Detecting an analyte in a flash and glow reaction

Patent metadata

Field	Value
Publication number	US-2017191111-A1
Application number	US-201615384207-A
Country	US
Kind code	A1
Filing date	Dec 19, 2016
Priority date	Dec 31, 2015
Publication date	Jul 6, 2017
Grant date	—

Abstract

Accurate measurements of the presence or absence of an analyte (e.g., MRSA) in a sample are provided. For example, the sample can be subjected to an activation reagent (potentially after an initial reagent has already been added), which can cause a flash signal that increases and then decreases over time. Signal data points can be measured from the flash signal using a detector. A quadratic regression function that fits the signal data points can be determined. An accuracy of the quadratic fit can be determined, as well as a signal-to-background ratio. A difference between a signal-to-background term and an accuracy term can be used as a score that is compared to a threshold to determine whether the analyte is present in the sample.

First claim

What is claimed is:

1 . A method for determining whether a sample is positive for an analyte, the method comprising:
collecting a plurality of background data points using a detector, the background data points collected during a background time interval; collecting a plurality of signal data points using the detector, wherein the plurality of signal data points is collected during a signal time interval that is after the background time interval, wherein the signal time interval includes a portion of time after the sample is activated by an activation reagent; computing, by a computer system, a signal-to-background ratio using the plurality of background data points and the plurality of signal data points; determining, by the computer system, a regression function that fits the plurality of signal data points, wherein the regression function includes at least a second order polynomial; computing, by the computer system, an accuracy value of the regression function fitting the plurality of signal data points; calculating, by the computer system, a score by adjusting an accuracy term with a signal-to-background term, the accuracy term including the accuracy value, and the signal-to-background term including the signal-to-background ratio; and comparing the score to a classification threshold to determine a classification of whether the sample is positive or negative for the analyte.

2 . The method of claim 1 , wherein:
if the score is less than the threshold, the sample is determined to be negative for the analyte, and if the score is greater than or equal to the threshold, the sample is determined is determined to be positive for the analyte.

3 . The method of claim 1 , further comprising:
computing a weighted sum of squared errors of the regression function relative to the plurality of signal data points; and calculating the accuracy term by multiplying the accuracy value by the weighted sum of squared errors.

4 . The method of claim 1 , wherein the accuracy value is selected from: include r-squared, adjusted r-squared, p-value of F-test, and predicted residual error sum of squares (PRESS) statistic.

5 . The method of claim 1 , wherein the regression function includes a parabolic coefficient of a second order term, and wherein the accuracy term includes the accuracy value multiplied by the parabolic coefficient.

6 . The method of claim 1 , further comprising:
computing the signal-to-background term by multiplying the signal-to-background ratio by a reciprocal of a first constant.

7 . The method of claim 1 , where the threshold is chosen to obtain a desired sensitivity and specificity with reference to representative sets of data having known results for presence of the analyte.

8 . The method of claim 1 , wherein the detector is a light detector, and wherein the plurality of signal data points form a luminescence signal.

9 . The method of claim 1 , wherein the analyte is methicillin-resistant staphylococcus aureus (MRSA).

10 . The method of claim 1 , wherein the background time interval is prior to activation of the sample by the activation reagent, and wherein the plurality of background data points are collected from the sample before the activation of the sample.

11 . The method of claim 1 , further comprising:
calculating a median of at least a portion of the background data points; calculating a deviation using the median; determining a first cutoff using a sum of the median and the deviation; counting a first number of background data points that are greater than the first cutoff; comparing the first number to a second cutoff; and when the first number is greater than the second cutoff, determining that an error exists.

12 . The method of claim 11 , further comprising:
repositioning the sample relative to the detector.

13 . The method of claim 12 , wherein the plurality of signal data points are not collected until the error is corrected as determined by the first number being less than or equal to the second cutoff.

14 . The method of claim 1 , further comprising:
calculating a mean of at least a portion of the background data points; comparing the mean to a predetermined factor; and when the mean is greater than the predetermined factor, determining that an error exists.

15 . The method of claim 14 , further comprising:
determining the predetermined factor using data points collected when a sensor of the detector is blocked.

16 . The method of claim 14 , wherein the plurality of signal data points are not collected until the error is corrected as determined by the first number being less than or equal to the second cutoff.

17 . The method of claim 1 , further comprising:
determining, by the computer system, a maximum value of the plurality of signal data points, wherein the regression function is determined using a time of the maximum value.

18 . The method of claim 17 , wherein the regression function is centered at the time of the maximum value.

19 . The method of claim 17 , further comprising:
determining a high-order regression function using the plurality of signal data points, the high-order regression function being a 4 th order polynomial or higher; computing an accuracy measure of the high-order regression function; comparing the maximum value of the plurality of signal data points to one or more initial thresholds; selecting an accuracy threshold based on the comparison to the one or more initial thresholds; and comparing the accuracy measure to the accuracy threshold to determine whether the sample is negative for the analyte based on the accuracy measure being less than the accuracy threshold.

20 . The method of claim 19 , further comprising:
determining that the sample is possibly positive for the analyte based on the accuracy measure being greater than the accuracy threshold; and proceeding to use the comparison of the score to the classification threshold to determine the classification of whether the sample is positive or negative for the analyte.

21 . The method of claim 19 , further comprising:
filtering the plurality of signal data points to obtain a filtered dataset by applying a statistical smoothing function to the plurality of signal data points; and using the filtered dataset to determine the high-order regression function.

22 . The method of claim 21 , wherein filtering the plurality of signal data points includes:
for each of at least a portion of the signal data points:
calculating a median value of adjacent data points and the signal data point; and
replacing the signal data point with the median value.

23 . The method of claim 22 , wherein each of the median values are determined from three signal data points.

24 . The method of claim 23 , wherein the portion of the signal data points is offset by one data point from a beginning signal data point and ending one data point before an ending signal data point of the signal time interval,
wherein the ending signal data point is replaced with a median of a last two signal data points, wherein the beginning signal data point is replaced by a corresponding value from the high-order regression function if the beginning signal data point is greater than the corresponding value plus a specified number times a standard deviation, and wherein the accuracy measure is of the high-order regression function fit to the filtered dataset.

Assignees

• Roche Molecular Systems Inc

Inventors

• Rey Diego Ariel
• Hayter Paul
• Sharma Manoj
• Reinig Nick

Classifications

• C12Q1/14 Primary

  Streptococcus; Staphylococcus · CPC title

• G16C20/70 Primary

  Machine learning, data mining or chemometrics · CPC title