JMP for antibody titration curves

[#1]

Hi

 Hi All, im a very novice jmp user and am trying to analyze an antibody titration curve using JMP. I tried to fit a 4PHill. However i got some very weird values on my estimates with my inflection point (-1.72)  and lower asymptote (-26453.17) having negative values while the upper asymptote showed 97.11.  In additions the P value was not significant for both the inflection point and lower asymptote (approx 0.98 on both). Am i using the wrong type of curve? How do i approach this? any help is appreciated! Thanks!

[#2]

Please post your data (JMP data table) if possible.

Also, are you using the Hill model in the model library associated with the Analyze > Specialized Modeling > Nonlinear platform? Or are you using the Analyze > Specialized Modeling > Fit Curve platform?

[#3]

@Mark_Bailey..i tried it both ways and it still gave me the negative values:-(

[#4]

Hi @Apssub 

as @Mark_Bailey posted already, a sample would be really helpful, you can also kind of anonymize it to avoid unveiling confidential information.

Did you have a look at the examples, e.g. in the scripting index you will find that script ( > Help >Scripting Index > search for 4P HILL).

Simply execute and have a look at the result.

Names Default To Here( 1 );
dt = Open( "$SAMPLE_DATA/Nonlinear Examples/Bioassay.jmp" );
obj = dt << Fit Curve(
	Y( :Toxicity ),
	X( :log Conc ),
	Group( :formulation )
);
obj << Fit Logistic 4P Hill;

[#5]

Hi,

Thanks for responding

Please see attached file.

Not sure what to understand from the P values since both lower asymptote and inflection point have higher P values suggesting no significance on this curve..how can i use this to determine the optimal mass of antibody from the titration 

Thanks!

[#6]

Hi @Apssub ,

I think the problem is the model chosen for the data.  A 3 parameter model would be more appropriate and why you are getting negative values.  Notice the shape of the curve. Going from 0 to some number and then tapering off at some asymptote. That is the shape of a 3 parameter model.

Hope that helps.

[#7]

@Chris...thanks so much..the Logistic 3P works great and resolved the issue of negative values...now it looks better with P<0.001..however i am still lost as to how i extrapolate this to derive the optimal titered antibody mass.

[#8]

Hi,

To add to @Chris_Kirchberg, I wonder if you actually need a logistic model at all because your data is essentially an exponential growth curve.

See below the comparison of the Exponential 3P, Logistic 3P, and Logistic 4P Hill models: the Exponential 3P model is the most effective at describing your data (lowest AICc).

Best,

TS

[#9]

<@Thierry_S..thanks for the additional tips...a colleague recommended a bivariate fit as well and weighting to determine the linear region. But once that is determined what concentration of antibody should be chosen from the linear region (in my case i get 3 points in the linear region) . I am still confused on what to choose>

[#10 · Accepted solution]

Adding to @Chris_Kirchberg's comment, another common cause for failing to converge to chosen model is that it is over-specified. For example, the dilution curve might be a 4PLC, but if experimentally only a portion of the curve is observed, then it is difficult to estimate all of the parameters, and a simpler model will be more successful. Otherwise, you must extend the domain to observe more of the curve if you want to fit the more complex model.