Estimation of delta-Cq values with confidence from qPCR data

Patent metadata

Field	Value
Publication number	US-9779206-B2
Application number	US-201113704655-A
Country	US
Kind code	B2
Filing date	May 16, 2011
Priority date	Jun 17, 2010
Publication date	Oct 3, 2017
Grant date	Oct 3, 2017

Abstract

The invention describes how to estimate delta-Cq values from measured (raw-)Cq values gained from PCR measurements and how to calculate confidence intervals for them. This is realized by the following processing steps: A noise model, which might be constructed on some training PCR data, calculates the distribution of the true target material concentration of a single well for an observed measurement results. Said distribution is calculated for all types of measurement results including “Numeric” raw-Cq values as well as Cq being “Undetected”, which denotes that no fluorescence signal was detected during all cycles and thus corresponds to no or very few target molecules.

First claim

The invention claimed is:

1. A method for determining a quality control criterion for a target raw-Cq value of a target analyte obtained from quantitative polymerase chain reaction (PCR) measurements, the method comprising:
determining, by a processor of a computing device, a target raw-Cq value from quantitative PCR measurements, the target raw-Cq value being representative of a concentration or activity of a target analyte from a first sample;
determining by the processor, a reference raw-Cq value from quantitative PCR measurements, the reference raw-Cq value being representative of a concentration or activity of a reference analyte from at least one of (i) the first sample, and (ii) one or more other samples;
calculating, by the processor, an unbounded delta-Cq value and respective confidence information based on a calculated difference between the target raw-Cq value and the reference raw-Cq value;
calculating, by the processor, a bounded delta-Cq value and respective confidence information, wherein the bounded delta-Cq value is bounded by applying an upper bound, a lower bound, or both, the confidence information corresponding to the bounded delta-Cq value being used to determine a quality control criterion for the target raw-Cq value of the target analyte; and
using the quality control criterion to perform a quality control action.

2. The method of claim 1 , wherein the applying of the upper bound, lower bound or both, generates a variance of the bounded delta-Cq value which is smaller than a variance of the unbounded delta-Cq value, for at least one observable Cq value.

3. The method of claim 1 , wherein the target analyte is a nucleic acid selected from the group consisting of DNA, RNA, or mRNA.

4. The method of claim 1 , wherein the target raw-Cq value and the reference raw-Cq value comprise threshold cycle (cT) values.

5. The method of claim 1 , wherein each of the target raw-Cq and the reference raw-Cq measurement values comprise one of:
(i) a numeric value representative of the concentration or activity of the target analyte or the reference analyte, respectively;
(ii) an “Undetected” value representative of a concentration or activity below a limit of detection (LOD) of the target analyte or the reference analyte, respectively; or
(iii) an “Invalid” value indicating the occurrence of a severe problem during measurement and/or that the measurement value should be ignored.

6. The method of claim 1 , further comprising:
measuring, by the processor, one or more replicates for each of the target analyte and/or the reference analyte; and
combining, by the processor, the one or more replicates of each of the target analyte and/or the reference analyte into each of the unbounded and/or bounded delta-Cq values.

7. The method of claim 6 , further comprising:
eliminating, by the processor, outliers within the one or more replicates using an automated procedure before determining the target raw-Cq value.

8. The method of claim 1 , wherein the lower bound is determined based on a training data set by constrained optimization,
wherein the lower bound is optimized to be as high as possible to make a variance of bounded normalized values as small as possible,
wherein a constraint depends on an application of normalized values, and
wherein the lower bound is constrained to be sufficiently low such that for said application, no distinction between values at and below the lower bound needs to be made.

9. The method of claim 1 , wherein the calculating of the unbounded delta-Cq value and respective confidence information is based on a noise model, and
wherein a true value is modeled as a normal distribution with a mean equal to one of the measured raw-Cq values and a standard deviation calculated from a parameterized function comprising parameters fitted based on training data.

10. The method of claim 1 , wherein the calculating of the unbounded delta-Cq value and respective confidence information is based on a noise model, and
wherein a measured value is modeled as a normal distribution with a mean equal to a true value and a standard deviation calculated from a parameterized function comprising parameters fitted based on training data.

11. The method of claim 5 , wherein, in the event that one of the target raw-Cq and the reference raw-Cq values comprises an “Undetected” value, a noise model assigns a fixed distribution to a true value, said fixed distribution having a finite or an infinite expected value, a finite or an infinite variance, or a bounded, half-bounded or unbounded support.

12. The method of claim 5 , wherein the quality control action comprises at least one of the following:
(i) report or reject the one or more of the unbounded and/or bounded target raw-Cq values;
(ii) repeat or not repeat measurements of the target raw-Cq value and the reference raw-Cq value;
(iii) decide whether the one or more of the unbounded and/or bounded target raw-Cq values are used for a decision; or
(iv) determine whether the one or more of the unbounded and/or bounded target raw-Cq values and/or a combination thereof exceeds a threshold.

13. The method of claim 1 , wherein the upper bound is determined based on a training data set by constrained optimization,
wherein the upper bound is optimized to be as low as possible to make a variance of bounded normalized values as small as possible,
wherein a constraint depends on an application of normalized values, and
wherein the upper bound is constrained to be sufficiently high such that for said application, no distinction between values at and above the upper bound needs to be made.

14. The method of claim 1 , wherein the unbounded and bounded delta-Cq values are (i) normalized values, or (ii) estimated values calculated using an estimator.

15. The method of claim 1 , wherein the reference raw-Cq value is a mean of the PCR measurement of the reference analyte from each of the at least one of (i) the first sample, or (ii) the one or more other samples, and
wherein the calculating the unbounded delta-Cq value and the bounded delta-Cq value comprises calculating the difference between a predetermined constant and the calculated difference between the target raw-Cq value and the reference raw-Cq value.

16. A system for determining a quality control criterion for a target raw-Cq value of a target analyte obtained from quantitative polymerase chain reaction (PCR) measurements, comprising:
memory storing instructions that are executable to perform operations comprising:
determining a target raw-Cq value from quantitative PCR measurements, the target raw-Cq value being representative of the concentration or activity of a target analyte from a first sample;
determining a reference raw-Cq value from quantitative PCR measurements, the reference raw-Cq value being representative of the concentration or activity of a reference analyte from at least one of (i) the first sample, and (ii) one or more other samples;
calculating an unbounded delta-Cq value and respective confidence information based on the target raw-Cq value and the reference raw-Cq value;
calculating a bounded delta-Cq value and respective confidence information, wherein the bounded delta-Cq value is bounded by applying an upper bound, a lower bound, or both, the confidence information corresponding to the bounded delta-Cq value being used to determine a quality control criterion for the target raw-Cq value of the target analyte; and
using the quality control criterion to perform a quality control action.

17. The system of claim 16 , wherein the apply

Assignees

• Weber Karsten
• Feder Inke Sabine
• Siemens Healthcare Diagnostics Inc

Inventors

• Weber Karsten
• Feder Inke Sabine

Classifications

• C12Q1/6851

  Quantitative amplification · CPC title

• G16B40/00 Primary

  ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding · CPC title

• G16B25/00

  ICT specially adapted for hybridisation; ICT specially adapted for gene or protein expression · CPC title

• G06F19/24 Primary

  Physics · mapped topic

• G06F19/20

  Physics · mapped topic