring serum sodium) to a molal concentration (the most common
method of reporting sodium concentration today). In other words,
laboratories today measure ion activity but report molal
concentration, because the latter is what doctors are used to.
5. aspartate aminotransferase (AST). PRO has aspartate
aminotransferase, cytoplasmic (PRO:000008153 ) vs. aspartate
aminotransferase, mitochondrial (PRO:000008154) but no plain aspartate
aminotransferase (the parent of the two isoenzymes is just Protein
(PRO:000000001). It is canonical to have molecules of AST in your
blood, so you can’t say that canonically every instance of these two
isoenzymes is located in some cytoplasm or mitochondrion,
respectively.
We need to request a new term from PRO that is the parent of the two
existing terms. This link: http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=cm&part=A3139
says: “Two distinct forms [of AST] have been identified: a
cytoplasmic, or soluble isoenzyme, and a mitochondrial isoform.
Selective measurement of these isoenzymes has no currently
demonstrated clinical application.”
6. alanine aminotransferase (ALT). Same problem as for AST, except
the isoenzymes are ALT1 (PRO:000008208) and ALT2 (PRO:000008209).
7. bilirubin. The total bilirubin test measures the mass of all
bilirubin molecules that are either free (indirect), part of some
(glucosyluronic acid)bilirubin molecule (direct), or part of some
protein-bilirubin conjugated molecule (delta bilirubin). In ChEBI,
bilirubin and (glucosyluronic acid)bilirubin do not share a parent,
appropriately, since one is part-of the other (and not is-a the
other). Nevertheless the measurement is the mass of just the
bilirubin components per unit volume of serum, not the mass of the
glucuronide or protein.
Having said all that, I’m comfortable with just bilirubin as the
analyte.
8. globulin and albumin/globulin ratio. Globulin is simply
calculated as total protein – albumin. Then the A/G ratio is albumin/
globulin, i.e., albumin/(total protein – albumin). How do we
represent the fact that some observations are calculations derived
from measurements?
Furthermore, there’s no PRO term for globulin.
9. lactate dehydrogenase (or LDH). No PRO term for it, or any of its
isoenzymes.
Bill
Melanie Courtot
Thanks for trying that out and sending back your results, and
apologies for not coming back to you earlier.
Here is a try at answering some of your questions, maybe it will
inspire others :)
I think this specific topic is overlapping IAO and OBI: the template
proposal (QTT) has been developed by Philippe in the context of OBI,
and your assays would belong in OBI, but some terms (e.g.,
measurements) would have their place into IAO.
On 21-Jan-10, at 7:14 AM, Bill Hogan wrote:
> Folks,
>
> At the last call, I volunteered to try out the spreadsheet for filling
> out template information to create representational units for
> laboratory tests.
>
> I uploaded my effort thus far to the IAO site:
>
> http://information-ontology.googlegroups.com/web/OBI-QTT-Analyte-Assay.xls?gda=0kEZD0sAAAAImI6OMgNJbexEGlJ2K2Cug78pZLm7_OpWii9ot153V94kVQLiOMRXMFnDK3A3iTL2dGxO9rdPgy5uB-Yu-3XKBkXa90K8pT5MNmkW1w_4BQ&gsc=7zzxyQsAAABRDyxnS1TcjiWtmfxke330
>
> A summary of the issues I encountered:
>
> 1. Derived units of measure: although the spreadsheet says code UO:
> 0000300 is mmol/L, I cannot find this code in any extant version of UO
> available for download on the WWW. Similarly, I cannot find a code
> for mg/dL. I realize there are more fundamental issues here, i.e.,
> whether we want to create a code for every derived unit. My gestalt
> though is that for clinical lab testing, this set is relatively finite
> and would not lead to overbearing combinatorial explosion.
I don't know about UO_0000300 specifically, but I would think as well
that there shouldn't be any issue adding units like mg/dL etc.
I had a look at the UO file (http://obo.cvs.sourceforge.net/*checkout*/obo/obo/ontology/phenotype/unit.obo
), and there seems to be some very specific units already in, for
example:
[Term]
id: UO:0000160
name: microeinstein per square meter per second
def: "An irradiance unit which is equal to one microeinstein per
square meter per second or 10^[-6] microeinstein/
sm^[2]." [Wikipedia:Wikipedia "http://www.wikipedia.org/"]
synonym: "micromole per second and square meter mmol/sm^2" EXACT []
synonym: "umicroeinstein/sm^[2]" EXACT []
is_a: UO:0000154 ! irradiance unit
If the number of units to be created were to explode it may be worth
thinking about a way to build those on the fly if it were possible,
but I don't think this should concern us for now.
>
> 2. Blood urea nitrogen. First, despite its name, it is performed on
> serum (historically, it used to be done on blood). But that's not the
> problem. The issue is that the test measures the mass per unit volume
> (of serum) of all nitrogen atoms that are part of some urea molecule.
> Thus, this test is not as straightforward as the others, and does not
> fit the template. The analyte is really: nitrogen and (some part-of
> urea).
I would say in that case that even though you are pretty close, you
don't completely adhere to this specific pattern. I would prefer to
have a distinct term submission in that case, rather than try to
accommodate and end up with a wrong restriction. I suspect it should
be possible to create the class nitrogen and (some part-of urea) and
then use that, but in my opinion the template was done to ease entry
of a batch of data, and we should take only what directly complies,
and treat the rest separately, or we risk to spend more time trying to
make things fit than time it would have taken to process manually, and
we increase risk for error with each modification.
>
> 3. BUN/Creatinine ratio. How do we handle the ratio of two measured
> values? Also, I don’t have a use case at this point, but it would not
> surprise me if some tests measure the ratio directly, and thus you
> have a measurement of a ratio as opposed to the ratio of two
> measurements. For this test, however, we just need the ratio of two
> measurements. Also, this test has no units of measure as they “cancel
> out” (mg/dL / mg/dL = 1). Do we need to do anything special with
> “unitless” measurements? Also, I assume the evaluant still makes
> sense as serum, as we could report this ratio for, say, urine.
I would go for same than above, if it doesn't fit straight away let's
not try and force it. If you have several of those type of assays (the
one above sounded more like a very specific case), we may create a new
pattern. The current pattern was really analyte in evaluant gets
measured and outputs scalar measurement datum.
Based on http://en.wikipedia.org/wiki/BUN-to-creatinine_ratio: In
medicine, the BUN-to-creatinine ratio, also BUN-creatinine ratio, BUN/
creatinine ratio or BUN:Cr, is the ratio of two serum laboratory
values, the blood urea nitrogen (BUN) and serum creatinine.
So you first measure BUN, then Creatinine, and then compute a ratio.
Could you have BUN assay, serum creatinine assay using the pattern,
and then the ratio would be the output of a data transformation?
Regarding units, we have measurement datum with subclasses like
"ration of collected to emitted light" (in OBI). Scalar measurement
datum is the specific subclass which has measurement value and
measurement unit.
I am afraid I don't have your clinical training, so I am somehow
struggling here :)
Based on http://en.wikipedia.org/wiki/Calcium_metabolism
The serum level of calcium is closely regulated with a normal total
calcium of 2.2-2.6 mmol/L (9-10.5 mg/dL) and a normal ionized calcium
of 1.1-1.4 mmol/L (4.5-5.6 mg/dL). The amount of total calcium varies
with the level of serum albumin, a protein to which calcium is bound.
The biologic effect of calcium is determined by the amount of ionized
calcium, rather than the total calcium. Ionized calcium does not vary
with the albumin level, and therefore it is useful to measure the
ionized calcium level when the serum albumin is not within normal
ranges, or when a calcium disorder is suspected despite a normal total
calcium level.
So we have total calcium, part of it is inactive because bound to
albumin, and you are interested in the remaining part, i.e. the
ionized calcium?
>
> 5. aspartate aminotransferase (AST). PRO has aspartate
> aminotransferase, cytoplasmic (PRO:000008153 ) vs. aspartate
> aminotransferase, mitochondrial (PRO:000008154) but no plain aspartate
> aminotransferase (the parent of the two isoenzymes is just Protein
> (PRO:000000001). It is canonical to have molecules of AST in your
> blood, so you can’t say that canonically every instance of these two
> isoenzymes is located in some cytoplasm or mitochondrion,
> respectively.
>
> We need to request a new term from PRO that is the parent of the two
> existing terms. This link: http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=cm&part=A3139
> says: “Two distinct forms [of AST] have been identified: a
> cytoplasmic, or soluble isoenzyme, and a mitochondrial isoform.
> Selective measurement of these isoenzymes has no currently
> demonstrated clinical application.”
>
> 6. alanine aminotransferase (ALT). Same problem as for AST, except
> the isoenzymes are ALT1 (PRO:000008208) and ALT2 (PRO:000008209).
I don't see anything controversial here. The PRO tracker is at http://sourceforge.net/tracker/?group_id=266825&atid=1135711
, do you want to go ahead?
>
> 7. bilirubin. The total bilirubin test measures the mass of all
> bilirubin molecules that are either free (indirect), part of some
> (glucosyluronic acid)bilirubin molecule (direct), or part of some
> protein-bilirubin conjugated molecule (delta bilirubin). In ChEBI,
> bilirubin and (glucosyluronic acid)bilirubin do not share a parent,
> appropriately, since one is part-of the other (and not is-a the
> other). Nevertheless the measurement is the mass of just the
> bilirubin components per unit volume of serum, not the mass of the
> glucuronide or protein.
>
> Having said all that, I’m comfortable with just bilirubin as the
> analyte.
>
> 8. globulin and albumin/globulin ratio. Globulin is simply
> calculated as total protein – albumin. Then the A/G ratio is albumin/
> globulin, i.e., albumin/(total protein – albumin). How do we
> represent the fact that some observations are calculations derived
> from measurements?
I would go for output of some data transformation which takes the
measurements as input.
>
> Furthermore, there’s no PRO term for globulin.
>
> 9. lactate dehydrogenase (or LDH). No PRO term for it, or any of its
> isoenzymes.
I'm happy to help with the PRO submissions if needed, let me know.
Hope some of those help, if there is interest we could maybe either
use an IAO call or have a distinct one with only the people interested
in QTT/lab tests.
Cheers,
Melanie
>
> Bill
> --
> informatio...@googlegroups.com
> To change settings, visit
> http://groups.google.com/group/information-ontology
---
Mélanie Courtot
TFL- BCCRC
675 West 10th Avenue
Vancouver, BC
V5Z 1L3, Canada
Bill Hogan
Thank you! This response is helpful. My specific responses inline below.
Hi Bill,
Thanks for trying that out and sending back your results, and apologies for not coming back to you earlier.
Here is a try at answering some of your questions, maybe it will inspire others :)
I think this specific topic is overlapping IAO and OBI: the template proposal (QTT) has been developed by Philippe in the context of OBI, and your assays would belong in OBI, but some terms (e.g., measurements) would have their place into IAO.
I don't know about UO_0000300 specifically, but I would think as well that there shouldn't be any issue adding units like mg/dL etc.
On 21-Jan-10, at 7:14 AM, Bill Hogan wrote:
Folks,
At the last call, I volunteered to try out the spreadsheet for filling
out template information to create representational units for
laboratory tests.
I uploaded my effort thus far to the IAO site:
http://information-ontology.googlegroups.com/web/OBI-QTT-Analyte-Assay.xls?gda=0kEZD0sAAAAImI6OMgNJbexEGlJ2K2Cug78pZLm7_OpWii9ot153V94kVQLiOMRXMFnDK3A3iTL2dGxO9rdPgy5uB-Yu-3XKBkXa90K8pT5MNmkW1w_4BQ&gsc=7zzxyQsAAABRDyxnS1TcjiWtmfxke330
A summary of the issues I encountered:
1. Derived units of measure: although the spreadsheet says code UO:
0000300 is mmol/L, I cannot find this code in any extant version of UO
available for download on the WWW. Similarly, I cannot find a code
for mg/dL. I realize there are more fundamental issues here, i.e.,
whether we want to create a code for every derived unit. My gestalt
though is that for clinical lab testing, this set is relatively finite
and would not lead to overbearing combinatorial explosion.
Great!
I had a look at the UO file (http://obo.cvs.sourceforge.net/*checkout*/obo/obo/ontology/phenotype/unit.obo), and there seems to be some very specific units already in, for example:
[Term]
id: UO:0000160
name: microeinstein per square meter per second
def: "An irradiance unit which is equal to one microeinstein per square meter per second or 10^[-6] microeinstein/sm^[2]." [Wikipedia:Wikipedia "http://www.wikipedia.org/"]
synonym: "micromole per second and square meter mmol/sm^2" EXACT []
synonym: "umicroeinstein/sm^[2]" EXACT []
is_a: UO:0000154 ! irradiance unit
If the number of units to be created were to explode it may be worth thinking about a way to build those on the fly if it were possible, but I don't think this should concern us for now.
I agree with you. The number of units needed for clinical lab measurements won't "break the bank".
I would say in that case that even though you are pretty close, you don't completely adhere to this specific pattern. I would prefer to have a distinct term submission in that case, rather than try to accommodate and end up with a wrong restriction. I suspect it should be possible to create the class nitrogen and (some part-of urea) and then use that, but in my opinion the template was done to ease entry of a batch of data, and we should take only what directly complies, and treat the rest separately, or we risk to spend more time trying to make things fit than time it would have taken to process manually, and we increase risk for error with each modification.
2. Blood urea nitrogen. First, despite its name, it is performed on
serum (historically, it used to be done on blood). But that's not the
problem. The issue is that the test measures the mass per unit volume
(of serum) of all nitrogen atoms that are part of some urea molecule.
Thus, this test is not as straightforward as the others, and does not
fit the template. The analyte is really: nitrogen and (some part-of
urea).
Yes, the options are to modify the template, to model BUN separately outside the template, or to create a defined class of nitrogen atoms that are part of some urea molecule and then it fits the template again. I agree that this test, although very very commonly performed, nevertheless is somewhat uncommon among tests in terms of "breaking the template". Thus, BUN would appropriately be handled outside the template. I don't know if I like the 'defined class' approach.
I would go for same than above, if it doesn't fit straight away let's not try and force it. If you have several of those type of assays (the one above sounded more like a very specific case), we may create a new pattern. The current pattern was really analyte in evaluant gets measured and outputs scalar measurement datum.
3. BUN/Creatinine ratio. How do we handle the ratio of two measured
values? Also, I don’t have a use case at this point, but it would not
surprise me if some tests measure the ratio directly, and thus you
have a measurement of a ratio as opposed to the ratio of two
measurements. For this test, however, we just need the ratio of two
measurements. Also, this test has no units of measure as they “cancel
out” (mg/dL / mg/dL = 1). Do we need to do anything special with
“unitless” measurements? Also, I assume the evaluant still makes
sense as serum, as we could report this ratio for, say, urine.
Ratios are extremely common and I recommend the creation of a template to handle them.
Based on http://en.wikipedia.org/wiki/BUN-to-creatinine_ratio: In medicine, the BUN-to-creatinine ratio, also BUN-creatinine ratio, BUN/creatinine ratio or BUN:Cr, is the ratio of two serum laboratory values, the blood urea nitrogen (BUN) and serum creatinine.
So you first measure BUN, then Creatinine, and then compute a ratio. Could you have BUN assay, serum creatinine assay using the pattern, and then the ratio would be the output of a data transformation?
Regarding units, we have measurement datum with subclasses like "ration of collected to emitted light" (in OBI). Scalar measurement datum is the specific subclass which has measurement value and measurement unit.
I don't like at all the approach of creating units that are specific to each type of ratio. Where the units of measure cancel, I think we either leave it empty or have some unit of measure called 'unity' or something (which I am not sure makes sense even).
The question is how to differentiate the two tests? Is it by analyte or evaluant? I choose the latter. The evaluant is albumin-free serum.
5. aspartate aminotransferase (AST). PRO has aspartate
aminotransferase, cytoplasmic (PRO:000008153 ) vs. aspartate
aminotransferase, mitochondrial (PRO:000008154) but no plain aspartate
aminotransferase (the parent of the two isoenzymes is just Protein
(PRO:000000001). It is canonical to have molecules of AST in your
blood, so you can’t say that canonically every instance of these two
isoenzymes is located in some cytoplasm or mitochondrion,
respectively.
We need to request a new term from PRO that is the parent of the two
existing terms. This link: http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=cm&part=A3139
says: “Two distinct forms [of AST] have been identified: a
cytoplasmic, or soluble isoenzyme, and a mitochondrial isoform.
Selective measurement of these isoenzymes has no currently
demonstrated clinical application.”
6. alanine aminotransferase (ALT). Same problem as for AST, except
the isoenzymes are ALT1 (PRO:000008208) and ALT2 (PRO:000008209).
I don't see anything controversial here. The PRO tracker is at http://sourceforge.net/tracker/?group_id=266825&atid=1135711, do you want to go ahead?
I'll submit the request when I get a chance.
I would go for output of some data transformation which takes the measurements as input.
7. bilirubin. The total bilirubin test measures the mass of all
bilirubin molecules that are either free (indirect), part of some
(glucosyluronic acid)bilirubin molecule (direct), or part of some
protein-bilirubin conjugated molecule (delta bilirubin). In ChEBI,
bilirubin and (glucosyluronic acid)bilirubin do not share a parent,
appropriately, since one is part-of the other (and not is-a the
other). Nevertheless the measurement is the mass of just the
bilirubin components per unit volume of serum, not the mass of the
glucuronide or protein.
Having said all that, I’m comfortable with just bilirubin as the
analyte.
8. globulin and albumin/globulin ratio. Globulin is simply
calculated as total protein – albumin. Then the A/G ratio is albumin/
globulin, i.e., albumin/(total protein – albumin). How do we
represent the fact that some observations are calculations derived
from measurements?
I'm happy to help with the PRO submissions if needed, let me know.
Furthermore, there’s no PRO term for globulin.
9. lactate dehydrogenase (or LDH). No PRO term for it, or any of its
isoenzymes.
Thanks. I'll try it and call for help when needed.
Hope some of those help, if there is interest we could maybe either use an IAO call or have a distinct one with only the people interested in QTT/lab tests.
Philippe Rocca-Serra
This is a great effort, Thanks. Melanie has addressed a number of issues
already.
I am also cc-ing OBI developers for heads-up and keeping the thread
I have 2 comments so please see inline (you have to scroll down a long
way!) about the evaluant which are missing from OBI and about an
alternative way to model enzymatic activity + a proposal for another
template that would include an extension to hook up biological process /
patholocal process to the assays being carried out.
> values? Also, I don�t have a use case at this point, but it
> would not
> surprise me if some tests measure the ratio directly, and thus you
> have a measurement of a ratio as opposed to the ratio of two
> measurements. For this test, however, we just need the ratio
> of two
> measurements. Also, this test has no units of measure as they
> �cancel
> out� (mg/dL / mg/dL = 1). Do we need to do anything special with
> �unitless� measurements? Also, I assume the evaluant still makes
> involved in an ionic (or mostly ionic, let�s not discuss the
> fact that
> no bond is 100% ionic) bond, and thus the analyte was Ca(2+). Of
> course, from my clinical training, I knew the relevance of the
> difference for patient care, but not the difference at an
> atomic and
> molecular level. Nevertheless, suffice it to say that ~50% of Ca
> atoms are bound up so tightly in serum that they have no relevant
> physiological effect in the metabolism of Ca. So then, is the
> difference b/w total and ionized Ca testing the analyte? Or
> is it the
> �space� or �compartment� in which the analyte is contained?
> At this
> point I favor the latter, since I don�t think any Calcium atom
> in the
> body has all its electrons entirely to itself. The ionized Ca
> test
> seems to be measuring the concentration of Ca(2+) ions in
> solution,
> where the albumin-binding of Ca removes a substantial portion
> of Ca
> atoms out of solution. The issue then would be how to model the
> �water-space-with-no-albumin� of blood, because that�s
> definitely not
> the specimen, the specimen is serum, and then the methodology
> of the
> test somehow isolates and thus measures just the portion of Ca not
> bound up and thus free in solution.
>
> Interestingly, there is a few mmol/L difference between total
> sodium
> and ionized sodium. There�s lots of interesting literature on
Do I get right that in order to carry out an 'ionized calcium' test,
serum needs to be treated to deplete it of albumin ?
in other words plain serum can not be used. This points to the need to
add "albumin-free serum" to OBI as a defined class (possibly relying the
'lacks' relation, still under discussion if I recall correctly).
This brings another point I came across when going over the IUPAC
Clinical Chemistry nomenclature (very much in LOINC format):
A number of tests record that the evaluant is Blood but Blood is further
qualified to indicate the fact that blood is drawn from fasted patient.
Again, in order to capture this, we would need to record the
physiological status of the patient from which the evaluant comes from.
SO we would need 'fasted subject/organism' in OBI. I will submit the
tickets on the term tracker.
>
>
>
> 5. aspartate aminotransferase (AST). PRO has aspartate
> aminotransferase, cytoplasmic (PRO:000008153 ) vs. aspartate
> aminotransferase, mitochondrial (PRO:000008154) but no plain
> aspartate
> aminotransferase (the parent of the two isoenzymes is just Protein
> (PRO:000000001). It is canonical to have molecules of AST in your
> blood, so you can�t say that canonically every instance of
> these two
> isoenzymes is located in some cytoplasm or mitochondrion,
> respectively.
>
> We need to request a new term from PRO that is the parent of
> the two
> existing terms. This link:
> http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=cm&part=A3139
> <http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=cm&part=A3139>
> says: �Two distinct forms [of AST] have been identified: a
> cytoplasmic, or soluble isoenzyme, and a mitochondrial isoform.
> Selective measurement of these isoenzymes has no currently
> demonstrated clinical application.�
>
> 6. alanine aminotransferase (ALT). Same problem as for AST,
> except
> the isoenzymes are ALT1 (PRO:000008208) and ALT2 (PRO:000008209).
>
>
> I don't see anything controversial here. The PRO tracker is at
> http://sourceforge.net/tracker/?group_id=266825&atid=1135711
> <http://sourceforge.net/tracker/?group_id=266825&atid=1135711>, do
> you want to go ahead?
>
>
> I'll submit the request when I get a chance.
Could we represent this using the Material class Protein (PRO:000000001)
bearing alanine aminotransferase function GO:0004021 ?
What it fails to capture is the fact that this activity can be detected
in the blood is caused by loss of cellular integrity since most of
alanine aminotransferase activity is confined to cell compartments.
In order to say so, we would have to represent the fact that "those
instances of protein with alanine aminotransferase function" are the
specific output of a process "cell lysis || cell leakage || loss of
cellular integrity" and whose input could be (but not restricted to)
hepatocytes (I am assuming a measurement in human blood sample).
Detecting elevated levels of ALT in blood could be directly related to
an elevation of liver cell damage.
This would require extending the template to include relation such
"is_proxy_for".
In OBI, I've 'experimented" a bit working on a "glucose tolerance test"
trying to relate the fact of measurement glucose concentration following
a bolus of glucose and the biolocal process of insulin resistance
(covered in the Human Phenotype ontology).
Cheers
Philippe
Bill Hogan
From: Bill Hogan <hog...@gmail.com>
Date: Wed, Jan 27, 2010 at 1:45 PM
Subject: Re: [IAO] Laboratory test modeling using spreadsheet template
To: Philippe Rocca-Serra <ro...@ebi.ac.uk>
values? Also, I don’t have a use case at this point, but it
would not
surprise me if some tests measure the ratio directly, and thus you
have a measurement of a ratio as opposed to the ratio of two
measurements. For this test, however, we just need the ratio
of two
measurements. Also, this test has no units of measure as they
“cancel
out” (mg/dL / mg/dL = 1). Do we need to do anything special with
“unitless” measurements? Also, I assume the evaluant still makes
sense as serum, as we could report this ratio for, say, urine.
I would go for same than above, if it doesn't fit straight away
let's not try and force it. If you have several of those type of
assays (the one above sounded more like a very specific case), we
may create a new pattern. The current pattern was really analyte
in evaluant gets measured and outputs scalar measurement datum.
Ratios are extremely common and I recommend the creation of a template to handle them.
Based on http://en.wikipedia.org/wiki/BUN-to-creatinine_ratio: In
medicine, the BUN-to-creatinine ratio, also BUN-creatinine ratio,
BUN/creatinine ratio or BUN:Cr, is the ratio of two serum
laboratory values, the blood urea nitrogen (BUN) and serum creatinine.
So you first measure BUN, then Creatinine, and then compute a
ratio. Could you have BUN assay, serum creatinine assay using the
pattern, and then the ratio would be the output of a data
transformation?
Regarding units, we have measurement datum with subclasses like
"ration of collected to emitted light" (in OBI). Scalar
measurement datum is the specific subclass which has measurement
value and measurement unit.
I don't like at all the approach of creating units that are specific to each type of ratio. Where the units of measure cancel, I think we either leave it empty or have some unit of measure called 'unity' or something (which I am not sure makes sense even).
4. I wasted hours investigating the difference between total
calcium
and ionized calcium. It seemed to me that any calcium atom
would be
involved in an ionic (or mostly ionic, let’s not discuss the
fact that
no bond is 100% ionic) bond, and thus the analyte was Ca(2+). Of
course, from my clinical training, I knew the relevance of the
difference for patient care, but not the difference at an
atomic and
molecular level. Nevertheless, suffice it to say that ~50% of Ca
atoms are bound up so tightly in serum that they have no relevant
physiological effect in the metabolism of Ca. So then, is the
difference b/w total and ionized Ca testing the analyte? Or
is it the
“space” or “compartment” in which the analyte is contained?
At this
point I favor the latter, since I don’t think any Calcium atom
in the
body has all its electrons entirely to itself. The ionized Ca
test
seems to be measuring the concentration of Ca(2+) ions in
solution,
where the albumin-binding of Ca removes a substantial portion
of Ca
atoms out of solution. The issue then would be how to model the
“water-space-with-no-albumin” of blood, because that’s
definitely not
the specimen, the specimen is serum, and then the methodology
of the
test somehow isolates and thus measures just the portion of Ca not
bound up and thus free in solution.
Interestingly, there is a few mmol/L difference between total
sodium
and ionized sodium. There’s lots of interesting literature on
No, that is not necessary. A Ca(2+)-selective electrode is used to measure ion activity of Ca(2+). Because all the calcium bound to albumin (and otherwise bound and thus not 'available' or 'diffusable') does not affect the measurement, voila, you get a measurement of ionized calcium. Then, the lab uses a conversion formula (I don't know what it is) to convert ion activity to mg/dL.
in other words plain serum can not be used. This points to the need to add "albumin-free serum" to OBI as a defined class (possibly relying the 'lacks' relation, still under discussion if I recall correctly).
It uses plain serum. It is just clever for measuring only those ions that are "active in solution".
This brings another point I came across when going over the IUPAC Clinical Chemistry nomenclature (very much in LOINC format):
A number of tests record that the evaluant is Blood but Blood is further qualified to indicate the fact that blood is drawn from fasted patient. Again, in order to capture this, we would need to record the physiological status of the patient from which the evaluant comes from.
An interesting problem, too. Yes, recording the events that preceded the test is necessary and presents lots of new issues.
SO we would need 'fasted subject/organism' in OBI. I will submit the tickets on the term tracker.
5. aspartate aminotransferase (AST). PRO has aspartate
aminotransferase, cytoplasmic (PRO:000008153 ) vs. aspartate
aminotransferase, mitochondrial (PRO:000008154) but no plain
aspartate
aminotransferase (the parent of the two isoenzymes is just Protein
(PRO:000000001). It is canonical to have molecules of AST in your
blood, so you can’t say that canonically every instance of
these two
isoenzymes is located in some cytoplasm or mitochondrion,
respectively.
We need to request a new term from PRO that is the parent of
the two
existing terms. This link:
http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=cm&part=A3139
<http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=cm&part=A3139>
says: “Two distinct forms [of AST] have been identified: a
cytoplasmic, or soluble isoenzyme, and a mitochondrial isoform.
Selective measurement of these isoenzymes has no currently
demonstrated clinical application.”
6. alanine aminotransferase (ALT). Same problem as for AST,
except
the isoenzymes are ALT1 (PRO:000008208) and ALT2 (PRO:000008209).
I don't see anything controversial here. The PRO tracker is at
http://sourceforge.net/tracker/?group_id=266825&atid=1135711
<http://sourceforge.net/tracker/?group_id=266825&atid=1135711>, do
you want to go ahead?
I'll submit the request when I get a chance.
Could we represent this using the Material class Protein (PRO:000000001) bearing alanine aminotransferase function GO:0004021 ?
Although the protein was named according to its function, and this representation of its function is correct, it still requires a PRO term that refers to its structure.
What it fails to capture is the fact that this activity can be detected in the blood is caused by loss of cellular integrity since most of alanine aminotransferase activity is confined to cell compartments.
In order to say so, we would have to represent the fact that "those instances of protein with alanine aminotransferase function" are the specific output of a process "cell lysis || cell leakage || loss of cellular integrity" and whose input could be (but not restricted to) hepatocytes (I am assuming a measurement in human blood sample).
Detecting elevated levels of ALT in blood could be directly related to an elevation of liver cell damage.
This would require extending the template to include relation such "is_proxy_for".
Two thoughts. (1) if we stop to represent this level of detail for every test, we'll make insufficient progress on covering lots of commonly performed tests. Not that we shouldn't ultimately take the time, it's just a matter of priorities. (2) From the following link "Since aminotransaminases are ubiquitous in their cellular distribution, serum elevations may occur with a variety of nonhepatobiliary disorders." http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=cm&part=A3139
In OBI, I've 'experimented" a bit working on a "glucose tolerance test" trying to relate the fact of measurement glucose concentration following a bolus of glucose and the biolocal process of insulin resistance (covered in the Human Phenotype ontology).
There are lots of challenge tests and we ought to ultimately work toward a template for these as well.
Also, there are "peak" and "trough" drug levels that will be complicated as well. That said, my own personal priority is to get as many tests done as fit the existing template. But if that's not the priority of the group, I'm happy to redirect my efforts.
Cheers
Philippe
Philippe Rocca-Serra
thanks for the answers.
>
> Two thoughts. (1) if we stop to represent this level of detail for
> every test, we'll make insufficient progress on covering lots of
> commonly performed tests. Not that we shouldn't ultimately take the
> time, it's just a matter of priorities.
I am with you on the need to be fast. The Templates and MappingMaster
plugin should allow to achieve this. We can revisit on due time (and
funding permitting) if we want to drill further.
> (2) From the following link "Since aminotransaminases are ubiquitous
> in their cellular distribution, serum elevations may occur with a
> variety of nonhepatobiliary disorders."
> http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=cm&part=A3139
> <http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=cm&part=A3139>
Correct. I have been cautious though and did mention that the enzymes
expression was not restricted to hepatocytes.
The link also states that: "However, elevations exceeding 10 to 20 times
the reference are uncommon in the absence of hepatic cell injury. Since
the concentration of ALT is significantly less than AST in all cells
except hepatic cytosol, ALT serum elevations are less common in
nonhepatic disorders"
>
>
> In OBI, I've 'experimented" a bit working on a "glucose tolerance
> test" trying to relate the fact of measurement glucose
> concentration following a bolus of glucose and the biolocal
> process of insulin resistance (covered in the Human Phenotype
> ontology).
>
>
> There are lots of challenge tests and we ought to ultimately work
> toward a template for these as well.
> Also, there are "peak" and "trough" drug levels that will be
> complicated as well. That said, my own personal priority is to get as
> many tests done as fit the existing template. But if that's not the
> priority of the group, I'm happy to redirect my efforts.
Your priority is right and by all means, I don't want to put you off! I
only wanted to know if you felt the template was enough or could do with
extension (in well defined cases) in order to relate to capture extra
knowledge.
best wishes
Philippe