Systems and methods for processing PCR signal cross talk

A machine learning-based correction matrix addresses cross-talk interference in PCR detection systems, improving signal differentiation and accuracy by training on prior data and penalizing negative signals, especially considering temperature variations.

WO2026236171A1PCT designated stage Publication Date: 2026-11-12ROCHE DIAGNOSTICS INTERNATIOAL AG +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/US2026/026639
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-09
Filing Date
2026-05-05
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Existing PCR detection systems face challenges in correcting cross-talk signals between multiple channels due to overlapping wavelength bands, which affect the accuracy of signal detection and differentiation, and traditional methods fail to adapt well across different devices and assays.

Method used

A machine learning system is employed to train a correction matrix using prior PCR data to mitigate cross-talk interference, utilizing a loss function to minimize differences in corrected data and penalize negative signals, with temperature modulation considered to enhance accuracy.

Benefits of technology

The system effectively corrects cross-talk interference, improving the accuracy of PCR signal detection by distinguishing between positive and negative signals, thereby enhancing the reliability of PCR test results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure US2026026639_12112026_PF_FP_ABST
    Figure US2026026639_12112026_PF_FP_ABST
Patent Text Reader

Abstract

A method for detecting the presence of a molecule in a biological sample with a Polymerase Chain Reaction (PCR) detection system, including applying a trained machine learning correction matrix to obtained PCR signals from a plurality of channels, the matrix trained with PCR training data to correct signal cross-talk between all of the plurality channels. The training includes applying a loss function representing differences in cross-talk corrected PCR data respective to the plurality of channels. Determining the presence or lack of presence of a target molecule in the biological sample is based on determining whether the corrected PCR signals represent positive or negative signals.
Need to check novelty before this filing date? Find Prior Art

Description

SYSTEMS AND METHODS FOR PROCESSING PCR SIGNAL CROSS TALK CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Provisional Patent Application No. 63 / 802758, filed May 9, 2025, the entire contents of which is herein incorporated by reference.FIELD

[0002] The application relates to polymerase chain reaction (PCR) signal processing.BACKGROUND

[0003] The Polymerase Chain Reaction (PCR) is an in vitro method for enzymatically synthesizing or amplifying defined nucleic acid sequences. The reaction typically uses two oligonucleotide primers that hybridize to opposite strands and flank a template or target DNA sequence that is to be amplified. Elongation of the primers is catalyzed by a heat-stable DNA polymerase. A repetitive series of cycles involving template denaturation, primer annealing, and extension of the annealed primers by the polymerase results in an exponential accumulation of a specific DNA fragment.Fluorescent probes are typically used in the process to facilitate detection and quantification of the amplification process.

[0004] Typical PCR detection systems for analyzing radiation emissions from PCR experiments include two or more filters that are each operable to isolate a wavelength range for further analysis. For example, each optical filter typically allows substantially all radiation in a defined wavelength range to pass. However, the probes or markers typically emit with partially overlapping wavelength bands, and a filter's band pass typically includes a region of this overlap such that each detection channel will typically receive signal emitted from other probes. Such cross-talk signals tend to affect the real signal(s) of interest. Accordingly, it is desirable to correct for such cross-talk signals in each detection channel. One traditional way of doing this is to determine quantitative cross-talk coefficients that can be used to correct for cross-talk signals in each detection channel. However, such coefficients may not adapt well between individual devices, assays, and tests where determination of the coefficients was tailored to or a generalization for use across or among these multiple factors.

[0005] It is therefore desirable to provide systems and methods for determining crosstalk coefficients in curves, such as sigmoid-type or growth curves, and PCR curves in particular, which overcome the above and other problems.SUMMARY

[0006] A PCR detection system is configured to mitigate cross-talk interference between multiple PCR channels utilizing a machine learning system. The machine learning system is trained utilizing prior PCR data to create a correction matrix of correction parameters that can be applied to PCR curve data used to test biological samples for particular molecules (e.g., indicative of conditions or diseases).

[0007] In some embodiments, a loss function is utilized / minimized that represents differences in cross-talk corrected data operated on by a correction matrix for signals identified / labeled as negative (indicating a negative signal / test). A signal may be labeled as negative based on determining that a signal has a magnitude that remains below a predetermined threshold over at least a predetermined period during a PCR cycle (e g., FIG. 2B).

[0008] In some embodiments, the loss function comprises a summation represented by:|max (Xccc[i]) - min (Xccc[i])|i, label[i]=negative wherein cc represents a matrix of cross-talk corrected PCR signal data.

[0009] Minimizing loss may be performed utilizing a gradient descent process, for example. The loss function generally penalizes a presence of a relatively non-flat corrected curve for a “negative” (or non-activated) fluorescence signal. In some embodiments, a weighted term penalizes the presence of a negative intensity values in corrected data and can include the expression:W' Si, label[i]=negative mit (0, Tj - bj)wherein w represents a learnable weight factor, r represents relative fluorescence intensity of a corrected PCR curve, and b represents the uncorrected baseline PCR signal.

[0010] In some embodiments, the training includes using a dataset that is relatively evenly distributed between positively and negatively labeled data and anapproximately proportional number of positively and negatively labeled signals for each of the plurality of channels.

[0011] In some cases, temperature modulation is a significant aspect of performing PCR tests and may influence cross-talk between channels. A cross-talk correction matrix may be trained utilizing temperature information and trained with parameters to correct cross-talk based on temperature utilized and monitored during PCR cycle(s). The temperature-based correction training may be performed by including temperature in the training data and correction parameters in the loss function and correction matrices described herein.

[0012] In some embodiments, separate correction matrices are trained and maintained for separate PCR assays to correct for cross-talk when utilizing the different assays in further testing. A software tool may be integrated with a PCR testing platform to generate separate cross-talk correction matrices for separate assays and / or other parameters / conditions that may influence PCR signal cross-talk.BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Objects, features, and advantages of the disclosure will be readily apparent from the following description of certain embodiments taken in conjunction with the accompanying drawings in which:

[0014] FIG. 1 is an illustrative flow diagram of a PCR testing process.

[0015] FIG. 2A is an illustrative chart of a frequency domain spectrum for five PCR channels.

[0016] FIG. 2B is an illustrative chart of PCR signal curves for five channels.

[0017] FIG. 3 is an illustrative flow chart for performing PCR with cross-talk correction according to some embodiments.

[0018] FIG. 4 is an illustrative flow chart for performing cross-talk correction while penalizing negative RI interference according to some embodiments.

[0019] FIG. 5 A is an illustrative chart of PCR signal curves without cross-talk correction.

[0020] FIG. 5B is an illustrative chart of PCR signal curves corrected for cross-talk interference according to some embodiments.

[0021] FIG. 6 is an illustrative chart representing cycles of training cross-talk correction according to some embodiments.

[0022] FIG. 7 is an illustrative chart of temperatures utilized during PCR testing cycles for particular tests according to some embodiments.

[0023] FIG. 8 is a chart of uncorrected and corrected PCR signals utilizing some embodiments.

[0024] FIG. 9 is an illustrative diagram of a computer system that may be utilized to implement processes and systems disclosed herein.

[0025] It is contemplated that elements disclosed in one embodiment may be beneficially utilized in other embodiments without specific recitation. Moreover, the description of an element in the context of one drawing is applicable to other drawings illustrating that element.DETAILED DESCRIPTION

[0026] A PCR optical detection system is configured to correct cross-talk interference between multiple PCR channels utilizing a machine learning process with a matrix of correction parameters applied to PCR signals obtained in reference to a sample under test (e.g., for testing of the sample for the presence or lack of molecules indicative of particular conditions or diseases). The matrix is trained with prior PCR data and loss function to correct signal cross-talk between all of multiple PCR channels.

[0027] FIG. 1 is an illustrative flow diagram of a PCR testing process. A biological sample (e.g., patient fluid or tissue sample) is obtained at block 100 for testing with a PCR testing system (e.g., one of Roche Diagnostic cobas® 58xx PCR testing systems). The PCR system includes multiple channels each adapted to receive electromagnetic signals respective to a particular spectrum (e.g., color) and used to signal the presence of particular molecules.

[0028] The PCR system initiates / performs PCR amplification at block 110. During amplification, at block 120, PCR signals are generated while the system collects and stores the signals through each of the channels. PCR signals through each of the channels can be represented as PCR curves representing frequencies or “colors” 130a, 130b, ..., 130e (e.g., as illustrated in FIG. 2B).

[0029] FIG. 2A is an illustrative chart of a frequency domain spectrum for five PCR channels. A spectrum for signals collected by each channel may overlap with and / or interfere with detecting and differentiating signals attributable to different sources (e.g., fluorophores) which particular channels are configured to detect.

[0030] FIG. 2B is an illustrative chart of PCR signal curves for five channels. A signal received by a channel may be considered “positive” (detected) or “negative” (not detected) signal, depending on such factors including whether the received signal exceeded a particular threshold of intensity and / or the shape of the signal curve. For example, a positive signal may be represented by a curve of intensity that typically increases after a number of PCR cycles (e.g., heating / cooling) and then plateaus such as illustrated by signal 210 with color 130b.

[0031] “Negative” signals should be represented by a curve maintaining a relatively flat intensity below a particular threshold. However, cross-talk between spectra for separate channels may influence these curves such as by an increased intensity attributable to a source that bleeds into or partially overlaps with a spectrum associated with another channel. For example signals 220 with colors 130a, 130c, 130d, and 130e should represent negative signals but may have been influenced by interference from the source of signal 210. In some cases, the interference could be significant and influence whether the signal could be considered a valid negative or positive signal.

[0032] FIG. 3 is an illustrative flow chart for performing PCR with cross-talk correction according to some embodiments. At block 310, PCR signals across multiple channels are obtained (e.g., 5 channels), such as from a PCR testing device. The PCR signals may represent and be obtained over multiple PCR cycles (e.g., repeated heating and cooling to amplify a specific DNA sequence) and remain generally unmodified (e.g., unfiltered / uncorrected) from the form generated by a device.

[0033] The signals may represent PCR data for multiple sets of samples analyzed separately using the same or similar devices for training a machine learning model to correct for cross-talk between channels. Each set of signals for a sample can be represented as a PCR curve, that is a curved line representing intensity values of the data over time that is generated by fitting and / or interpolation between points (e.g., least squares, polynomial interpretation, splines, linear and / or non-linear regression).

[0034] In some embodiments, the number of samples used for training exceeds around 500 or more, a relatively even distribution between negative and positive signals, and can involve use of a particular assay (e.g., for testing for the presence of COVID, influenza). Depending on the device(s), assay(s), and / or type(s) of samples tested, and / or deviation from previously established correction parameters, thenumber of samples used for training may be adapted to include more or fewer samples for training.

[0035] At block 320, the sample data is used to train a correction parameter matrix to be applied to signals for later-tested samples. That training data is labeled with predetermined characteristics (e.g., what accurate PCR testing of the samples should reflect as “negative” or positive signals). In some embodiments, the parameter matrix is two dimensional with the number of rows and columns each representing the number of channels (e.g., 5). The training includes iterating multiple passes through each of the training samples obtained at block 320 (i.e., each pass representing an epoch), and adjusting the matrix parameters based on a loss function. In some embodiments, the loss function is represented by:

[0036] label[i]=negative |m3X (Xccc[l]) mit (Accc[l])|

[0037] where Xccc = MNXN • Jlraw,

[0038] where a “negative” represents a PCR curve labeled for a channel on the basis that a sample with predetermined characteristics tested should not cause a significant modification or increase in the respective channel’s signal over the cycles of a PCR test (i.e., the PCR curve should remain relatively flat and below a threshold intensity level), where MAW represents a parameter correction matrix, Xraw represents raw or unfiltered PCR data and Xcccrepresents corrected data after application of the correction matrix to unfiltered data.

[0039] An initial correction matrix MAW may represent an estimate of correction parameters such as based on observations of the impact of positive signals on negative signals (e.g., bleeding / overlap) for various channels.

[0040] A loss is calculated using a loss function representing differences in the “negative” signals after application of a given correction matrix MAMV. Minimizing loss may be determined such as by using gradient descent or other methods known to one of ordinary skill in the art while adjusting values (e.g., incrementally) of the correction parameters in the matrix based on calculated loss determined from the loss function. Multiple passes through the training data may occur (where each pass represents an epoch) until the loss stabilizes or plateaus. Once the calculated loss is minimized in magnitude and variability to a satisfactory degree, further iterations of adjusting the correction matrix may end.

[0041] At block 330, raw or unfiltered PCR data is obtained with respect to further samples (e.g., under test for particular patient(s)). At block 340, the correction parameter matrix optimized from block 320 is used to correct the raw or unfiltered PCR data pertaining to the further samples, using the equation Xccc = MNXN • aw. At block 350, the corrected PCR data for a sample under test is returned and further analysis of the corrected data may occur (e.g., identifying the outcome of a particular test for presence of a targeted molecule such as a particular target DNA molecule). The results of the test may be obtained / retrieved such as through a testing platform used for testing patients (e.g., in a laboratory or hospital).

[0042] FIG. 4 is an illustrative flow chart for performing cross-talk correction while penalizing negative RI interference according to some embodiments. Depending on a particular PCR testing platform and / or assay, interference may include anomalies other than from signals that overlap between multiple channels. For example, interference may include a negative radiofrequency influence not associated with an overlapping signal. In some embodiments, such influence not associated with overlap or bleed is weighted negatively so that overlap interference is not obscured.

[0043] At block 410, unfiltered PCR training data is obtained from tested samples and labeled with predetermined characteristics (e.g., what accurate PCR testing of the samples should reflect as “negative” or positive signals). A correction parameter matrix MNXN is established with initial values for further training / refinement and correction of raw or unfiltered PCR signal data, where rows and columns for each of the channels of a PCR system.

[0044] At block 420, the training data is iterated fully through (an epoch) with a loss function multiple times for refining the correction matrix similar to the function and process described with respect to FIG. 3 and further including a weighted term for penalizing the influence of a negative radiofrequency (RF) interference:

[0045] ArgminMi, iabei[i\=negative |max (Xccc[i]) - min (Xccc[i])| +W' Si, label[i]=negative mit (0, r(- - bj)

[0046] where Xccc= MAW • Xraw, where MAW represents a parameter correction matrix, aw represents raw or unfiltered PCR data and Xcccrepresents corrected data after application of the correction matrix to unfiltered data, n is the level of measured radiofrequency (rf) intensity and bi is a baseline measure of background rf measuredfor a particular channel where the channel’s signal is labeled as a “negative” signal, and w is a weight factor for a term representing.

[0047] At block 430, a refined correction matrix MAW is output after iterating through the training data with the loss function such as with use of gradient descent. At block 440, new unfiltered PCR data is obtained and the correction matrix is applied to the unfiltered data and corrected data Xccc is obtained.

[0048] FIG. 5 A is an illustrative chart of PCR signal curves without cross-talk correction. The curves represent fitting to raw unfiltered PCR data from five different channels. A signal 500 is illustrative of a typical positive PCR signal that begins to increase and plateau after a number PCR cycles. The remaining curves of PCR signals 510 reflect typical “negative” signals for channels in which the signals do not reach a particular threshold of intensity and remain generally flat.

[0049] FIG. 5B is an illustrative chart of PCR signal curves corrected for cross-talk interference according to some embodiments. Curves 520 represent PCR curves that are corrected from unfiltered curves 510 of negative signals using corrected matrix MNXN. The corrective impact may include correction of intensities of negative signals based on cross-talk from a higher-intensity positive signal 500, for example, and may thereby avoid misinterpretation of a negative signal as a positive signal.

[0050] FIG. 6 is an illustrative chart representing cycles of training cross-talk correction according to some embodiments. A set of labeled training data is processed using a loss function such as described with respect to FIGs. 3 and 4. As each iteration (epoch) through training data occurs and the correction parameters are refined, the level of loss calculated by the function may decrease as shown by training curve 610. The loss eventually flattens as the correction parameters do not significantly change over further epochs. A separate set of testing data (apart from the training data) is tested using the loss function and a testing curve 620 is shown representing the calculated loss. Ideally, losses calculated from training and testing data should be relatively close and stabilize correspondingly.

[0051] FIG. 7 is an illustrative chart of temperatures utilized during PCR testing cycles for particular tests according to some embodiments. In some embodiments, assays may be used for PCR testing that allows for multiple different tests to be performed. The assay(s) may be configured to prompt activation of signals for particular channels based on the presence of particular target molecules (e.g., DNA, antigen, protein).

[0052] Parameter correction matrices may be generated / configured based on the additional influence of different temperatures applied during testing (e.g., for testing different molecules). The correction matrices may be applied such as with the equation Xccc= KNXKXK MAMV Xraw., with a loss function such as described above in connection with FIGs. 3 and 4, where a temperature correction matrix KNXKXK has two dimensions representing the number of temperatures K used and a dimension representing the number N channels used.

[0053] FIG. 8 is a chart of uncorrected and corrected PCR signals utilizing some embodiments. A set of unfiltered negative PCR signals is represented by curves 810. Varying temperatures are applied across the signals (58° C, 80° C, 93° C). The differences between the negative signal intensities can vary significantly depending on the temperature utilized for a particular test, while ideally (e.g., absent crosstalk) negative signals should represent relatively flat signals close to each other in intensity regardless of temperature. Utilizing training data with negative / positive labeled signals and temperatures, a parameter correction matrix is generated (e.g., as described above in connection with FIG. 7) to correct intensity for crosstalk and is shown as cross-talk corrected curves 820.

[0054] FIG. 9 is an illustrative diagram of a computer system that may be utilized to implement processes and systems disclosed herein. Computer systems may be utilized for assisting in performing the methods described herein, such as for controlling devices, receiving, transmitting, and processing data from PCR devices, controlling inflow and / or outflow of samples and detection agents to and from the devices, and controlling the manufacturing processes. The computer systems may utilize any suitable number of subsystems. Examples of such subsystems are shown in FIG. 9 in computer system 10.

[0055] In some embodiments, a computer system includes a single computer apparatus, where the subsystems can be the components of the computer apparatus. In other embodiments, a computer system can include multiple computer apparatuses, each being a subsystem, with internal components. A computer system can include desktop and laptop computers, tablets, mobile phones and other mobile devices. In some embodiments, a cloud infrastructure (e.g., Amazon Web Services), a graphical processing unit (GPU), etc., can be used to implement the disclosed techniques.

[0056] The subsystems shown in FIG. 9 are interconnected via a system bus 75. Additional subsystems such as a printer 74, keyboard 78, storage device(s) 79, monitor 76, which is coupled to display adapter 82, and others are shown. Peripherals and input / output (I / O) devices, which couple to I / O controller 71, can be connected to the computer system by any number of means known in the art such as input / output (I / O) port 77 (e.g., USB, FireWire®). For example, VO port 77 or external interface 81 (e.g. Ethernet, Wi-Fi, etc.) can be used to connect computer system 10 and PCR device 90 to a wide area network such as the Internet, a mouse input device, or a scanner. The interconnection via system bus 75 allows the central processor 73 to communicate with each subsystem and to control the execution of a plurality of instructions from system memory 72 or the storage device(s) 79 (e.g., a fixed disk, such as a hard drive, or optical disk), as well as the exchange of information between subsystems. The system memory 72 and / or the storage device(s) 79 may embody a computer readable medium. Another subsystem is a data collection device 85, such as a camera, microphone, accelerometer, and the like. Any of the data mentioned herein can be output from one component to another component and can be output to the user.

[0057] A computer system can include a plurality of the same components or subsystems, e.g., connected together by external interface 81 or by an internal interface. In some embodiments, computer systems, subsystem, or apparatuses can communicate over a network. In such instances, one computer can be considered a client and another computer a server, where each can be part of a same computer system. A client and a server can each include multiple systems, subsystems, or components.

[0058] Aspects of embodiments can be implemented in the form of control logic using hardware (e.g. an application specific integrated circuit or field programmable gate array) and / or using computer software with a generally programmable processor in a modular or integrated manner. As used herein, a processor includes a single-core processor, multi-core processor on a same integrated chip, or multiple processing units on a single circuit board or networked. Based on the disclosure and teachings provided herein, a person of ordinary skill in the art will know and appreciate other ways and / or methods to implement embodiments of the present invention using hardware and a combination of hardware and software.

[0059] Machine learning models utilized herein, such as for analyzing magnetic sensor data for determining the presence of MNPs and attached molecules, may include one or more of a Naive Bayes (NB) model, a logistic regression (LR) model, a random forest (RF) model, a support vector machine (SVM) model, an artificial neural network model, a multilayer perceptron (MLP) model, a convolutional neural network (CNN), a Large Language model (LLM), and / or other machine learning or deep leaning models, etc. The machine learning models can be updated / trained using a supervised learning technique, an unsupervised learning technique, etc.

[0060] Any of the software components or functions described in this application may be implemented as software code to be executed by a processor using any suitable computer language such as, for example, Java, C, C++, C#, Objective-C, Swift, or scripting language such as Perl or Python using, for example, conventional or object-oriented techniques. The software code may be stored as a series of instructions or commands on a computer readable medium for storage and / or transmission. A suitable non-transitory computer readable medium can include random access memory (RAM), a read only memory (ROM), a magnetic medium such as a harddrive or a floppy disk, or an optical medium such as a compact disk (CD) or DVD (digital versatile disk), flash memory, and the like. The computer readable medium may be any combination of such storage or transmission devices.

[0061] Such programs and / or data may also be encoded and transmitted using carrier signals adapted for transmission via wired, optical, and / or wireless networks conforming to a variety of protocols, including the Internet. As such, a computer readable medium may be created using a data signal encoded with such programs. Computer readable media encoded with the program code may be packaged with a compatible device or provided separately from other devices (e.g., via Internet download). Any such computer readable medium may reside on or within a single computer product (e.g. a hard drive, a CD, or an entire computer system), and may be present on or within different computer products within a system or network. A computer system may include a monitor, printer, or other suitable display for providing any of the results mentioned herein to a user.

[0062] Any of the methods described herein may be at least partially performed with a computer system including one or more processors, which can be configured to perform the steps. Thus, embodiments can be directed to computer systems configured to perform elements of the steps of any of the methods described herein,potentially with different components performing a respective steps or a respective group of steps. Steps of methods herein can be performed at a same time or in a different order. Additionally, portions of these steps may be used with portions of other steps from other methods. Also, all or portions of a step may be optional.Additionally, any of the steps of any of the methods can be performed with modules, units, circuits, or other means for performing these steps.

[0063] The specific details of particular embodiments may be combined in any suitable manner without departing from the spirit and scope of embodiments of the invention. However, other embodiments of the invention may be directed to specific embodiments relating to each individual aspect, or specific combinations of these individual aspects.

[0064] The above description of example embodiments of the invention has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise form described, and many modifications and variations are possible in light of the teaching above.

[0065] A recitation of "a", "an" or "the" is intended to mean "one or more" unless specifically indicated to the contrary. The use of “or” is intended to mean an “inclusive or,” and not an “exclusive or” unless specifically indicated to the contrary. Reference to a “first” component does not necessarily require that a second component be provided. Moreover reference to a “first” or a “second” component does not limit the referenced component to a particular location unless expressly stated.

[0066] All patents, patent applications, publications, and descriptions mentioned herein are incorporated by reference in their entirety for all purposes. None is admitted to be prior art.

Claims

CLAIMS1. A method for detecting a presence of a molecule in a biological sample with a Polymerase Chain Reaction (PCR) detection system, the method comprising:applying one or PCR testing cycles to a biological sample under test utilizing a plurality of PCR channels of a PCR optical detection system;obtaining PCR signals from the plurality of PCR channels in response to applying the one or more PCR testing cycles;applying a trained machine learning correction matrix of correction parameters to the obtained PCR signals, the matrix trained with PCR training data to correct signal cross-talk between all of the plurality channels, wherein the training comprised applying a loss function representing differences in cross-talk corrected PCR data respective to the plurality of channels;receiving an output of corrected signals based on application of the matrix of correction parameters;based on the output of corrected signals, determining that one or more of the PCR signals represents a positive or negative PCR signal;based on determining that one or more of the PCR signals represents a positive or negative signal, respectively determining the presence or lack of presence of a target molecule in the biological sample.

2. The method of claim 1 wherein the training comprised applying a gradient descent process to the loss function.

3. The method of claim 1 wherein the loss function comprises a summation represented by:Si, label\i\=negative |m3X (Xccc[i]) — min (Xccc[i])|wherein Xcccrepresents a matrix of cross-talk corrected PCR signal data.

4. The method of claim 3 wherein a signal is labeled as negative based on determining that a signal has a magnitude that remains below a predetermined threshold over at least a predetermined period during a PCR cycle.

5. The method of claim 1 wherein the loss function comprises a weighting aspect that penalizes a presence of a negating impact on correcting a PCR curve.

6. The method of claim 5 wherein the weighting aspect comprises:W' Si, label[i]=negative mit (0, r(- - hj)wherein w represents a learnable weight factor, r represents a corrected PCR signal curve, and b represents a baseline PCR signal.

7. The method of claim 1 wherein the training comprised using a relatively even distribution of negative and positive signal training data.

8. The method of claim 1 wherein the even distribution comprises an approximately proportional number of positive and negative signals for each of the plurality of channels.

9. The method of claim 1 wherein the PCR training data comprised temperature information pertaining to PCR testing cycles of the PCR training data, the machine learning matrix configured to correct for cross-talk based on temperature information obtained with respect to PCR testing cycles.

10. The method of claim 1 wherein the trained machine learning correction matrix is selected from a plurality of machine learning matrices of correction parameters separately trained with respect to a plurality of corresponding separate PCR assays.

11. The method of claim 10 wherein the corresponding separate PCR assays are configured to perform different tests for the presence of particular molecules indicative of conditions or diseases.

12. A Polymerase Chain Reaction (PCR) detection system for detecting a presence of a molecule in a biological sample, the system comprising:one or processors programmed and configured to:obtain PCR signals of a plurality of PCR channels from a test of a biological sample by a PCR testing device;apply a trained machine learning correction matrix of correction parameters to the obtained PCR signals, the matrix trained with PCR training data to correct signal cross-talk between all of the plurality channels, whereinthe training comprised applying a loss function representing differences in cross-talk corrected PCR data respective to the plurality of channels;receive an output of corrected signals based on application of the matrix of correction parameters;based on the output of corrected signals, determine that one or more of the PCR signals represents a positive or negative PCR signal;based on determining that one or more of the PCR signals represents a positive or negative signal, respectively determine the presence or lack of presence of a target molecule in the biological sample.

13. The system of claim 12 wherein the training comprises applying a gradient descent process to the loss function.

14. The system of claim 12 wherein the loss function comprises a summation represented by:Si, label\i\=negative |m3X (Xccc[l]) mit (7fcccld])lwherein cc represents a matrix of cross-talk corrected PCR signal data.

15. The system of claim 14 wherein a signal is labeled as negative based on determining that a signal has a magnitude that remains below a predetermined threshold over at least a predetermined period during a PCR cycle.

16. The system of claim 12 wherein the loss function comprises a weighting aspect that penalizes a presence of a negating impact on correcting a PCR curve.

17. The system of claim 16 wherein the weighting aspect comprises:W' Si, label[i]=negative mit (0, Tj - bj)wherein w represents a learnable weight factor, r represents a corrected PCR signal curve, and b represents a baseline PCR signal.

18. The system of claim 12 wherein the training comprised using a relatively even distribution of negative and positive signal training data.

19. The system of claim 12 wherein the even distribution comprised an approximately proportional number of positive and negative signals for each of the plurality of channels.

20. The system of claim 12 wherein the PCR training data comprised temperature information pertaining to PCR testing cycles of the PCR training data, the machine learning matrix configured to correct for cross-talk based on temperature information obtained with respect to PCR testing cycles.

21. The system of claim 12 wherein the trained machine learning correction matrix is selected from a plurality of machine learning matrices of correction parameters separately trained with respect to a plurality of corresponding separate PCR assays.

22. The system of claim 21 wherein the corresponding separate PCR assays are configured to perform different tests for the presence of particular molecules indicative of conditions or diseases.

23. A non-transitory computer readable medium comprising instructions for causing one or more processors to:obtain PCR signals of a plurality of PCR channels from a test of a biological sample by a PCR testing device;apply a trained machine learning correction matrix of correction parameters to the obtained PCR signals, the matrix trained with PCR training data to correct signal cross-talk between all of the plurality channels, wherein the training comprised applying a loss function representing differences in cross-talk corrected PCR data respective to the plurality of channels;receive an output of corrected signals based on application of the matrix of correction parameters;based on the output of corrected signals, determine that one or more of the PCR signals represents a positive or negative PCR signal;based on determining that one or more of the PCR signals represents a positive or negative signal, respectively determine the presence or lack of presence of a target molecule in the biological sample.