WO2026003151A1 - Advanced signal extraction for single-molecule-detection - Google Patents

Abstract

A method and a system to improve the quantitative accuracy and sensitivity of spectral channels of a high-sensitivity optical spectrometry instrument via a computer, comprising the following steps of Using the high-sensitivity optical spectrometry instrument to scan at least one sample in a well plate and forwarding the resulting signal data to the computer; Removing scan defects in the signal data via a software hosted by the computer by identifying photon count bins in the signal data using statistical analysis of underlying photon counting distribution with outlying bins; Extracting signal responses in multiscale via the software to increase the detection sensitivity by maximizing the response SNR, and to widen the response dynamic range by adaptively extending the signal response at the top of standard curve; and Providing real-time feedback of the extracted signal response applying it via the software to the high-sensitivity optical spectrometry instrument while the raw photon count trace is being collected during the scan of the well plate, thereby improving the quantitative accuracy and sensitivity of its spectral channels.

Description

Advanced Signal Extraction for Single-Molecule-Detection

BACKGROUND OF THE INVENTION

[0001] The disclosed invention relates to a method and a system for quantifying fluorescent signal generated byfluorescent label molecules.

[0002] The invention belongs to the technical field of Protein Detection.

[0003] Sensitive optical instrumentation, such as fluorescence microscopes, single molecule detection instruments, or Raman spectrometers have very specific requirements in order to operate with full efficiency. At the ultra-low concentrations, when a fluorescent-labeled molecule is excited by laser light in the femtoliter-level confocal volume, the fluorescence photon burst at the given bintime has to be accurately detected for reaching high sensitivity. At the high concentrations where there is much higher likelihood for multiple molecules traversing through the confocal volume at any given interval, the measured fluorescence response is therefore the sum of responses from multiple molecules which causes the plateau and saturation response and reduces the dynamic range of assay measurements.

[0004] One example of such a high sensitivity instruments is known from the yet unpublished US Provisional application 63/605,656 which discloses a Method and a detection system for detecting protein concentrations in a prepared assay solution consisting of several components, arranged one after another, the components being a fiber coupled laser for emitting laser light, a laser tube lens to focus the laser light, a laser cleanup filter for blocking non-laser wavelengths, a first dichroic filter for passing the laser light and reflecting fluorescence light, an objective lens to focus the laser light to a small excitation volume, the objective also collecting fluorescence light emitted from a sample, a pre-confocal and a post-confocal lens with a pinhole positioned between them to focus and re-focus the fluorescent light, a second dichroic filter, a focus photodiode for auto-focusing on the sample, at least one bandpass filter for spectrum filtering of the fluorescent light, at least one fiber focus lensfor coupling the fluorescent light to an optical fiber, at least one fiber patch cables to couple fluorescently focused light to at least one computer-based detector for signal detection, wherein the sample is located after the objective lens, the second dichroic filter, the focus photodiode and the at least one bandpass filter are forming a De-Multiplexer for the incoming fluorescence light and the at least one bandpass filter, at least one fiber focus lens, at least one fiber patch cables and at least one computer-based detector always consist of same the number and one of each of them are located along a respective fluorescence light.

[0005] For such a high-sensivity instrument the mentioned problems with the plateau and saturation response and the reduction of the dynamic range of assay measurements, need to be addressed somehow. A known approach is the application of a mostly software based, signal extraction algorithm.

[0006] As the closest prior art regarding this approach of signal extraction algorithms can be considered the Erenna and Clarity instruments. Both take the raw photon counts binned by the instrument and estimate how many molecules generated that signal or responses extracted from that raw photon count signal. Specifically, both are looking for flashes of fluorescence emission light with bin counts being significantly brighter than the typical background counts. Each flash of light termed as a DE (detected event) corresponding to be a single molecule passing through the confocal volume. As the sample concentration goes up there is a significant higher probability that two molecules will be in the confocal volume in the same counting bin time. In this case one flash of light represents two or more molecules which cannot be separated by DE only, instead the brightness in number of photons over background noise called EP (Event Photons) is also used to detect molecular events. At very high concentrations of bright samples the noise estimation by the lower 9% of bins floats up due to many molecules passing through the confocal volume at the same time which leads to few bins above the threshold or few detections of molecules. A third metric called TP (total photons) is created to avoid the hook effect of both DE and EP, but increases the uncertainty and efforts required to join three different signals during downstream quantification.

[0007] The task of this invention is therefore to disclose a method to increase the fluorescence detection sensitivity and dynamic range of a high sensitivity protein detection instrument avoiding the problems of the known prior art.

BRIEF SUMMARY OF THE INVENTION

[0008] This task can be solved by a method to improve the quantitative accuracy and sensitivity of spectral channels of a high-sensitivity optical spectrometry instrument via a computer, comprising the following steps of Using the high- sensitivity optical spectrometry instrument to scan at least one sample in a well plate and forwarding the resulting signal data to the computer; Removing scan defects in the signal data via a software hosted by the computer by identifying photon count bins in the signal data using statistical analysis of underlying photon counting distribution with outlying bins; Extracting signal responses in multiscale via the software to increase the detection sensitivity by maximizing the response SNR, and to widen the response dynamic range by adaptively extending the signal response at the top of standard curve; and Providing real-time feedback of the extracted signal response applying it via the software to the high-sensitivity optical spectrometry instrument while the raw photon count trace is being collected during the scan of the well plate, thereby improving the quantitative accuracy and sensitivity of its spectral channels. This advanced signal extraction method is specifically developed to improve the quantitative accuracy and sensitivity of individual spectral channels. First, the algorithm removes the scan defects by identifying those abnormal photon count bins in the well scan signal using statistical analysis of underlying photon counting distribution with outlying bins. Second, the algorithm extracts signal responses in multiscale (multiple integration bintimes) to increases the fluorescence detection sensitivity by maximizing fluorescence response SNR, and widens the fluorescence response dynamic range by adaptively extending the signal response at the top of standard curve. Third, the algorithm is computationally optimized to run and provide realtime feedback of signal response while the raw photon count trace is being collected during the well scan. This approach provides a comprehensive approach to improving the detection capabilities of optical spectrometry instruments. By scanning a sample, analyzing photon count data for anomalies, extracting signal responses at multiple scales, and providing real-time feedback during the scan, this method ensures high precision and enhanced sensitivity in measurement outcomes.

[0009] Advantageous and therefore preferred further developments of this invention emerge from the associated sub claims and from the description and the associated drawings.

[0010] One of those preferred further developments of the disclosed method comprise that the at least one sample in the well plate reflects fluorescence light when being excited by laser light and for the high-sensitivity optical spectrometry instrument the fluorescence detection sensitivity is increased and the fluorescence response dynamic range is widened. This specific feature improves the instrument's ability to detect fluorescence signals by increasing the sensitivity and widening the dynamic range, thus enabling more accurate detection of low- intensity fluorescence signals and better performance across a broader range of sample concentrations.

[0011] Another one of those preferred further developments of the disclosed method comprise that the at least one sample in the well plate is a fluorescent- labeled molecule concentrated in a femtoliter-level confocal volume at ultra-low concentrations. This feature focuses on the ability to accurately detect and analyze extremely low concentrations of fluorescent-labeled molecules, crucial for high-precision biological and chemical assays where sample volumes and molecule concentrations are minimal.

[0012] Another one of those preferred further developments of the disclosed method comprise that for identifying photon count bins in the signal data a single feature value (RE) is created by the software to represent and quantify the raw photon counting trace of the scan of the at least one sample with full dynamic range from ultra-low to very high concentrations. By synthesizing complex data into a single quantifiable feature, the data interpretation is simplified and enhances the ability to accurately measure across a wide range of concentrations, improving versatility and usability of the spectrometry data.

[0013] Another one of those preferred further developments of the disclosed method comprise that the creation of the single feature value is done by the software by extracting a single response value by creating multiple responses using a range of integration time scales and then combining the multiple responses into a single response weighted by the Signal-to-Noise Ratio (SNR) that optimally reflects the relative strength of extracted responses at different scales for different concentrated samples. This embodiment optimizes data accuracy and reliability by combining multiple measurements into a single, robust dataset that reflects the relative strengths of responses at different scales, enhancing the precision in detecting various sample concentrations.

[0014] Another one of those preferred further developments of the disclosed method comprise that the statistical analysis of underlying photon counting distribution is done by employing a rigorous statistical framework to estimate the underlying background photon count distribution via iteratively detecting and removing count bins that are likely from fluorescence molecules. This accurate estimate of background noise level directly leads to accurate detection of molecules by providing a method for accurately separating signal from noise, crucial for low-concentration sample detection, thereby ensuring high fidelity in measurement and analysis results.

[0015] Another one of those preferred further developments of the disclosed method comprise that the software utilizes machine learning algorithms to adaptively adjust the statistical analysis parameters based on data collected from previous scans, thereby improving the predictive accuracy of the signal extraction for similar future samples. This enhances the predictive accuracy and efficiency of the instrument by learning from past data, thereby improving future analyses and reducing the likelihood of errors in signal extraction.

[0016] Another one of those preferred further developments of the disclosed method comprise that the signal extraction includes a normalization step where the signal data from various wells of the well plate are normalized against a control sample to correct for any inter-well variability. This addresses possible variability between different sample wells, ensuring consistent and reliable measurements across an entire plate, crucial for high-throughput and multisample experiments.

[0017] Another one of those preferred further developments of the disclosed method comprise that additional steps include calibrating the high-sensitivity optical spectrometry instrument before scanning the at least one sample to ensure consistent performance across different uses and time. That guarantees reliability and accuracy of the instrument by standardizing performance before use, which is essential for maintaining consistent data quality in longitudinal studies or varied experimental conditions.

[0018] A further component of the claimed invention is a system to improve the quantitative accuracy and sensitivity of spectral channels of a high-sensitivity optical spectrometry instrument, comprising of a computer and the high- sensitivity optical spectrometry instrument which is connected to the computer, wherein the system is configured to perform the previously described method steps. This system integrates both hardware and software components to form a complete system capable of executing the described methods, ensuring seamless operation and user experience from data collection to analysis.

[0019] Another one of those preferred further developments of the disclosed system comprise that the computer is integrated into the instrument and acts as a control unit for the instrument. The integration of the computer into the instrument, serving as a control unit, simplifies the system architecture and enhances control over the instrument's functions, leading to improved handling, reduced complexity, and potentially lower maintenance requirements. [0020] Another one of those preferred further developments of the disclosed system comprise that the system includes a user interface on the computer that displays real-time analytics of the signal extraction process, including visualization of the raw photon count trace and the steps of statistical analysis being applied. Via the inclusion of a user interface that displays real-time analytics of the signal extraction process the user interaction with the system can be enhanced by providing immediate visual feedback and detailed analysis, facilitating better understanding and quicker decision-making during experiments.

[0021] Another one of those preferred further developments of the disclosed system comprise that the system is capable of automatically adjusting the laser intensity based on the feedback from the software regarding the detected fluorescence intensity, thereby optimizing the excitation energy for different samples and preventing photobleaching. This automatic adjustment of the laser intensity based on software feedback regarding detected fluorescence intensity not only prevents photobleaching but also optimizes excitation energy for different samples, thus preserving sample integrity and improving the quality of fluorescence data collected.

BRIEF DESCRIPTION OF SEVERAL VIEWS OF THE DRAWINGS

[0022] Figure 1 : A schematic of the optical path of a confocal detection instrument including its components.

[0023] Figure 2: The steps of the invented method for a preferred embodiment.

DETAILED DESCRIPTION OF THE INVENTION

[0024] The invention will be explained in more detail by presenting one preferred exemplary embodiment. [0025] The invented method can be applied for different high-sensitivity optical spectrometry instruments, in particular single molecule detection (SMD) instruments like the newest SMCxPRO® and SMC® assay product line mentioned in the unpublished US Provisional application 63/605,656.

[0026] The invented method can be applied for different high-sensitivity optical instruments, but preferably for the mentioned SMCxPRO® instrument. An example of the optical path of a confocal detection instrument is shown in Figure 1.

[0027] The invention is now further explained by one preferred embodiment, as described in Figure 2. The invented method in the preferred embodiment consists of the following method steps:

[0028] Start: The starting point of the algorithm.

[0029] Scan Well Plate: The high-sensitivity optical spectrometry instrument scans the well plate.

[0030] Identify Abnormal Photon Count Bins: Using statistical analysis, the algorithm identifies photon count bins that deviate from the expected distribution (outlying bins).

[0031] Remove Scan Defects: The identified abnormal photon count bins are removed from the scan data.

[0032] Extract Signal Responses in Multiscale: The algorithm processes the cleaned data to extract signal responses at multiple integration times.

[0033] Maximize Fluorescence Response SNR: Enhance the signal-to-noise ratio of the fluorescence response.

[0034] Extend Signal Response Dynamically: Adaptively extend the signal response at the top of the standard curve to widen the dynamic range.

[0035] Provide Real-Time Feedback: As the well scan continues, the algorithm optimizes computations to provide real-time feedback on the signal response.

[0036] End: The algorithm completes its execution. [0037] Regarding the required hardware and software environment the invention in this preferred embodiment is implemented as part of Belysa analysis software (Belysa-DEV build #3344) as well as part of the embedded software (control server build #77.0.258) in the instrument computer. The Belysa III has an option to switch between current and this invention algorithm. To switch on, go to Tools -^Settings -^Analysis and check the checkbox "Use advanced signal extraction for XPD files". Once it is checked, next time when an xpd file is loaded, the invention algorithm will be automatically invoked. To switch back to current algorithm, simply check off the same checkbox and reload the xpd file which will be analyzed by the current algorithm.

[0038] The working principle itself is as following: As mentioned, at the ultra-low concentrations, when a fluorescent-labeled molecule is excited by laser light in the femtoliter-level confocal volume, the fluorescence photon burst at the given bintime has to be accurately detected for reaching high sensitivity. The most saliant feature by this invention is to create a single feature value called RE (response) to represent and quantify the raw photon counting trace from well scan of samples with full dynamic range from ultra-low to very high concentrations. The most technically challenging problem this invention solved is to extract a single response value by creating multiple responses using a range of integration time scales (bintimes) and then combining into single response weighted by signal-to-noise ratio (SNR) that optimally reflects the relative strength of extracted responses at different scales adapted automatically for different concentrated samples. Another important algorithmic problem in Erenna and Clarity is the over-estimating of the background noise level by heuristically calculating only the lowest 9% bin counts that can include non-negligible count bins from fluorescence molecules at ultra-low or high concentration samples. This invention solves this problem by employing rigorous statistical framework to estimate the underlying background photon count distribution by iteratively detecting and removing count bins that are likely from fluorescence molecules. The accurate estimate of background noise level directly leads to accurate detection of molecules with a given statistical significance level. [0039] The algorithm optimization approach and its potential in achieving better assay sensitivity and precision are validated using SMCxPRO commercial assay kits as well as those that are currently been developed for both RED channel (647nm) and NIR channel (730nm). The most significant result established by this invention is the dynamic range of responses (ratio between the responses at the top and the background) outperforms the current algorithm by 1 .5 to 5 times depending on assay types. Also, the slope of the standard curve generated using this invention increases by more than 10% compared to the current algorithm. These two advantages are directly related to assay linear range and assay sensitivity with lower LOD and LLOQ and higher IILOQ.

Claims

1. A Method to improve the quantitative accuracy and sensitivity of spectral channels of a high-sensitivity optical spectrometry instrument via a computer, the following steps comprising:

• Using the high-sensitivity optical spectrometry instrument to scan at least one sample in a well plate and forwarding the resulting signal data to the computer;

• Removing scan defects in the signal data via a software hosted by the computer by identifying photon count bins in the signal data using statistical analysis of underlying photon counting distribution with outlying bins;

• Extracting signal responses in multiscale via the software to increase the detection sensitivity by maximizing the response SNR, and to widen the response dynamic range by adaptively extending the signal response at the top of standard curve; and

• Providing real-time feedback of the extracted signal response applying it via the software to the high-sensitivity optical spectrometry instrument while the raw photon count trace is being collected during the scan of the well plate, thereby improving the quantitative accuracy and sensitivity of its spectral channels.

2. The Method according to claim 1 , wherein the at least one sample in the well plate reflects fluorescence light when being excited by laser light and for the high-sensitivity optical spectrometry instrument the fluorescence detection sensitivity is increased and the fluorescence response dynamic range is widened.

3. The Method according to claim 2, wherein the at least one sample in the well plate is a fluorescent-labeled molecule concentrated in a femtoliter-level confocal volume at ultralow concentrations.

4. The Method according to claims 1 to 3, wherein for identifying photon count bins in the signal data a single feature value (RE) is created by the software to represent and quantify the raw photon counting trace of the scan of the at least one sample with full dynamic range from ultra-low to very high concentrations.

5. The Method according to claim 4, wherein the creation of the single feature value is done by the software by extracting a single response value by creating multiple responses using a range of integration time scales and then combining the multiple responses into a single response weighted by the Signal- to-Noise Ratio (SNR) that optimally reflects the relative strength of extracted responses at different scales for different concentrated samples.

6. The Method according to claims 1 to 5, wherein the statistical analysis of underlying photon counting distribution is done by employing a rigorous statistical framework to estimate the underlying background photon count distribution via iteratively detecting and removing count bins that are likely from fluorescence molecules.

7. The Method according to claims 1 to 6, wherein the software utilizes machine learning algorithms to adaptively adjust the statistical analysis parameters based on data collected from previous scans, thereby improving the predictive accuracy of the signal extraction for similar future samples.

8. The Method according to claims 1 to 7, wherein the signal extraction includes a normalization step where the signal data from various wells of the well plate are normalized against a control sample to correct for any inter-well variability.

9. The Method according to claims 1 to 8, wherein additional steps include calibrating the high-sensitivity optical spectrometry instrument before scanning the at least one sample to ensure consistent performance across different uses and time.

10. A System to improve the quantitative accuracy and sensitivity of spectral channels of a high-sensitivity optical spectrometry instrument, comprising of a computer and the high-sensitivity optical spectrometry instrument which is connected to the computer, wherein the system is configured to perform the method steps of claims 1 to 9.

11 . The system according to claim 10, wherein the computer is integrated into the instrument and acts as a control unit for the instrument.

12. The System according to claims 10 to 11 , wherein the system includes a user interface on the computer that displays real-time analytics of the signal extraction process, including visualization of the raw photon count trace and the steps of statistical analysis being applied.

13. The System according to claims 10 to 12, wherein the system is capable of automatically adjusting the laser intensity based on the feedback from the software regarding the detected fluorescence intensity, thereby optimizing the excitation energy for different samples and preventing photobleaching.