Method and system for improving the evaluation of an interaction between an analyte and a ligand using a biosensor

1. A method for improved evaluation of an interaction between an analyte in a fluid sample and a ligand immobilized on a sensor surface of a biosensor, comprising the steps of:

providing a first sensor surface, and providing a second sensor surface having a ligand immobilized thereon,

allowing a plurality of fluid samples to flow across said first and said second sensor surface wherein the fluid samples comprise a solvent at known concentration,

measuring changes of refractive index at the first and second sensor surface during the flow of each fluid sample,

determining a response value and a reference-subtracted response value for each fluid sample based on said changes and creating a data set comprising said values, each fluid sample forming a point in the data set,

selecting a reliable set comprising at least five samples from the data set, plotting the response value against the reference-subtracted response value for the reliable set and fitting a curve to them,

and further comprising steps a)-f) below, said steps being performed in alphabetical order for every other fluid sample in the data set

a) calculating leverages h i ,

b) labelling points with h i >h i cut as outliers, and if the number of outliers, n o , is less than n o max =min(n−p−1, floor((n−1)/2)), continue; otherwise, go to step f) below,

c) forming a clean set with outliers removed and calculating n clean Studentized residuals t i ext using a regression based on this clean set, where n clean represents size of the clean set,

d) labelling points with t i ext >t(a/(2n clean ), n clean −p−1) as outliers, and if n o <n o max , continue; otherwise, go to step f) below,

e) forming a clean set with outliers removed again, if outliers were found in previous step d); if χ 2 >χ 2 cut , where χ 2 is calculated from regression of the clean set, calculating the n clean Studentized residuals t i ext for all points in the clean set and labelling outliers just as was done in previous step d),

f) if any high-leverage outliers were detected in step b) and n o <n o max , calculating t i pred for these high-leverage points using a regression based on the clean, outlier-free set, and removing the outlier label for these points if t red <t(α/2, n clean −p).

2. The method according to claim 1 , wherein the method further comprises the steps of:

g) calculating an influence measure, referred to as DFFITS, for each point and labeling points with DFFITS i >f(t(α/(2n clean ), n clean −p−1), h i cut ) as outliers; if the number of outliers, n o , is less than n o max =min(n−p−1, floor((n−1)/2)), continue; otherwise, go to step f),

and wherein said step g) is performed before or after any of the steps a) and b).

3. The method according to claim 1 , wherein the method further comprises determining a clean set by

i) calculating a slope for all point area data inside each point,

ii) calculating a difference between average and median of the data inside each point,

iii) calculating a difference between the slope calculated in step i) and median of all slopes calculated in step i),

iv) sorting the data from step iii),

v) determining a clean set of the sorted data from step iv) by removing data with largest difference to median,

vi) using the clean set to calculate normal difference to median,

vii) classifying a point as outlier if difference against median is larger than a first predetermined value and data from step ii) is larger than a second predetermined value.

4. The method according to claim 3 , wherein said first predetermined value is 20 times the value determined in step vi).

5. The method according to claim 3 , wherein said second predetermined value is 1 resonance units.

6. The method according to claim 1 , wherein χ 2 =max(χ o 2 , χ 2 rel ).

7. The method according to claim 1 , wherein the solvent is an organic solvent, preferably dimethyl sulfoxide.

8. A biosensor system, comprising a processor, for improved evaluation of an interaction between an analyte in a fluid sample and a ligand immobilized on a sensor surface of a biosensor, characterized in being arranged to perform the steps of the method according to claim 1 .

9. A biosensor instrument for evaluation of an interaction between an analyte in a fluid sample and a ligand immobilized on a sensor surface of a biosensor, the instrument having a biosensor including a first sensor surface, and a second sensor surface having a ligand immobilized thereon,

means for causing a plurality of fluid samples to flow across said first and said second sensor surface wherein the fluid samples comprise a solvent at known concentration, a

a measuring device for measuring changes of refractive index at the first and second sensor surface during the flow of each fluid sample,

a processor operable to determine a response value and a reference-subtracted response value for each fluid sample based on said changes and creating a data set comprising said values, each fluid sample forming a point in the data set and for selecting a reliable set comprising at least five samples from the data set, plotting the response value against the reference-subtracted response value for the reliable set and fitting a curve to them, the processor being further operable to carry out steps a)-f) below, said steps being performed in alphabetical order for every other fluid sample in the data set:

a) calculating leverages h i ,

b) labelling points with h i >h i cut as outliers, and if the number of outliers, no, is less than n o max =min(n−p−1, floor((n−1)/2)), continue; otherwise, go to step f) below,

c) forming a clean set with any-outliers removed and calculating n clean Studentized residuals t i ext using a regression based on this clean set, where n clean represents size of the clean set,

d) labelling points with t i ext >t(a/(2n clean ), n clean −p−1) as outliers, and if n o <n o max, continue; otherwise, go to step f) below,

e) forming a clean set with outliers removed again, if outliers were found in previous step d); if χ 2 >χ 2 cut , where χ 2 is calculated from regression of the clean set, calculating the n clean Studentized residuals t i ext for all points in the clean set and labelling outliers just as was done in previous step d),

f) if any high-leverage outliers were detected in step b) and n o ≤n o max , calculating t i pred for these high-leverage points using a regression based on the clean, outlier-free set, and removing the outlier label for these points if t red <t(α/2, n clean −p),

thereby improving the evaluation of an interaction between an analyte in a fluid sample and a ligand immobilized on a sensor surface of a biosensor.