Forums

I love units. I spend an inordinate amount of time trying to convince
experimentalists to pay more attention to units. But the current proposal to
REQUIRE units consistency in a valid SBML model should be rejected.

First, it is entirely possible to have different units in different parts of
a model. A simple example would be expressing cell surface receptor binding
using extracellular agonist concentrations and binding constants (Kd)
expressed in mg/dl while all the intracellular concentrations are expressed
in Molar or nM. No inconsistency is created so long as the total binding
capacity has units compatible with the rate laws in the second messenger
signaling system model. This is true because the concentration of the
extracellular signal and its Kd appear in the equilibrium binding expression
as a unitless ratio. I've made this point on multiple occasions, notably in
a long SBML-discuss thread on simple Michaelis-Menten rate laws, but the
units police apparently have not been convinced.

It's a different story when you have A = B and the units on the two sides
are inconsistent. It seems to me it would be far simpler for us assume that
when an equation appears in a SBML model, the parameters on the two sides of
the equal sign MUST have units that result in equality. If the STATED
parameter units do NOT result in equality, the units are wrong. But you
don't have to fix the units for the numerical solver to get the right
answer. In some important sense, the solver knows what the units must be and
acts accordingly.

It is true that there are some apparent inconsistencies in the published
units in the Voit paper cited as an example of inconsistent units. But it
should be obvious that all the units in the published paper were added after
the hard modeling work was completed. Solvers don't track units.

Has anyone had the courtesy to write to Eberhard asking him to comment on
the units? To me, these apparent unit inconsistencies simply demonstrate
that the stated units are wrong either because of a publisher's or printer's
error or an oversight. But the model accounts for the data and, really, only
the numerical values of the parameters are relevant to the numerical
solutions.

Something would be deeply wrong with SBML if it forced us to exclude the
Voit model because the units police claim it says A = B when A and B have
different units even though it is obvious that the numerical integration has
no need of units and the numerical solution is clearly internally consistent
and accounts for the data.

Vote SHOULD, not MUST!

RDP

PS As an aside, it should be clear that if you are importing a parameter
value into your model that was measured by someone else in some related
experiment, then you MUST convert that parameter to the units used in your
model before you enter it in your SBML file. Naturally, this conversion
will, in general, change its numerical value.

-----Original Message-----
From: sbml-discuss-bounces@caltech.edu
[mailto:sbml-discuss-bounces@caltech.edu] On Behalf Of Michael Hucka
Sent: Friday, December 14, 2007 7:07 PM
To: sbml-discuss@caltech.edu
Subject: [sbml-discuss] SBML survey: should consistency of units be
required?

The topic of this vote is:
Should consistency of units be required?

The URL for the voting page is:
http://www.surveymonkey.com/s.aspx?sm=znse9ys51PMv5mMlma808w_3d_3d

This survey ends 28 December 2007.

The SBML Level 2 Version 3 Release 2 specification is at
http://sbml.org/specifications/sbml-level-2/version-3/

Section 4.10.4, 4.11.3, 4.11.4 and 4.14.4 (and corresponding
validation rules) of the L2v3 specification currently state
a requirement for the consistency of units of the left- and
right-hand sides of mathematical expressions. The statement
is that the units of measurement of the left- and right-hand
sides *must* be equal. The specification is not explicit
about how strictly this requirement is supposed to be
interpreted. We suspect that many SBML developers and users
assumed that the statements really meant the units only
"*should*" be equal. We are therefore running this vote to
determine how strict the SBML community prefers the
requirement about unit consistency to be.

The fundamental issue can be put as follows. Given two
quantities A and B, with units of measurement given for
both, should SBML specify that an expression

A = B

*must* have consistent units, or should it specify only
that it *should* have consistent units? Put another
way, if A and B have identical numerical values but
different units, should SBML stipulate that the expression
above be considered an invalid statement?

Currently, the SBML specification makes it clear that pure
numbers appearing in formulas (e.g., the "1" in an
expression "A = B*1") and parameters have undefined units by
default. (Parameters can be assigned units, but they have
no default units defined.) This is not the same as being
dimensionless. If a number or parameter without units
appears in a formula, they may make it impossible to deduce
the units of the overall expression. This is related,
though not identical, to the present issue; in the present
case, we are concerned with situations in which all the
units are defined.

Physically speaking, two quantities A and B are unequal
if they have different units. However, in the domain of
SBML, there are published models in the literature in which
units were either deliberately or mistakenly not considered
in the system of equations, resulting in inconsistent or
impossible units. For example, in Ferreira AEN, Ponces
Freire AMJ, and Voit EO, Biochem. J. 376:109-121, 2003
(available as BioModels DB model #BIOMD0000000053), some of
the rate laws given in table 1 have units of 1/(M*h) and
others have M/h.

The conundrum is that if SBML stipulates unit
consistency is required for any and all valid SBML models,
then these published models can't be represented in SBML,
yet at the same time, such inconsistencies in a model could
be argued to mean that the model contains errors (because
A != B if units are taken into account).

So what should SBML specify in this area? We would like
your opinion on this matter. Please fill out the short
survey at

http://www.surveymonkey.com/s.aspx?sm=znse9ys51PMv5mMlma808w_3d_3d

This survey ends 28 December 2007.

Thanks,
The SBML Editors

____________________________________________________________
To manage your sbml-discuss list subscription, visit
https://utils.its.caltech.edu/mailman/listinfo/sbml-discuss

For a web interface to the sbml-discuss mailing list, visit
http://sbml.org/forums/

For questions or feedback about the sbml-discuss list,
contact sbml-team@caltech.edu.



____________________________________________________________
To manage your sbml-discuss list subscription, visit
https://utils.its.caltech.edu/mailman/listinfo/sbml-discuss

For a web interface to the sbml-discuss mailing list, visit
http://sbml.org/forums/

For questions or feedback about the sbml-discuss list,
contact sbml-team@caltech.edu.