One can not calculate the true Km value if this is much lower than the substrate concentration. Go for a free binding assay and find a curve and saturation level where the graph become flat, go on y axis and find a maximum velocity. 1/ maximum velocity will represent a simple Km, but it will needs at least equivalent concentration.  I telling you a simple method you should do two experiments, in one experiment keep enzyme concentration in increasing mode and substrate constant, and in another experiment use substrate concentration using increasing and enzyme constant, now you will get two maximum points of activity and both graphs will cut each other at one point, but you will get a different saturation level. Now can calculate the difference of Km from original concentration from the graph, not by calculation itself.

There seem to be two questions here, one general and one specific.

1. The general question seems to be, "Can Km be determined from data that have [S]max lower than the actual Km?"  The answer is a qualified "Yes".  The caveat is that the lower you go with [S]max in relation to Km, the more precision and accuracy is required for the measurements.  I have seen very highly accurate data sets where [S]max < 0.5 Km and you could still get the Km very nicely.  In that case you probably need better than 2% accuracy and precision, and/or a good number of data points covering the available concentration region.  Replicates are always key.  I would not attempt such measurements without doing at least duplicates, but triplicates are always better.

2. The specific question seems to be have something to do with the graph of data you attached.  I am not sure I understand fully what the question really is, but let me make a couple of suggestions for how you could "make" the Km very much higher (if that's what you expected):

• First, delete the last data point at approximately 350 nM and repeat the analysis.  I fully expect that the Km will increase up to several times.  The problem here is that you do you do not have replicates, and so a single data point might have a disproportionate impact.  
• Second, let's assume that there is a tiny bit of baseline rate, which does not come from the enzymatic reaction but is an artifact of some kind.  That happens...  If so, fit the data (after first deleting the last data point) to this equation:

                                    v = V(base) + V(max) [S]/([S] + Km)

where V(base) is an adjustable model parameter.  Use V(base) = 0.02 nmol/sec as the initial estimate.  What will happen then is that your Km will probably "increase" more than ten-fold.  The reason I predict this is that if you remove the last data point (which would be a fluke) then the rest of the points more-or-less sit on a straight line with a small intercept, about equal to 0.02 nmol/sec.

And, as you probably know, if the saturation curve resembles a straight line, the Km is effectively "infinite".

If you want me to show you all this using the DynaFit software, please post the data.  Or you can grab DynaFit from the link below (it's free for academic use) and you can experiment yourself:

       http://www.biokin.com/dynafit/

It sounds to me like you are immobilizing the substrate on the surface, and the limited surface binding capacity is the reason why the substrate concentration can't be increased further. If this is the case, there is a concern that the substrate concentration may be ill-defined because it is not evenly distributed throughout the reaction volume.

I will have to respectfully disagree with Theo's assertion that the coefficient of determination  R2 ("R-square") is a suitable measure of goodness of fit.  This is, unfortunately, a very common misconception. 

Please see the Figure 1 in the attached preprint of an article to appear in Analytical Biochemistry.   The third paragraph in the "Results" section points out that for the very clearly nonlinear data plotted in Figure 1 the linear fit gives R2 = 0.99.  The second paragraph in "Discussion" is quoted below:

[S]ome investigators mistakenly believe that the R2 value measures the "goodness of fit between theoretical curves and experimental data" [18]. In fact, R2 measures how strongly the dependent variable, in this case fluorescence intensity, varies with the independent variable, in this case reaction time. However the qualitative nature of this dependence (e.g. "goodness of fit" to a linear, exponential, hyperbolic, or another fitting model ) is not reflected in R2 at all. That is why it is possible to observe R2 = 0.99 (presumably a "good fit") and a decidedly nonrandom distribution of residuals (a "poor fit", as seen in the bottom panel of Fig 1).

Hope this helps...

Thanks for all the input. It isn't critical for me to calculate the kinetic data for the assay since its meant to be used only as an initial HT screen for inhibitor compounds. However, in the absence of reporting standard Km and Vmax data, does anyone know of a more basic yet somewhat informative kinetic analysis I can do for the assay? I was thinking some sort of kinetic analysis would be good to show in my paper to demonstrate the robustness of the method. I am including a comparison of IC50 data for certain know inhibitors tested with established methods and showing my assay produces similar results. Also, the assay is meant to test a proteolytic enzyme that has multiple subtypes so I was thinking maybe just showing reaction velocities at a constant enzyme and substrate concentration for each subtype may be informative. I originally hoped to calculate the Km for each subtype but I don't think that is doable.  

@ Brian re: "more basic yet somewhat informative kinetic analysis":

Recall that at low enough [S] (in particular [S]

Thank you Dr. Kuzmic. One other question if anyone knows or has a resource: Have any studies been done about the kinetic effects of immobilized substrate, such as in my assay? It would appear to me that the actual apparent concentration at the bulk fluid interface with the immobilized wall would be much higher than expected, however this may be counteracted by exclusion of enzyme from the interface. I'm curious if MM kinetic assumptions even hold in this case. I tried looking around for studies on the topic for elisa or some other similar immobilized substrate assay but came up empty. 

@ Brian re: "Have any studies been done about the kinetic effects of immobilized substrate"

Yes.  Sorry, I lost track of the fact that your substrate is immobilized.  But apparently some of my collaborators managed to wrangle the DynaFit software to do that, using the usual mechanistic assumptions.  You might want to check out these references:

"Detection of enzyme-catalyzed polysaccharide synthesis on surfaces"

Cle, C; Martin, C; Field, RA; Kuzmic, P; Bornemann, S (2010)

Biocatal. Biotransform. 28, 64-71

http://www.biokin.com/publications/pdf/Clex1064.pdf

"DynaFit-a software package for enzymology"

Kuzmic, P (2009)

Methods Enzymol. 467, 247-280.  starting on page 261

http://www.biokin.com/publications/pdf/Kuzm0947.pdf

In that paper the situation is similar to SPR, because the liquid phase was flowing: there is a number of "foundational" studies from early days of SPR showing that at slow flow rates the binding kinetics is similar to bulk solution phase. 

But if you just let the fluid sit, that's probably another ballgame.  Why exactly is the substrate immobilized, in this case?

Hi Dr. Kuzmic, thanks for your help. The basic idea behind my assay is that I am assaying a protease. The substrate construct I made contains an avitag on one end for biotin/streptavidin plate binding and GFP on the free end. As the substrate is cleaved, GFP moves into the bulk solvent giving a progress curve. There is a low initial signal due to the GFP bound to the plate, but since it's on the periphery I still get a very strong signal far above background as the reaction proceeds. There are other FRET based substrates available to do the reaction in solution, but the substrates are very expensive (~$400 for 100 rxns). Also, you have to purchase a separate substrate for every protease subtype (for which there are 8 subtypes) With my assay I can express as much substrate as I want and my construct has the cleavage site for all 8 subtypes. I have known inhibitors that I tested with both the conventional assay as well as my own and I get similar results. 

@ Brian:

Sounds very interesting.  I would not worry (much) about the Km business in this case.  However, you may want to spend some time thinking about what inhibitor concentrations to use, especially if this is a larger-scale screening project.  Yes, you can make your construct so you don't have to spend four bucks on each well, but time and material resources are always finite.   That said, you might want to check out this paper:

http://www.biokin.com/publications/articles.html ... "Optimal Design"

• "Optimal design for the dose-response screening of tight-binding enzyme inhibitors"

Kuzmic, P (2011)

Anal. Biochem. 419, 117-122

http://www.biokin.com/publications/pdf/Kuzm1117.pdf

There is also an online calculator for "optimal design" of the inhibitor dilution series:

• http://www.biokin.com/tools/design.html

Feel free to contact me via http://www.biokin.com/ for further discussion and/or to run your data by me.  No obligation...

Dr. Kuzmic. Thank you for all of your helpful responses. I will definitely not hesitate to run my data by you. I appreciate the offer.

Dr. Nikiforov, Unfortunately the substrate is quite large (over 100 AA) and is mostly linear and unstructured so I believe the distance would be far to great. FRET based methods for this assay exist, but rely on truncated portions of the substrate spanning the cleavage site for specific subtypes of my protease.