Santoro-Bolen equation

[leofl...@yahoo.com.br]

Hi all,

I'm new on fityk and I'm trying to define a simple nonlinear equation
to fit data equilibrium of protein unfolding, as described by Santoro
and Bolen ( http://www.genominfo.org/html/UploadFile/article_7_200609.pdf
), page 2.

I'm trying this:

define Santoro (a0=intercept, a1=slope, b0=intercept, b1=slope,
e=intercept, f=slope) =

((a0+a1*x)+(b0+b1*x)*exp(-(e/0.592090122+f*x/0.592090122)))/(1+exp(-(e/
0.592090122+f*x/0.592090122)))

(the constant is RT adjusted to my experimental conditions)

and is expected to fit very well in sigmoid-like data, as in fig. 5 of
the pdf. However, I can only get a linear fitting, not shaped to data,
even adjusting parameters manually.

LM give an error. NM and GA works but gives me non fitted curve...

Miss I something in editing the equation, is there a tip for this?

Thanks for any help.

[Marcin Wojdyr]

The definition of the function looks good, but there is a problem with
initial values.
It seems that for the initial values you used the Lev-Mar algorithm is
stuck. Perhaps because the function has some derivatives equal zero,
but I'm not sure what's the reason, I haven't investigated it.
Anyway, if you change the initial parameters (e.g. to a0=intercept,
a1=slope, b0=0, b1=0, e=0, f=0) , there is a chance that the fitting
will work.
Let us know if this helps.

BTW, 3 years ago I considered adding a support for sigmoid-like functions:
http://groups.google.com/group/fityk-users/browse_thread/thread/01588d56015a862f
but I haven't found users who would be interested in such a feature.

Marcin

--
Marcin Wojdyr | http://www.unipress.waw.pl/~wojdyr/

[leofl...@yahoo.com.br]

You're right!

I normalized my experimental data (raging from 0 to 1) and choose
initial reasonable values for a0, a1, b0 and b1. e and f remained as
intercept and slope, just because those were the parameters that I
want to obtain from the fit. And it just fitted perfectly. The problem
is that I had poorly sampled data in linear parts of the curve, that
are mainly controlled by the two linear equation a0... and b0... So
the program had little to use to guess about this part of the fit.

Consider to add support for sigmoid curves, because there are a lot of
biological events that has a sigmoid shape.

On Sep 6, 1:33 pm, Marcin Wojdyr <woj...@gmail.com> wrote:
> The definition of the function looks good, but there is a problem with
> initial values.
> It seems that for the initial values you used the Lev-Mar algorithm is
> stuck. Perhaps because the function has some derivatives equal zero,
> but I'm not sure what's the reason, I haven't investigated it.
> Anyway, if you change the initial parameters (e.g. to a0=intercept,
> a1=slope,  b0=0, b1=0, e=0, f=0) , there is a chance that the fitting
> will work.
> Let us know if this helps.
>

> BTW, 3 years ago I considered adding a support for sigmoid-like functions:http://groups.google.com/group/fityk-users/browse_thread/thread/01588...

> but I haven't found users who would be interested in such a feature.
>
> Marcin
>
> On Sun, Sep 6, 2009 at 15:29,
>
>
>
> leofloat...@yahoo.com.br<leofloat...@yahoo.com.br> wrote:
>
> > Hi all,
>
> > I'm new on fityk and I'm trying to define a simple nonlinear equation
> > to fit data equilibrium of protein unfolding, as described by Santoro

> > and Bolen (http://www.genominfo.org/html/UploadFile/article_7_200609.pdf