with the only proviso being that they exhibited typical experimental
scatter (without obvious errors in the sample preparation) and were
from a range of different years.
It is important to consider that these data were not collected
with systematic analysis or multimethod benchmarking in mind.
Firstly, for systematic analysis, the stock solutions would be ideally
identical (to control for variability in sample preparation). However, the stocks for these experiments were often prepared fresh for
individual experiments (HEWL solutions, NAG3 dilutions), or
stocks were refreshed at least once a year (NAG3 stocks). Only
the stock of fluorescently labeled HEWL for MST measurements
was constant for all experiments. Secondly, for systematic analysis, a
larger sample of different instruments, operated by experts, would
be required. Therefore, the analyses in this chapter are emphatically
not intended as a statistically reliable exploration of the real error
and consistency in determining the K d for this interaction by
different methods. Rather, the data are an example of the typical
variation that one might expect when using these protocols in the
context of teaching, at a single location (see Note 3).
There are a number of different ways in which the data from
independent replicate experiments could be analyzed, and there is
not sufficient space here to explore them all. Nor are the data
necessarily of sufficient quality and consistency to make such a
comparison worthwhile. Therefore, to assess the robustness of the
assay over the replicates presented for each technique, we have
chosen a single approach: to globally fit the five datasets to a shared
value of K d . If all of the replicates for each technique are measuring
the same interaction under the same conditions, with sufficient data
quality, it should be possible to fit the datasets to a single shared
value of K d .
In performing a global fit with more than one fitting parameter
(in addition to experimental dependent and independent variables),
one must make a decision on whether to fit other parameters in the
fitting equation as a global parameter (a shared value across all
datasets) or a local parameter (an individual value for each dataset).
This decision depends on whether that parameter is measured by an
instrument with an absolute calibration, and upon the apparent
consistency of that absolute calibration.
For example, in fitting fluorescence intensity data from different plate readers, with different (fixed) bandwidths of excitation
and emission, it is not expected that the fluorescence intensities of
free and bound HEWL would be consistent between datasets, even
after normalization, since the amplitude of signal change will be
bandwidth-dependent to some degree. Therefore, it seems sensible
to allow the fluorescence intensities of free and bound HEWL to be
fitted as a local parameter. A similar treatment has been applied here
to MST data (where small variations between datasets can be seen in
the signals for free and bound HEWL, even for the same
Interactions by Multiple Methods
51
scatter (without obvious errors in the sample preparation) and were
from a range of different years.
It is important to consider that these data were not collected
with systematic analysis or multimethod benchmarking in mind.
Firstly, for systematic analysis, the stock solutions would be ideally
identical (to control for variability in sample preparation). However, the stocks for these experiments were often prepared fresh for
individual experiments (HEWL solutions, NAG3 dilutions), or
stocks were refreshed at least once a year (NAG3 stocks). Only
the stock of fluorescently labeled HEWL for MST measurements
was constant for all experiments. Secondly, for systematic analysis, a
larger sample of different instruments, operated by experts, would
be required. Therefore, the analyses in this chapter are emphatically
not intended as a statistically reliable exploration of the real error
and consistency in determining the K d for this interaction by
different methods. Rather, the data are an example of the typical
variation that one might expect when using these protocols in the
context of teaching, at a single location (see Note 3).
There are a number of different ways in which the data from
independent replicate experiments could be analyzed, and there is
not sufficient space here to explore them all. Nor are the data
necessarily of sufficient quality and consistency to make such a
comparison worthwhile. Therefore, to assess the robustness of the
assay over the replicates presented for each technique, we have
chosen a single approach: to globally fit the five datasets to a shared
value of K d . If all of the replicates for each technique are measuring
the same interaction under the same conditions, with sufficient data
quality, it should be possible to fit the datasets to a single shared
value of K d .
In performing a global fit with more than one fitting parameter
(in addition to experimental dependent and independent variables),
one must make a decision on whether to fit other parameters in the
fitting equation as a global parameter (a shared value across all
datasets) or a local parameter (an individual value for each dataset).
This decision depends on whether that parameter is measured by an
instrument with an absolute calibration, and upon the apparent
consistency of that absolute calibration.
For example, in fitting fluorescence intensity data from different plate readers, with different (fixed) bandwidths of excitation
and emission, it is not expected that the fluorescence intensities of
free and bound HEWL would be consistent between datasets, even
after normalization, since the amplitude of signal change will be
bandwidth-dependent to some degree. Therefore, it seems sensible
to allow the fluorescence intensities of free and bound HEWL to be
fitted as a local parameter. A similar treatment has been applied here
to MST data (where small variations between datasets can be seen in
the signals for free and bound HEWL, even for the same
Interactions by Multiple Methods
51