Dependence of base detection on flow cell tilt

Abstract

During sequencing while synthesizing, the tilt of the flow cell and the flatness variation of the flow cell introduce out-of-focus. Techniques related to dependence of base detection on flow cell tilt are used, reducing the impact of this defocus and improving base detection quality. For example, flow cell surface height is measured throughout the flow cell. During the sequencing, the focus height of an imager having a sensor for sequencing is optionally adaptively set one or more times. Each image captured by the sensor is partitioned, for example, based on a difference between a focus height and a measured flow cell surface height across a region of the sensor. A filter associated, for example, with defocus correction is selected based at least in part on a difference between the focus height and a measured flow cell surface height at a particular region of the image being corrected for defocus.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 350,776, filed on June 9, 2022, the entire disclosure of which is hereby incorporated by reference herein in its entirety. Technical Field

[0003] The technology disclosed in the present invention relates to sequencing by synthesis for determining a genetic sequence in parallel by performing parallel base calling on many nucleotides of the genetic sequence. Base calling is enhanced by focusing / tilting of a flow cell that relies on retaining portions of the genetic material. Base calling enhancement is relative to image processing.

[0004] Reference Merge

[0005] The following documents are incorporated by reference as if fully set forth herein for all purposes:

[0006] U.S. non-provisional patent No. US8,422,031B2, filed on April 16, 2013, entitled “Focusing Methods and Optical Systems and Assemblies Using the Same”

[0007] U.S. Nonprovisional Application No. 15 / 936,365, filed on March 26, 2018, entitled “DETECTION APPARATUS. HAVING AMICROFLUOROMETER, A FLUIDIC SYSTEM, AND A FLOW CELL LATCH CLAMP MODULE”;

[0008] U.S. non-provisional patent application No. 16 / 567,224, entitled “FLOW CELLS AND METHODS RELATED TO SAME,” filed on September 11, 2019;

[0009] U.S. non-provisional patent application No. 16 / 439,635, entitled “DEVICE FOR LUMINESCENT IMAGING,” filed on June 12, 2019;

[0010] U.S. non-provisional patent application No. 15 / 594,413, entitled “INTEGRATED OPTOELECTRONIC READ HEAD AND FLUIDIC CARTRIDGE USEFUL FOR NUCLEIC ACID SEQUENCING,” filed on May 12, 2017;

[0011] U.S. non-provisional patent application No. 16 / 351,193, entitled “ILLUMINATION FOR FLUORESCENCE IMAGING USING OBJECTIVE LENS,” filed on March 12, 2019;

[0012] U.S. non-provisional patent application No. 12 / 638,770, entitled “DYNAMIC AUTOFOCUS METHOD AND SYSTEM FORASSAY IMAGER,” filed on December 15, 2009;

[0013] U.S. non-provisional patent application No. 13 / 783,43, entitled “KINETIC EXCLUSION AMPLIFICATION OF NUCLEICACID LIBRARIES,” filed on March 1, 2013;

[0014] U.S. Non-Provisional Patent Application No. 13 / 006,206, entitled “DATA PROCESSING SYSTEM AND METHODS,” filed on January 13, 2011;

[0015] U.S. non-provisional patent application No. 14 / 530,299, entitled “IMAGE ANALYSIS USEFUL FOR PATTERNEDOBJECTS,” filed on October 31, 2014;

[0016] U.S. non-provisional patent application No. 15 / 153,953, entitled “METHODS AND SYSTEMS FOR ANALYZING IMAGEDATA,” filed on December 3, 2014;

[0017] U.S. non-provisional patent application Ser. No. 14 / 20,570, filed Sep. 6, 2013, entitled “CENTROID MARKERS FOR IMAGE ANALYSIS OF HIGHDENSITY CLUSTERS IN COMPLEX POLYNUCLEOTIDE SEQUENCING”;

[0018] U.S. non-provisional patent application No. 14 / 530,299, entitled “IMAGE ANALYSIS USEFUL FOR PATTERNEDOBJECTS,” filed on October 31, 2014;

[0019] U.S. non-provisional patent application No. 12 / 565,341, entitled “METHOD AND SYSTEM FOR DETERMINING THEACCURACY OF DNA BASE IDENTIFICATIONS,” filed on September 23, 2009;

[0020] U.S. non-provisional patent application No. 12 / 295,337, entitled “SYSTEMS AND DEVICES FOR SEQUENCE BYSYNTHESIS ANALYSIS,” filed on March 30, 2007;

[0021] U.S. non-provisional patent application No. 12 / 20,739, entitled “IMAGE DATA EFFICIENT GENETIC SEQUENCING METHOD AND SYSTEM,” filed on January 28, 2008;

[0022] U.S. nonprovisional patent application Ser. No. 13 / 833,619, entitled “BIOSENSORS FOR BIOLOGICAL OR CHEMICAL ANALYSIS AND SYSTEMS AND METHODS FOR SAME,” filed on Mar. 15, 2013 (Attorney Docket No. IP-0626-US);

[0023] U.S. non-provisional patent application No. 15 / 175,489, entitled “BIOSENSORS FOR BIOLOGICAL OR CHEMICAL ANALYSIS AND METHODS OF MANUFACTURING THE SAME,” filed on June 7, 2016 (Attorney Docket No. IP-0689-US);

[0024] U.S. nonprovisional patent application Ser. No. 13 / 882,088, filed on April 26, 2013, entitled “MICRODEVICES AND BIOSENSOR CARTRIDGES FOR BIOLOGICAL OR CHEMICAL ANALYSIS AND SYSTEMS AND METHODS FOR THE SAME” (Attorney Docket No. IP-0462-US);

[0025] U.S. non-provisional patent application No. 13 / 624,200, entitled “METHODS AND COMPOSITIONS FOR NUCLEIC ACID SEQUENCE,” filed on September 21, 2012 (Attorney Docket No. IP-0538-US);

[0026] U.S. non-provisional patent application No. 17 / 308,35, filed on May 4, 2021, entitled “EQUALIZATION-BASED IMAGE PROCESSING AND SPATIAL CROSSTALK ATTENUATOR” (Attorney Docket No. ILLM 1032-2 / IP-1991-US).

[0027] U.S. Provisional Patent Application No. 62 / 821,602, entitled “Training Data Generation for Artificial Intelligence-Based Sequencing,” filed on March 21, 2019 (Attorney Docket No. ILLM 1008-1 / IP-1693-PRV);

[0028] U.S. Provisional Patent Application No. 62 / 821,618, entitled “Artificial Intelligence-Based Generation of Sequencing Metadata,” filed on March 21, 2019 (Attorney Docket No. ILLM 1008-3 / IP-1741-PRV);

[0029] U.S. Provisional Patent Application No. 62 / 821,681, entitled “Artificial Intelligence-Based Base Calling,” filed on March 21, 2019 (Attorney Docket No. ILLM 1008-4 / IP-1744-PRV);

[0030] U.S. Provisional Patent Application No. 62 / 821,724, entitled “Artificial Intelligence-Based QualityScoring,” filed on March 21, 2019 (Attorney Docket No. ILLM 1008-7 / IP-1747-PRV);

[0031] U.S. Provisional Patent Application No. 62 / 821,766, entitled “Artificial Intelligence-Based Sequencing,” filed on March 21, 2019 (Attorney Docket No. ILLM 1008-9 / IP-1752-PRV);

[0032] Netherlands Patent Application No. 2023310, entitled “Training Data Generation for Artificial Intelligence-Based Sequencing,” filed on June 14, 2019 (Attorney Docket No. ILLM1008-11 / IP-1693-NL);

[0033] Netherlands Patent Application No. 2023311, entitled “Artificial Intelligence-Based Generation of Sequencing Metadata,” filed on June 14, 2019 (attorney docket number ILLM 1008-12 / IP-1741-NL);

[0034] Netherlands Patent Application No. 2023312, entitled “Artificial Intelligence-Based Base Calling,” filed on June 14, 2019 (Attorney Docket No. ILLM 1008-13 / IP-1744-NL);

[0035] Netherlands Patent Application No. 2023314, entitled “Artificial Intelligence-Based QualityScoring,” filed on June 14, 2019 (Attorney Docket No. ILLM 1008-14 / IP-1747-NL); and

[0036] Netherlands Patent Application No. 2023316, entitled “Artificial Intelligence-Based Sequencing,” filed on June 14, 2019 (Attorney Docket No. ILLM 1008-15 / IP-1752-NL).

[0037] U.S. nonprovisional patent application No. 16 / 825,987, entitled “Training Data Generation for Artificial Intelligence-Based Sequencing,” filed on March 20, 2020 (Attorney Docket No. ILLM 1008-16 / IP-1693-US);

[0038] U.S. nonprovisional patent application No. 16 / 825,991, entitled “Training Data Generation for Artificial Intelligence-Based Sequencing,” filed on March 20, 2020 (Attorney Docket No. ILLM 1008-17 / IP-1741-US);

[0039] U.S. nonprovisional patent application No. 16 / 826,126, entitled “Artificial Intelligence-Based Base Calling,” filed on March 20, 2020 (Attorney Docket No. ILLM 1008-18 / IP-1744-US);

[0040] U.S. non-provisional patent application No. 16 / 826,134, entitled “Artificial Intelligence-Based QualityScoring,” filed on March 20, 2020 (Attorney Docket No. ILLM 1008-19 / IP-1747-US);

[0041] U.S. nonprovisional patent application No. 16 / 826,168, entitled “Artificial Intelligence-Based Sequencing,” filed on March 21, 2020 (Attorney Docket No. ILLM 1008-20 / IP-1752-PRV);

[0042] U.S. non-provisional patent application No. 17 / 511,483, entitled “Intensity Extraction with Interpolation and Adaptation for Base Calling,” filed on October 26, 2021 (Attorney Docket No. ILLM 1053-1 / IP-2214-US);

[0043] U.S. non-provisional patent application No. 17 / 687,586, entitled “Artificial Intelligence-Based Base Callerwith Contextual Awareness,” filed on March 4, 2022 (Attorney Docket No. ILLM1033-2 / IP-2007-US);

[0044] U.S. nonprovisional patent application No. 16 / 826,126, entitled “Artificial Intelligence-Based Base Calling,” filed on March 30, 2020 (Attorney Docket No. ILLM 1008-18 / IP-1744-US);

[0045] U.S. Non-Provisional Patent No. 10,830,700B2, entitled “Solid Inspection Apparatus and Method of Use,” filed on March 1, 2019;

[0046] U.S. nonprovisional patent application No. 17 / 179,395, entitled “Data Compression for Artificial Intelligence-Based Base Calling,” filed on February 18, 2021 (attorney docket No. ILLM 1029-2 / IP-1964-US);

[0047] U.S. non-provisional patent application No. 17 / 180,480, entitled “Split Architecture for Artificial Intelligence-Based Base Caller,” filed on February 19, 2021 (Attorney Docket No. ILLM 1030-2 / IP-1982-US);

[0048] U.S. non-provisional patent application No. 17 / 180,513, entitled “Bus Network for Artificial Intelligence-Based Base Caller,” filed on February 19, 2021 (attorney docket No. ILLM 1031-2 / IP-1965-US);

[0049] U.S. Provisional Patent Application No. 62 / 849,091, entitled “Systems and Devices for Characterization and Performance Analysis of Pixel-Based Sequencing,” filed on May 16, 2019 (Attorney Docket No. ILLM 1011-1 / IP-1750-PRV);

[0050] U.S. Provisional Patent Application No. 62 / 849,132, entitled “Base Calling Using Convolutions,” filed on May 16, 2019 (Attorney Docket No. ILLM 1011-2 / IP-1750-PR2);

[0051] U.S. Provisional Patent Application No. 62 / 849,133, entitled “Base Calling Using Compact Convolutions,” filed on May 16, 2019 (Attorney Docket No. ILLM 1011-3 / IP-1750-PR3);

[0052] U.S. Provisional Patent Application No. 62 / 979,384, entitled “Artificial Intelligence-Based Base Calling of Index Sequences,” filed on February 20, 2020 (Attorney Docket No. ILLM 1015-1 / IP-1857-PRV);

[0053] U.S. Provisional Patent Application No. 62 / 979,414, entitled “Artificial Intelligence-Based Many-To-ManyBase Calling,” filed on February 20, 2020 (Attorney Docket No. ILLM 1016-1 / IP-1858-PRV);

[0054] U.S. Provisional Patent Application No. 62 / 979,385, entitled “Knowledge Distillation-Based Compression of Artificial Intelligence-Based Base Caller,” filed on February 20, 2020 (Attorney Docket No. ILLM 1017-1 / IP-1859-PRV);

[0055] U.S. Provisional Patent Application No. 62 / 979,412, filed on February 20, 2020, entitled “Multi-Cycle Cluster Based Real Time Analysis System” (Attorney Docket No. ILLM 1020-1 / IP-1866-PRV);

[0056] U.S. Provisional Patent Application No. 62 / 979,411, entitled “Data Compression for Artificial Intelligence-Based Base Calling,” filed on February 20, 2020 (Attorney Docket No. ILLM 1029-1 / IP-1964-PRV);

[0057] U.S. Provisional Patent Application No. 62 / 979,399, entitled “Squeezing Layer for Artificial Intelligence-Based Base Calling,” filed on February 20, 2020 (Attorney Docket No. ILLM 1030-1 / IP-1982-PRV);

[0058] U.S. Provisional Patent Application No. 63 / 228,954, entitled “Self-Learned Base Caller,” filed on August 3, 2021 (Attorney Docket No. ILLM 1021-1 / IP-1856-PRV);

[0059] U.S. Provisional Application No. 63 / 300,531, entitled “Dynamic Detilt Focus Tracking,” filed on January 18, 2022 (Attorney Docket No. IP-2205-PRV); and

[0060] U.S. Provisional Patent Application No. 63 / 072,032, filed on August 28, 2020, entitled “Detecting and Filtering Clusters Based on Artificial Intelligence-Predicted Base Calls” (Attorney Docket No. ILLM 1018-1 / IP-1860-PRV).

[0061] Overall integration

[0062] The following documents filed with this provisional patent application are fully incorporated into this provisional patent application and should be considered a part of this provisional patent application:

[0063] Appendix, page 37. Background Art

[0064] The subject matter discussed in this section should not be considered to be prior art simply because it is mentioned in this section. Similarly, problems mentioned in this section or associated with the subject matter provided as background technology should not be considered to have been previously recognized in the prior art. The subject matter in this section merely represents different approaches, which themselves may also correspond to specific implementations of the technology protected by the claims.

[0065] Various schemes in biological or chemical research involve carrying out a large number of controlled reactions on a local support surface or in a predefined reaction chamber. The desired reaction can then be observed or detected, and subsequent analysis can help identify or reveal the characteristics of the chemical substances involved in the reaction. For example, in some multiple determinations, an unknown analyte with an identifiable label (e.g., a fluorescent label) can be exposed to thousands of known probes under controlled conditions. Each known probe can be placed in the corresponding well of a microplate. Any chemical reaction occurring between the known probe and the unknown analyte in the observation well can help identify or reveal the characteristics of the analyte. Other examples of such schemes include known DNA sequencing processes, such as sequencing while synthesizing or cycle array sequencing. In cycle array sequencing, a dense array of DNA features (e.g., template nucleic acids) is sequenced by repeated cycles of enzymatic operations. After each cycle, an image can be captured and then analyzed with other images to determine the sequence of the DNA feature.

[0066] As a first specific example, a known DNA sequencing system uses a pyrophosphate sequencing process and includes a chip with a fused fiber panel having millions of holes. A single capture bead with sstDNA cloned and amplified from a genome of interest is placed in each hole. After the capture beads are placed in the holes, the nucleotides are added to the holes in sequence by flowing a solution containing specific nucleotides along the panel. The environment in the hole is such that if the nucleotide flowing through a specific hole is complementary to the DNA chain on the corresponding capture bead, the nucleotide is added to the DNA chain. The colony of DNA chains is called a cluster. The binding of nucleotides to the cluster triggers a process that ultimately generates a chemiluminescent light signal. The system includes a CCD camera positioned directly adjacent to the panel and configured to detect light signals from the DNA clusters in the hole. Subsequent analysis of images captured throughout the pyrophosphate sequencing process can determine the sequence of the genome of interest.

[0067] However, the above-mentioned pyrophosphate sequencing system may have certain limitations, among other systems. For example, the fiber optic panel is acid-etched to form millions of small holes. Although these holes can be spaced approximately from each other, it is difficult to know the exact position of a hole relative to other adjacent holes. When the CCD camera is positioned directly adjacent to the panel, the holes are unevenly distributed along the pixels of the CCD camera, and therefore, the holes are not aligned with the pixels in a known manner. Spatial crosstalk is inter-hole crosstalk between adjacent holes, and makes it difficult to distinguish the real light signal from the hole of interest from other unwanted light signals in subsequent analysis. Moreover, fluorescence emission is substantially isotropic. As the density of analytes increases, it becomes increasingly challenging to manage or resolve unwanted light emissions (e.g., crosstalk) from adjacent analytes. Therefore, the data recorded during the sequencing cycle must be carefully analyzed.

[0068] As a second specific example related to sequencing by synthesis, a genetic sequence associated with a sample of DNA, RNA, protein, and / or other genetic material having a base sequence is determined. The genetic sequence is used for many purposes, including diagnosis and treatment of disease.

[0069] As a third specific example related to sequencing by synthesis, the tilt and / or non-planarity of the circulation cell introduces focusing changes across the circulation cell. Focusing and / or tilt adjustment techniques enable some sequencing imaging to be performed in the following manner: a best fit plane is established for the entire sample so that the entire sample remains within the DoF of the optical imaging system. However, the available DoF is reduced, such as due to an increase in the numerical aperture (NA). Therefore, global and / or local changes in tilt and / or height cause portions of the sample to deviate outside the DoF, resulting in out-of-focus image portions and therefore a decrease in data quality and / or loss of data. As a result, base calling accuracy is reduced.

[0070] Sequencing by synthesis is a parallel technique for determining genetic sequence, and operates in parallel on a large number of oligonucleotides (sometimes referred to as oligomers) of a sample at a time, one base position at a time for each of the oligomers. Some specific implementations of sequencing by synthesis operate by cloning oligomers on a substrate (such as a slide and / or a flow cell), which are, for example, arranged in multiple channels and imaged as corresponding blocks in each channel. In some specific implementations, the cloning is arranged to, such as preferentially, clone each of a plurality of starting oligomers into a corresponding cluster of oligomers, such as in a corresponding nanopore of a patterned flow cell.

[0071] Sequencing by synthesis is performed in a series of sequencing cycles (sometimes referred to as cycles). In each sequencing cycle, there are chemical reactions, image capture, and base detection actions. The result of these actions is a base determined in parallel for each oligomer in the oligomer (e.g., one of four amino acids: adenine (A), guanine (G), thymine (T), and cytosine (C)). The chemical reaction action is designed to add a dye-labeled complementary nucleotide (sometimes referred to as a fluorophore) to each clone (e.g., oligomer) in each cluster in each cycle. The image capture action typically focuses and aligns an imager (e.g., a camera) relative to a block of the channel of the circulation pool, irradiates the block (e.g., using one or more lasers) to stimulate the fluorophore to fluoresce, and captures multiple images of the fluorescence (e.g., one to four images, each image corresponding to the block and each having a different wavelength). The base detection action results in the identification of the base determined in parallel for each oligomer (e.g., one of A, G, T, and C). In some implementations, the image capture action corresponds to a discrete point-and-shoot operation, e.g., the imager and the flow cell are moved relative to each other, and then the image capture action is performed for the block. In some implementations, the image capture action corresponds to a continuous scanning operation, e.g., the imager and the flow cell are continuously moved relative to each other and image capture is performed during the movement. In various continuous scanning implementations, the block corresponds to any continuous area of ​​the sample.

[0072] Some specific implementations of sequencing by synthesis use fluorescently labeled nucleotides (such as fluorescently labeled deoxyribonucleoside triphosphates (dNTPs)) as fluorophores. During each sequencing cycle, a single fluorophore is added to each oligomer in parallel. An excitation source such as a laser stimulates many fluorophores to fluoresce in parallel, and the fluorescing fluorophores are imaged in parallel via one or more imaging operations. When the imaging of the fluorophores added in the sequencing cycle is completed, the fluorophores added in the sequencing cycle are removed and / or inactivated, and the sequencing proceeds to the next sequencing cycle. During the next sequencing cycle, the next single fluorophore is added to each oligomer in parallel, the excitation source stimulates many fluorophores added in the next sequencing cycle to fluoresce in parallel, and the fluorescing fluorophores are imaged in parallel via one or more imaging operations. The sequencing cycle is repeated as needed based on how many bases there are in the oligomer and / or other termination conditions.

[0073] The accuracy of base calling is critical for high-throughput DNA sequencing and downstream analysis such as read mapping and genome assembly. In various scenarios, the flow cell holder or the flow cell or its elements (e.g., the glass / substrate of the flow cell and / or the patterned nanopores of the flow cell) cause tilt and / or height that causes non-optimal focusing. In some embodiments, spatial crosstalk between adjacent clusters and / or focusing changes such as due to changes in tilt and / or flatness of the flow cell are the source of most sequencing errors. Therefore, by addressing and / or correcting spatial crosstalk in the cluster intensity data, and / or by addressing and / or correcting focusing changes such as due to tilt and / or non-planarity relative to the image, there is an opportunity to reduce DNA sequencing errors and improve base calling accuracy. Summary of the invention

[0074] The technology disclosed in the present invention relates to sequencing by synthesis to determine a genetic sequence by performing parallel base calling on many nucleotides of the genetic sequence. A circulation cell retains a portion of the genetic material. During the sequencing, the tilt of the circulation cell and the change in the flatness of the circulation cell introduce defocus. Using techniques related to the dependence of base calling on the tilt of the circulation cell, the effect of the defocus is reduced and the quality of base calling is improved. For example, the circulation cell surface height is measured throughout the circulation cell. During the sequencing, the focus height of an imager having a sensor for sequencing is optionally set adaptively once or multiple times. Each image captured by the sensor is partitioned, for example, based on the difference between the focus height and the measured circulation cell surface height across the area of ​​the sensor. A filter, for example, related to defocus correction is selected based at least in part on the difference between the focus height and the measured circulation cell surface height at a specific area of ​​the image that is being corrected for defocus. Defocus correction is performed using the selected filter, and base calling is performed using the resulting image information. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] This patent or patent application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee. Color drawing(s) are also available in PAIR via the Supplemental Content tab.

[0076] In the drawings, similar reference symbols generally refer to similar parts throughout the different views. In addition, the drawings are not necessarily drawn to scale, but rather focus on illustrating the principles of the technology disclosed in the present invention as a whole. In the following description, various specific implementations of the technology disclosed in the present invention are described with reference to the following drawings, in which:

[0077] Figure 1AA An example illustrating the dependence of base calling on flow cell tilt is shown.

[0078] Fig. 1AB Example Figure 1AA Depicted are examples of base calling's dependence on flow cell tilt related operations.

[0079] Figure 1AC Elements for imaging a flow cell are generally illustrated, including selected details related to flow cell tilt.

[0080] Figure 1AD Selected details related to the tilting of the flow cell are illustrated.

[0081] Figure 1AE Selected details related to the non-planarity of the flow cell are illustrated.

[0082] Figure 1A One specific implementation of generating a look-up table (LUT) / equalizer filter by training an equalizer is shown.

[0083] Figure 1B Describes the use Figure 1A A specific implementation of a LUT / equalizer filter to attenuate spatial crosstalk from sensor pixels and base call clusters using the crosstalk-corrected sensor pixels.

[0084] Figure 2 An example of a sequencing image containing a center / point source of at least five clusters / wells on the flow cell is visualized.

[0085] Figure 3 Visualize the Figure 2 An example of extracting a pixel patch (yellow) from the sequencing image such that the center of target cluster 1 (blue) is contained in the center pixel of that pixel patch.

[0086] Figure 4 An example of cluster-to-pixel signal is visualized.

[0087] Figure 5 An example of cluster-to-pixel signal overlap is visualized.

[0088] Figure 6 An example of a cluster signal pattern is visualized.

[0089] Figure 7 Visualizes the decay from Figure 3 An example of a sub-pixel LUT grid showing spatial crosstalk of pixel patches.

[0090] Figure 8 The sub-pixel position of the cluster / hole center within the pixel is shown. Figure 1B Select a LUT / Equalizer filter from the LUT library in the menu.

[0091] Fig. 9One implementation is illustrated in which the center of target cluster 1 (blue) is not substantially concentric with the center of the pixel.

[0092] Fig.10 One specific implementation of interpolating between a selected set of LUTs and generating corresponding LUT weights is depicted.

[0093] Fig.11 A weight core generator is shown which generates a weight core using the calculated weights of the LUTs 12, 7, 8 and 13.

[0094] Fig.12 An element-by-element multiplier is shown multiplying the interpolated pixel coefficients of the weight kernel element-by-element with the intensity values ​​of the pixels in the pixel patch, and then summing the intermediate products of the multiplications to produce an output.

[0095] Fig.13A , Fig. 13B , Fig. 13C , Fig.13D , Fig.13E and Fig.13F Examples of coefficients of LUTs 12, 7, 8, and 13 are shown.

[0096] Fig.14A An example of a weight kernel is depicted.

[0097] Fig. 14B and Fig. 14C One example of weight core generation logic used by the weight core generator to generate weight cores from the calculated weights of the LUTs 12 , 7 , 8 , and 13 is illustrated.

[0098] Fig.15A and Fig. 15B It is shown how the interpolated pixel coefficients of the weight kernel maximize the signal-to-noise ratio and recover the underlying signal of the target cluster 1 from the signals corrupted by the crosstalk from clusters 2, 3, 4, and 5.

[0099] Fig.16 One specific implementation of base-by-base Gaussian fits is shown, which contain at their centers the base-by-base intensity targets used as ground truth for error calculations during training.

[0100] Fig.17A is a block diagram of an example computer system.

[0101] Fig. 17B Training and production elements that implement various aspects of base calling that relies on flow cell tilt are illustrated.

[0102] Fig.18 One specific implementation of an adaptive equalization technique that can be used to train an equalizer is shown.

[0103] Fig.19A , Fig.19B , Fig.19C and Fig.19D Various performance metrics of the disclosed techniques are illustrated.

[0104] Fig. 20A A benchmark example is illustrated.

[0105] Fig. 20B Example benchmarks in various focus contexts are illustrated.

[0106] Fig. 20C Example cross-correlation relations for discrete functions are illustrated.

[0107] Fig.20D An example scoring relationship is illustrated. DETAILED DESCRIPTION

[0108] The following discussion is presented to enable any person skilled in the art to make and use the technology disclosed herein, and is provided in the context of one or more specific applications and their associated requirements. Various modifications to the specific implementations disclosed herein will be apparent to those skilled in the art, and the general principles defined herein may be applied to other specific implementations and applications without departing from the spirit and scope of the technology disclosed herein. Therefore, the technology disclosed herein is not intended to be limited to the specific implementations disclosed, but is intended to conform to the widest scope consistent with the principles and features disclosed herein.

[0109] The following is a detailed description with reference to the accompanying drawings. The description of the example implementation is intended to illustrate the technology disclosed in the present invention, rather than to limit its scope, which is defined by the claims. A person of ordinary skill in the art will recognize various equivalent variations to the following description.

[0110] Examples of selected terms

[0111] Depending on the specific implementation, the elements of the equalizer (e.g., spatial equalizer) (such as elements capable of convolution, capable of performing convolution and / or capable of managing lookup table information) and the layers, loss functions and / or objective functions correspond differently to one or more hardware elements, one or more software elements and / or various combinations of hardware elements and software elements. For the first example, the convolution element (such as an N×M×D convolution element) is implemented as a hardware logic circuit included in an application-specific integrated circuit (ASIC). For the second example, multiple convolution layers are implemented in the TensorFlow machine learning framework on a collection of Internet-connected servers. For the third example, the first one or more parts of the spatial equalizer (such as one or more convolution layers) are respectively implemented in the hardware logic circuit according to the first example, and the second one or more parts of the spatial equalizer (such as one or more convolution layers) are implemented on a collection of Internet-connected servers according to the second example. It is contemplated that various specific implementations using various combinations of hardware elements and software elements to provide corresponding price points and performance points.

[0112] The example implementation of real-time analysis (RTA) architecture (e.g., equalizer, such as spatial equalizer) includes various collections of software elements and / or hardware elements that perform operations together according to the RTA architecture. Various RTA implementations vary according to machine learning frameworks, programming languages, runtime systems, operating systems, and underlying hardware resources. The underlying hardware resources include one or more computer systems, such as any combination of a central processing unit (CPU), a graphics processing unit (GPU), a field programmable gate array (FPGA), a coarse-grained reconfigurable architecture (FIELD), an application-specific integrated circuit (ASIC), an application-specific instruction set processor (ASIP), and a digital signal processor (DSP), and a general computing system, for example, an element capable of executing programming instructions specified via a programming language. Various RTA implementations can store programming information (such as code and data) on a non-transient computer-readable medium, and can further execute code and reference data according to a program that implements the RTA architecture.

[0113] Examples of programming languages, codes and / or databases, and / or operating environments that can be used to implement the dependence of base calls on flow cell tilt (such as those associated with expression signal processing functions (e.g., equilibrium and / or expectation-maximization)) include Python, Numpy, R, Java, Javascript, C#, C++, Julia, Shell, Go, TypeScript, and Scala.

[0114] An example of image collection is to use an imager to simultaneously capture the light emitted by a plurality of fluorescently labeled nucleotides when the nucleotides fluoresce in response to excitation energy (such as laser excitation energy) as a collected image. The image has one or more dimensions, for example, a row of pixels or a two-dimensional array of pixels. These pixels are represented according to one or more values. For the first example, each pixel is represented by a single integer (such as an 8-bit integer) representing the intensity of the pixel (such as a grayscale). For the second example, each pixel is represented by a plurality of integers (such as three 24-bit integers), and each of these integers represents the intensity of the pixel according to a corresponding wavelength band (such as a corresponding color).

[0115] An example of being in focus is that the element being imaged (e.g., a block of a flow cell or a portion of a flow cell) is nominally coincident with the imager's plane of focus, such that the element is within the imager's depth of field (DoF). In focus corresponds to the element's nominal maximum sharpness or nominal maximum definition. An example of being out of focus is that the element is above the plane of focus, such that the element is above the DoF (e.g., the element is too close to the imager to be in focus). An example of being out of focus is that the element is below the plane of focus, such that the element is below the DoF (e.g., the element is too far from the imager to be in focus).

[0116] Dependence of base calling on flow cell tilt

[0117] In the present disclosure, training context and production context are described. In some specific implementations, laboratory instruments (sometimes referred to as biological sequencing instruments) are used in training contexts, and production instruments (sometimes also referred to as biological sequencing instruments) are used in production contexts. In some specific implementations, laboratory instruments and production instruments are used in training contexts. Training context and production context implement various RTA related processes (such as one or more equalizer functions related to the reliance on the tilt of the circulation cell for realizing base detection). In various specific implementations, all or any part of the RTA related processes of training contexts are implemented differently on any one or more laboratory instruments in laboratory instruments, any one or more production instruments in production instruments, and / or any one or more computer systems separated from laboratory instruments and production instruments. In various specific implementations, all or any part of the RTA related processes of production contexts are implemented differently on any one or more laboratory instruments in laboratory instruments, any one or more production instruments in production instruments, and / or any one or more computer systems separated from laboratory instruments and production instruments (such as one or more servers). In various specific implementations, all or any part of the RTA related processes of laboratory instruments are performed by one or more computer systems of laboratory instruments. Similarly, in various implementations, all or any portion of the RTA-related processing for a production instrument is performed by one or more computer systems of the production instrument. In various implementations, all or any portion of a laboratory instrument is primarily used for image collection, and the RTA-related processing for the associated training context is performed on one or more computer systems separate from the laboratory instrument primarily used for image collection.

[0118] The dependence of base calling on the tilt of the circulation cell enables enhanced sequencing by synthesis, which determines the base sequence in the genetic material with improved accuracy compared to ignoring the tilt of the circulation cell. In turn, the improved accuracy achieves improved performance and / or reduced cost. Sequencing by synthesis is performed in parallel one base at a time for each of the multiple oligomers attached to all or any portion of the circulation cell. The processing of each corresponding base of the multiple oligomers includes: imaging a block of the circulation cell and using base calling that depends on the tilt of the circulation cell to improve the accuracy of the base calling.

[0119] To review, sequencing by synthesis is performed in part by capturing and processing images (e.g., blocks of a circulation cell). Consider an imager that captures an image of a portion of a circulation cell, the image having multiple portions. In some scenarios, the focusing technique used during image capture brings one of the portions into sharp focus, but one or more other portions are not in sharp focus due to limited DoF and the portions being at different distances from the imager. In some cases, the portions are at different distances from the imager because the circulation cell is tilted relative to the imager optical plane and / or the circulation cell is not uniformly flat (e.g., not uniformly planar) and therefore has different heights. Tilt and / or lack of uniform flatness results in non-uniformity of focus between portions of a single image, between different images, and between images of different circulation cells. In some scenarios, the non-uniformity of focus results in a reduction in base calling accuracy.

[0120] Differently, the tilt is consistent with the scanning direction of the imager, orthogonal to the scanning direction, or oblique to any angle relative to the scanning direction. Differently, the tilt is substantially uniform across the circulation cell, substantially variable across the circulation cell, or variable between relatively uniform and relatively variable across the circulation cell. In various scenarios, the non-planarity of the circulation cell changes with the above-mentioned tilt changes. In some specific implementations, the tilt is regarded as a vector with a magnitude (e.g., how much tilt there is) and a direction (e.g., which direction the tilt is facing). In contrast, the height is a scalar with only a magnitude (e.g., how far away a point on the circulation cell surface is from the imager image plane). In some specific implementations, two height measurements at two corresponding points on the circulation cell surface can be used to determine the tilt as a vector with a magnitude and a direction.

[0121] To review, the tilt of the flow cell relative to the optical plane affects the focus due to the different distances between different positions of the flow cell and the optical plane. For the first example, the center portion is in focus and the edge portion is out of focus. For the second example, the center portion is in focus, the first edge portion is overfocused, and the second edge portion positioned orthogonally relative to the first edge portion is out of focus. Other examples are characterized by various partial focuses, overfocuses, and underfocuses. Flow cells are manufactured with different flatness tolerances. Therefore, in some scenarios, the above-mentioned focus variability is partly due to the fact that the flow cell is not completely flat. In addition, the flatness of the flow cell varies between various flow cells and within a single flow cell. Therefore, in some scenarios, the tilt varies from one flow cell to another and from one block of the flow cell to another block of the flow cell.

[0122] In some cases, reducing the flow cell flatness tolerance enables to reduce the cost of the flow cell. Reduced flow cell flatness increases maximum tilt and / or tilt variability. In some cases, increasing the production base detection flux requires increasing the cluster density on the flow cell. In some cases, the increased cluster density leads to a reduction in DoF (such as due to increased NA), thus increasing the influence of tilt and / or flow cell non-planarity. In some cases, the increased cluster density increases optical crosstalk, thus increasing the difficulty of performing accurate base detection.

[0123] Some imagers and / or imaging systems are capable of measuring and / or determining the tilt and / or height of a circulation pool. For a first example, a multi-point focus tracker measures the defocus at multiple positions in the image plane. The defocus measurements are processed to determine the tilt at multiple positions. For a second example, a grid of resolution features (such as isolated nanopores) is included in the circulation pool to enable monitoring of defocus. The monitored defocus is processed to determine the tilt at each position of the grid. For a third example, optical aberrations are introduced into the optical system of the imager (e.g., using a phase mask) so that the point spread function is asymmetric between overfocus and underfocus, thereby enabling easy discrimination between overfocus defocus and underfocus defocus. These discriminations are processed to determine the tilt information. For a fourth example, the height of the circulation pool is measured at multiple positions, and the height is used to create a surface map. The surface map is processed to determine the tilt at multiple positions. In various usage scenarios, the height of the circulation pool remains stable during sequencing of the entire circulation pool, channel and / or column, thereby enabling the surface map to be used to process the entire circulation pool, channel and / or column, respectively. In some implementations, the non-planarity of the flow cell is assessed via height measurement and / or height determination. In various implementations, the height of the flow cell is measured and / or determined according to various combinations of the aforementioned techniques to measure and / or determine the tilt of the flow cell.

[0124] In addition to the ability to measure the tilt and / or height of the circulation cell, base calling can also be adapted based on various imaging conditions. Various imaging conditions include the sub-pixel position of the cluster within the pixel, the ratio of signal light to background light, and the size and / or shape of the point spread function. The inventors recognized that various imaging conditions also include varying degrees of defocus, for example, the base calling technology can be further adapted based on varying degrees of defocus, such as within different parts of the field of view of the imager. More specifically, the inventors recognized that measuring and / or determining the tilt of the circulation cell (such as by processing focus / defocus information) and providing the circulation cell tilt measurement to inform the base calling achieves improved accuracy of the base calling.

[0125] According to various specific implementations, measurements of the tilt (and / or height) of the circulation cell and / or measurements of information for determining the tilt (and / or height) of the circulation cell are collected at various time points. According to various specific implementations, the determination of the tilt (and / or height) of the circulation cell is determined at various time points. According to various specific implementations, the measurement of the tilt (and / or height) of the circulation cell and / or the determination of the tilt (and / or height) of the circulation cell are notified to the base call at various time points. According to various specific implementations, focusing adjustment is optionally performed at various time points. Tilt adjustment is optionally performed at various time points. Various time points include one or more times within the life of the instrument, one or more times for each synthesis-by-sequencing run, one or more times for each synthesis-by-sequencing cycle, one or more times per circulation cell, one or more times per channel of the circulation cell, one or more times per column of the channel, one or more times per block, and / or one or more times per one or more parts of the circulation cell, which vary according to different specific implementations.

[0126] In various specific implementations, measurements of the tilt (and / or height) of the circulation cell and / or measurements of information for determining the tilt (and / or height) of the circulation cell are collected in a first time set, the determination of the tilt (and / or height) of the circulation cell is determined in a second time set, and the measurement of the tilt (and / or height) of the circulation cell and / or the determination of the tilt (and / or height) of the circulation cell is notified to the base caller in a third time set. Some specific implementations are arranged for the first time set, the second time set, and the third time set to have a certain pre-arranged relationship with each other. For example, at a first time, such as before any image is captured, the entire surface of the circulation cell is mapped. Then, once per block and based on the map, the base caller is notified of the tilt measurement of the block. Alternatively, four times per block, corresponding to each of the four quarters of the block, the base caller is notified of the tilt measurement of the corresponding quarter of the block. As another example, when each block is imaged, the tilt is determined for each block, and when the base caller processes each block, the measured tilt is notified to the base caller.

[0127] According to various specific implementations, the circulation cell tilt (and / or height) is determined differently at the aforementioned time point, in coordination with the measurement of the circulation cell tilt (and / or height) or alternatively at a time point different from the measurement time of the circulation cell tilt (and / or height).

[0128] According to various specific embodiments, the base call is informed differently of the flow cell tilt (and / or height) at the aforementioned time points, coordinated with the determination of the flow cell tilt (and / or height) or alternatively at time points different from those time points at which the flow cell tilt (and / or height) is determined.

[0129] For clarity, the dependence of base calling on flow cell tilt is described in the context of a hypothetical base calling implementation based on a spatial equalizer and collectively referred to herein as RTA-based base calling. The spatial equalizer implementation is then described in the context of a single base caller. Depending on the implementation, other implementations of the dependence of base calling on flow cell tilt use techniques other than the spatial equalizer technique.

[0130] Multiple base callers

[0131] Figure 1AA The example of the dependence of base detection on the tilt of the circulation cell is illustrated. The upper part of the figure illustrates the training context (such as using a laboratory synthesis while sequencing instrument), and the lower part illustrates the production context (such as using one or more synthesis while sequencing production instruments). The figure illustrates the circulation cell, imaging and RTA part from left to right. As illustrated in the figure, multiple base detectors are used to realize the RTA part, and each base detector is implemented as having a corresponding equalizer and LUT (lookup table) element.

[0132] Conceptually, base calling is performed using knowledge of the tilt of the circulation cell. The circulation cell tilt is measured (tilt measurement) and the circulation cell tilt is assessed (assessed tilt). The assessment determines whether the circulation cell (or any part thereof, such as a channel, column, block or part thereof) is overfocused, in focus or out of focus. Alternatively, the assessment determines whether all or any area of ​​the image (such as a patch, for example one or more clusters in the image, a continuous area of ​​pixels of the image, or one or more regularly partitioned patches of the image) is overfocused, in focus or out of focus. Further alternatively, the assessment is used to determine how to partition the image into regions based on overfocus, in focus or out of focus. A base caller is selected from a plurality of base callers based on the tilt assessment.

[0133] For the first example, if the image area is determined to be in focus, a base caller that is being trained for the in-focus area or has been previously trained for the in-focus area is selected and the image area is processed using the in-focus base caller.The '=Base Caller' element in the figure is an example of an in-focus base caller.

[0134] For the second example, if the image area is determined to be out of focus, a base caller that is being trained for out-of-focus areas or has been previously trained for out-of-focus areas is selected and the image area is processed using the out-of-focus base caller. The '+ Base Caller' element in the figure is an example of an out-of-focus base caller.

[0135] For the third example, if the image area is determined to be out of focus, a base caller that is being trained for out of focus areas or has been previously trained for out of focus areas is selected and the image area is processed using the out of focus base caller. The '-base caller' element in the figure is an example of an out of focus base caller.

[0136] During training, images are collected (e.g., one image for each of multiple blocks of one or more circulation cells) and used in conjunction with ground truth (GT) to learn parameters (sometimes referred to as weights) for training context base callers, such as to determine information stored as coefficients in a LUT. Depending on the specific implementation, each image is processed as a single element, or each image is partitioned into multiple elements. The processing includes: evaluating the tilt associated with each corresponding element in the single element or multiple elements. The evaluated tilt determines which base caller of the multiple base callers is trained for each corresponding element. Each base caller is associated with a corresponding set of GTs, and each base caller includes a corresponding set of LUTs. During training, each base caller is trained independently based on the tilt evaluation. After training is completed, the information in the LUT is provided to the production context RTA base caller for improving base calls compared to base calls without the benefit of training.

[0137] During production, images are collected, such as one image for each of a plurality of blocks of a flow cell, and then processed for base calling. Depending on the specific implementation, each image is processed as a single element, or each image is partitioned into multiple elements. As in training, the processing includes evaluating a tilt associated with a single element or each corresponding element in a plurality of elements. The evaluated tilt determines which base caller of a plurality of base callers is selected to perform base calling of the element. Since each base caller includes a corresponding set of LUTs, the set of LUTs used to determine base calling depends on the evaluated tilt.

[0138] In some implementations, initial training (sometimes referred to as pre-training) is performed in a dedicated training context, and additional training is performed in a production context, such as unique to each production instrument.

[0139] Further details of training base callers and using them to perform base calling during production are provided elsewhere herein, such as with respect to FIG. 1A to FIG. 19D Further details of evaluating skew are disclosed elsewhere herein.

[0140] Fig. 1AB Example Figure 1AADepicted example of base calling dependence on flow cell tilt related operations. These operations are repeated for all blocks of the flow cell. The operations begin by capturing an image of the block and the tilt information associated with the image. Optionally, the image is partitioned into multiple parts. The entire image is then processed as an image region (or each image portion is processed in turn) as follows.

[0141] Depending on the implementation, these portions are, for example: various geometrically regular portions (such as any of a 2×2, 3×3, or 4×4 grid of portions of substantially equal area); edge portions versus interior (non-edge) portions; and / or one or more continuous regions that together form the entire image, where each continuous region is determined to be within a corresponding tilt range (and / or focus category, such as overfocus, infocus, and underfocus).

[0142] The tilt and / or focus of the image region is evaluated, for example, based on the tilt information of the block or the tilt information of the image region as determined from the tilt information of the block.

[0143] In response to the image being determined to be in focus, an in-focus base calling technique (=base caller) is selected. The in-focus base calling technique is used appropriately for training or production, depending on the operational context. For training, the selected base caller is trained using the GT set (=GT) corresponding to the in-focus context, resulting in zero or more updates to coefficients stored in a LUT (=LUT) of the selected base caller.

[0144] In response to the image being determined to be out of focus, an out-of-focus base calling technique (+base caller) is selected. The out-of-focus base calling technique is used appropriately for training or production, depending on the operating context. For training, the selected base caller is trained using the GT set (+GT) corresponding to the out-of-focus context, resulting in zero or more updates to coefficients stored in the LUT (+LUT) of the selected base caller.

[0145] In response to the image being determined to be out of focus, an out-of-focus base calling technique (-base caller) is selected. The out-of-focus base calling technique is used appropriately for training or production, depending on the operating context. For training, the selected base caller is trained using a GT set (-GT) corresponding to the out-of-focus context, resulting in zero or more updates to coefficients stored in a LUT (-LUT) of the selected base caller.

[0146] Figure 1AA and Fig. 1ABThe aforementioned embodiments are specific to the tilt of the flow cell, which results in images that are variously over-focused, in-focused, and out-of-focus. Other embodiments are specific to the height of the flow cell, which results in images that are variously over-focused, in-focused, and out-of-focus. Conceptually, Figure 1AA The tilt assessment is instead an altitude assessment. Altitudes above the DoF use the (+base caller) technique, altitudes within the DoF use the (=base caller) technique, and altitudes below the DoF use the (-base caller) technique. Figure 1AE supply.

[0147] Figure 1AC Elements for imaging a circulation pool are generally illustrated, including selected details related to the tilt of the circulation pool. Depending on the specific implementation, a tilt measurement element is included, which conceptually represents one or more capabilities present in one or more dedicated elements, non-dedicated elements, or a combination of both. Depending on the specific implementation, tilt measurement is achieved by one or more direct and / or indirect measurements and / or determinations based on one or more factors. These factors include tilt, focus and / or distance. Section "Determining Tilt, Focus and / or Distance" (located elsewhere in this document) describes various techniques for measuring and / or determining tilt. In some specific implementations, the tilt measurement element includes capabilities related to tilt measurement as well as capabilities for height measurement. Additional description regarding Figure 1AE supply.

[0148] The circulation pool is usually planar and includes a plurality of substantially parallel channels that are sequentially imaged (aligned and photographed) as a series of blocks organized, for example, into one or more columns, or alternatively, a series of one or more blocks of these channels are continuously imaged (continuously scanned) and processed. The imager includes a sensor, a semi-reflective mirror, and an objective lens. In some specific implementations, the laser and the imager and a mirror positioned to direct the emission of the laser toward the semi-reflective mirror are arranged in a module.

[0149] In some implementations, the imager and flow cell are moved relative to each other (such as by advancing the flow cell along a predetermined path on a movable platform, or by repositioning the imager and laser relative to the flow cell as images are captured). In a continuous scanning implementation, continuous areas of a portion of a channel of the flow cell are imaged and correspond to elements of the series of patches.

[0150] In some specific implementations, the movable platform (sometimes referred to as a stage) includes a circulation cell receiving surface capable of supporting a circulation cell. In some specific implementations, a controller is coupled to the stage and the optical assembly. Some specific implementations of the controller are configured to move the stage and the optical assembly relative to each other in a step-and-shoot manner (sometimes referred to as a step-and-settling technique). In various specific implementations, all or various parts of the tilt measurement and / or height measurement are implemented in the controller. In various specific implementations, a biological sequencing instrument (such as a laboratory instrument or a production instrument) includes all or any part of the elements depicted in the figure. In various specific implementations, the biological sequencing instrument includes a stage, an optical assembly and / or a controller.

[0151] In operation, the imager and the flow cell are moved relative to each other, thereby repositioning the imager from alignment with the (previous) block to alignment with the (current) block. Imaging is performed by operating a laser. The emission of the laser is reflected from the mirror onto a semi-reflective mirror, and then reflected from the semi-reflective mirror to illuminate the block of the flow cell, as shown by the dotted arrow pointing ('Power'). In response to the illumination, the fluorophores of the block fluoresce. The light from the fluorescence passes through an objective lens for focusing and continues through the semi-reflective mirror, thereby forming an image ('Image'). The image is captured by a sensor ('Sensor').

[0152] For example, tilt and / or non-planar blur (conceptually illustrated in the figure by the curved double arrow "Tilt") is introduced by differences in the distances between the imager and various regions of the tile being imaged. For example, a nominally planar flow cell is not optically aligned (e.g., tilted) relative to the imager, such that different portions (e.g., one or more edges) of the same tile are at different distances from the imager. Thus, depending on the DoF of the imager, one of these portions is improperly focused and therefore degraded. As another example, an otherwise nominally planar flow cell has a defect that causes one portion of the tile to be closer to the imager than another portion of the tile.

[0153] The figure depicts an example tilt of the circulation cell. The circulation cell is tilted upward on the left side and tilted downward on the right side. The direction of movement of the imager relative to the circulation cell is from left to right. Therefore, the tilt is in a direction consistent with the direction of movement. Other scenarios can also occur, so that the tilt is in an arbitrary direction relative to the direction of movement. Returning to the figure and the tilt direction therein, the left strip of the circulation cell is overfocused, the center strip of the circulation cell is sharply focused, and the right strip of the circulation cell is out of focus. Overfocus, sharp focus, and out of focus strips are illustrated in the blocks of the circulation cell and in the image formed on the sensor. In response to measuring, determining, and / or evaluating the tilt and / or focus of various areas of the image, base calling depends on the tilt of the circulation cell. Specifically, in response to an image area being overfocused, a '+ base detector' is used for the image area (used in training to determine parameters, and used in production for base calling). In response to an image area being sharply focused, a '= base detector' is used for the image area. In response to an image area being out of focus, a '- base detector' is used for the image area.

[0154] Some implementations of the imager use point imaging techniques that collect relatively small sets of one or more pixels. Some implementations of the imager use area imaging techniques that collect relatively large sets of pixels, such as in a rectangle (e.g., a square shape). Some implementations of the imager use line imaging techniques that collect relatively large sets of pixels, such as in a relatively high aspect ratio rectangular area. Some implementations of the imager (such as some variations of area imaging) use an area sensor that is coplanar with the collection area, and there are very few optical components between the fluorescing fluorophores and the area sensor. Example area sensors are based on semiconductor technology, such as complementary metal oxide semiconductor (CMOS) chips.

[0155] Figure 1AD Selected details related to the tilting of the flow cell are illustrated. Figure 1AC and Figure 1AD Like-named elements in correspond to each other. The upper portion of the figure (top view) is a view of the sensor from above, and depicts various focus strips of the image: overfocus, sharp focus, and underfocus. The lower portion of the figure (side view) is a view observed from the objective side of the imager, and a portion of the flow cell is imaged. The tilt is such that the flow cell surface is above the image plane on the left and below the image plane on the right. Note that the figure is not drawn to scale and the tilt is exaggerated for ease of understanding. Note further that the flow cell is illustrated as being uniformly flat for ease of understanding. The sharp focus strips of the image correspond to the depth of field (DoF) of the imager. Image strips that are outside the DoF relative to the image plane (above the image plane or below the image plane) are blurred. As shown in Figure 1ACIn some implementations, the point spread function (PSF) is asymmetric with respect to the blurred image of overfocus versus the blurred image of underfocus, thereby enabling classification of the blurred image as overfocus or underfocus based on the difference in the PSF.

[0156] Figure 1AE Selected details related to the non-planarity of the flow cell are illustrated. Figure 1AC and Figure 1AE Like-named elements in correspond to each other. Figure 1AD middle, Figure 1AE The upper portion of the figure (top view) is a view of the sensor from above, and depicts the various focus bands of the image: overfocus, sharp focus, and underfocus. The lower portion of the figure (side view) is a view from the objective side of the imager, and a portion of the flow cell is imaged. The non-planarity of the flow cell makes (as shown in Figure 1AD (center) The flow cell surface is above the image plane on the left and below the image plane on the right. Note that the figure is not drawn to scale and that non-planarity is exaggerated for ease of understanding. Note further that the flow cell surface depicted is a two-dimensional cross-section of a three-dimensional object (the flow cell) and that the focused bands appear uniform in the third dimension for ease of understanding. The bands of sharp focus of the image correspond to the depth of field (DoF) of the imager. Bands of the image that are outside the DoF relative to the image plane (above or below the image plane) are blurred. As shown in Figure 1AC In some implementations, the point spread function (PSF) is asymmetric with respect to the blurred image of overfocus versus the blurred image of underfocus, thereby enabling classification of the blurred image as overfocus or underfocus based on the difference in the PSF.

[0157] Depending on the implementation, something like Figure 1AC The tilt measuring element is included. Figure 1AD The height measurement element is included, which conceptually represents one or more dedicated elements, one or more capabilities present in non-dedicated elements, or a combination of both. Depending on the specific implementation, the height measurement is achieved through one or more direct and / or indirect measurements and / or determinations based on one or more factors. These factors include tilt, focus, and / or distance. The section "Determining Tilt, Focus, and / or Distance" (located elsewhere in this document) describes various techniques for measuring and / or determining height.

[0158] Figure 1ADillustrates that tilt alone is not sufficient to determine whether an image is over-focused, in-focus, or out-of-focus. In this figure, the tilt is uniform across the entire image. However, the first portion of the image is over-focused, the second portion of the image is in-focus, and the third portion of the image is out-of-focus. In contrast, Figure 1AE It is illustrated that the height alone is sufficient to determine whether the image is in focus, out of focus, or over focus. The first part of the image is over focus, the second part of the image is in focus, and the third part of the image is out of focus.

[0159] Single base caller

[0160] The foregoing description assumes a base calling implementation based on a spatial equalizer and generally referred to herein as RTA-based base calling. Specific techniques for implementing RTA-based base calling using a spatial equalizer are described below. In various implementations, Figure 1A The equalizer base detector 104 (sometimes referred to as the equalizer 104) is Figure 1AA to Figure 1A Example implementations of '+', '=', and '-' base caller elements of D. Figure 1A The ground truth base call 112 is Figure 1AA to Figure 1A Examples of '+', '=' and '-' GT elements of D. Figure 1A The lookup table 106 (sometimes referred to as LUT 106 or LUT set 106) is Figure 1AA to Figure 1A Example implementations of the '+', '=' and '-' LUT elements of D. Correspondingly, Figure 1A The sequencing image 102 corresponds to Figure 1AA to Figure 1A D image element, and Figure 1A The trainer 114 corresponds to Figure 1AA The trainer components and Fig. 1AB Trainer element.

[0161] Lookup table generation

[0162] Figure 1A One specific implementation of generating a lookup table (LUT) (or LUT bank) 106 by training an equalizer 104 is shown. The equalizer 104 is also referred to herein as an equalizer-based base caller 104. The system 100A includes a trainer 114 that trains the equalizer 104 using least squares estimation. Additional details about the equalizer and least squares estimation can be found in the appendix included with this application.

[0163] Sequencing images 102 are generated during a sequencing run performed by a sequencing instrument such as Illumina's iSeq, HiSeqX, HiSeq 3000, HiSeq4000, HiSeq 2500, NovaSeq 6000, NextSeq 550, NextSeq 1000, NextSeq 2000, NextSeqDx, MiSeq, and MiSeqDx. In one specific implementation, Illumina sequencers employ cyclic reversible termination (CRT) chemistry for base calling. The process relies on growing a nascent chain complementary to a template chain with fluorescently labeled nucleotides while tracking the emission signal of each newly added nucleotide. The fluorescently labeled nucleotides have a 3' removable block that anchors the fluorophore signal of the nucleotide type.

[0164] Sequencing is performed in repeated cycles, each cycle consisting of three steps: (a) extending the nascent chain by adding fluorescently labeled nucleotides; (b) exciting the fluorophore using one or more lasers of the sequencing instrument's optical system and imaging through different filters of the optical system to produce a sequencing image; and (c) cleaving the fluorophore and removing the 3' block to prepare for the next sequencing cycle. The incorporation and imaging cycles are repeated until a specified number of sequencing cycles are reached, thereby defining the read length. Using this method, each cycle interrogates a new position along the template chain.

[0165] The great power of Illumina sequencers comes from its ability to simultaneously perform and sense millions or even billions of analytes (e.g., clusters) undergoing CRT reactions. Clusters include approximately one thousand identical copies of the template strand, but the size and shape of the clusters are different. Before the sequencing run, clusters from the template strands are grown by bridge amplification or exclusion amplification of the input library. The purpose of amplification and cluster growth is to increase the intensity of the emission signal because the imaging device cannot reliably sense the fluorophore signal of a single strand. However, the physical distance of the strands within the cluster is small, so the imaging device perceives the strand cluster as a single point.

[0166] Sequencing occurs in a flow cell (a small glass slide that holds the input chain). The flow cell is connected to an optical system that includes microscope imaging, excitation lasers, and fluorescence filters. The flow cell includes a plurality of chambers referred to as channels. The channels are physically separated from each other and may include different labeled sequencing libraries that can be distinguished without sample cross contamination. In some implementations, the flow cell includes a patterned surface. "Patterned surface" refers to the arrangement of different regions in or on the exposed layer of a solid carrier. For example, one or more regions in these regions may be features in which one or more amplification primers are present. Features may be separated by gap regions in which there is no amplification primer. In some implementations, the pattern may be an xy format of features in the form of rows and columns. In some implementations, the pattern may be a repeated arrangement of features and / or gap regions. In some implementations, the pattern may be a random arrangement of features and / or gap regions. Exemplary patterned surfaces useful for the methods and compositions described herein are described in U.S. Pat. No. 8,778,849, U.S. Pat. No. 9,079,148, U.S. Pat. No. 8,778,848, and U.S. Publication No. 2014 / 0243224, each of which is incorporated herein by reference.

[0167] In some implementations, the flow cell includes an array of holes or depressions in the surface. This can be manufactured using a variety of techniques as generally known in the art, including but not limited to photolithography, imprinting techniques, molding techniques, and micro-etching techniques. Those skilled in the art will appreciate that the techniques used will depend on the composition and shape of the array substrate.

[0168] The features in the patterned surface can be holes (e.g., micropores or nanopores) in a hole array on a glass, silicon, plastic or other suitable solid support having a patterned and covalently attached gel such as poly (N- (5-azidoacetamidopentyl) acrylamide-co-acrylamide) (PAZAM, see, e.g., U.S. Publication Nos. 2013 / 184796, WO 2016 / 066586, and WO 2015-002813, each of which is incorporated herein by reference in its entirety). The method produces a gel pad for sequencing that can be stable in sequencing runs with a large number of cycles. The covalent attachment of the polymer to the hole helps to maintain the gel as a structured feature during multiple uses and throughout the life of the structured substrate. However, in many specific implementations, the gel need not be covalently attached to the hole. For example, under some conditions, a silane-free acrylamide (SFA, see, e.g., U.S. Pat. No. 8,563,477, which is incorporated herein by reference in its entirety) that is not covalently attached to any portion of the structured substrate can be used as a gel material.

[0169] In a particular implementation, a structured substrate can be made by patterning a solid support material with pores (e.g., micropores or nanopores), coating the patterned support with a gel material (e.g., PAZAM, SFA, or a chemically modified variant thereof, such as an azidated version of SFA (azido-SFA)), and polishing the gel-coated support, for example, via chemical or mechanical polishing, thereby retaining the gel in the pores while removing or inactivating substantially all of the gel from the interstitial regions between the pores on the surface of the structured substrate. Primer nucleic acids can be attached to the gel material. A solution of target nucleic acids (e.g., a fragmented human genome) can then be contacted with the polished substrate so that each target nucleic acid will be seeded into each pore via interaction with the primers attached to the gel material; however, due to the absence of the gel material or the inactivation of the gel material, the target nucleic acid will not occupy the interstitial region. Amplification of the target nucleic acid will be confined to the pores because the absence of gel or the inactivation of the gel in the interstitial region will prevent the growing nucleic acid population from migrating outward. The process is manufacturable, scalable, and utilizes conventional micro- or nano-manufacturing methods.

[0170] The imaging device of the sequencing instrument (e.g., a solid-state imaging device such as a charge-coupled device (CCD) or a complementary metal oxide semiconductor (CMOS) sensor) takes snapshots at multiple locations along the channel, in a series of non-overlapping areas (called blocks). For example, there may be sixty-four blocks or ninety-six blocks per channel. Blocks hold hundreds of thousands to millions of clusters.

[0171] The output of a sequencing run is a sequencing image, each of which depicts the intensity emission of a cluster and its surrounding background. A sequencing image depicts the intensity emission due to the incorporation of nucleotides into the sequence during sequencing. These intensity emissions come from the associated analyte / cluster and its surrounding background.

[0172] The sequencing image 102 originates from a plurality of sequencing instruments, sequencing runs, cycles, circulation cells, blocks, holes, and clusters. In one specific implementation, the sequencing image is processed by an equalizer 104 on the basis of imaging channels. The sequencing run generates m images corresponding to m imaging channels in each sequencing cycle. In one specific implementation, each imaging channel corresponds to a filter wavelength band in a plurality of filter wavelength bands. In another specific implementation, each imaging channel corresponds to an imaging event in a plurality of imaging events at a sequencing cycle. In yet another specific implementation, each imaging channel corresponds to a combination of irradiation with a specific laser and imaging through a specific optical filter. In different specific implementations such as 4-channel chemistry, 2-channel chemistry, and 1-channel chemistry, m is 4 or 2. In other specific implementations, m is 1, 3, or greater than 4.

[0173] In another specific implementation, the input data is based on the pH change induced by the release of hydrogen ions during molecular expansion. The pH change is detected and converted into a voltage change proportional to the number of bases introduced (for example, in terms of IonTorrent). In another specific implementation, the input data is constructed according to nanopore sensing, which uses a biosensor to measure the interruption of the current when the analyte passes through the nanopore or approaches its orifice, while determining the type of the base. For example, Oxford Nanopore Technology (ONT) sequencing is based on the following concept: single-stranded DNA (or RNA) is passed through the membrane via a nanopore, and a voltage difference is applied across the membrane. The nucleotides present in the hole will affect the resistance of the hole, so the current measurement results over time can indicate the sequence of the DNA base passing through the hole. The current signal (called 'squiggle' due to its appearance when drawn) is the raw data collected by the ONT sequencer. These measurements are stored as 16-bit integer data acquisition (DAC) values ​​obtained at (for example) 4kHz frequency. With a DNA strand speed of about 450 base pairs per second, this gives an average of approximately nine raw observations per base. The signal is then processed to identify breaks in the pore opening signal corresponding to individual read segments. These raw signal segments are base called, the process of converting DAC values ​​into DNA base sequences. In some implementations, the input data includes normalized or scaled DAC values. Additional information about non-image-based sequencing data can be found in U.S. Provisional Patent Application No. 62 / 849,132, entitled “Base Calling Using Convolutions,” filed on May 16, 2019 (Attorney Docket No. ILLM 1011-2 / IP-1750-PR2), U.S. Provisional Patent Application No. 62 / 849,133, entitled “Base Calling Using Compact Convolutions,” filed on May 16, 2019 (Attorney Docket No. ILLM 1011-3 / IP-1750-PR3), and U.S. Non-Provisional Patent Application No. 16 / 826,168, entitled “Artificial Intelligence-Based Sequencing,” filed on March 21, 2020 (Attorney Docket No. ILLM 1008-20 / IP-1752-PRV).

[0174] train

[0175] The equalizer 104 generates a LUT library having a plurality of LUTs (equalizer filters) 106, which have sub-pixel resolution. In one specific implementation, the number of LUTs 106 generated by the equalizer 104 for the LUT library depends on the number of sub-pixels into which the sensor pixels of the sequenced image 102 are or can be divided. For example, if the sensor pixels of the sequenced image 102 can each be divided into n×n sub-pixels (e.g., 5×5 sub-pixels), the equalizer 104 generates n 2 There are LUTs 106 (eg, 25 LUTs).

[0176] In one implementation of training, data from the sequencing images are binned by well sub-pixel location. For example, for a 5×5 LUT, 1 / 25 of the wells are centered in bin (1,1) (e.g., the upper left corner of the sensor pixel), 1 / 25 of the wells are in bin (1,2), and so on. The equalizer coefficients for each well center bin are determined using a least squares estimate of the subset of data from the wells in each bin. The input to the equalizer 104 is the raw sensor pixels of the sequencing images for those bins. The resulting estimated equalizer coefficients are different for each bin.

[0177] Each LUT has a number of coefficients learned from training. In one implementation, the number of coefficients in the LUT corresponds to the number of sensor pixels used to base call the clusters. For example, if the local grid (image or pixel patch) of sensor pixels used to base call the clusters is of size p×p (e.g., a 9×9 pixel patch), then each LUT has p 2 coefficients (e.g., 81 coefficients).

[0178] The training produces equalizer coefficients that are configured to mix / combine the intensity values ​​of pixels that depict the intensity emission from the target cluster being base called and the intensity emission from one or more neighboring clusters in a way that maximizes the signal-to-noise ratio. The signal for which the signal-to-noise ratio is maximized is the intensity emission from the target cluster, and the noise for which the signal-to-noise ratio is minimized is the intensity emission from the neighboring clusters, i.e., spatial crosstalk, plus some random noise (e.g., to account for background intensity emissions). The equalizer coefficients are used as weights, and the mixing / combining includes performing an element-by-element multiplication between the equalizer coefficients and the intensity values ​​of the pixels to calculate a weighted sum of the intensity values ​​of the pixels.

[0179] According to one specific implementation, during training, the equalizer 104 learns to maximize the signal-to-noise ratio through least squares estimation. Using least squares estimation, the equalizer 104 is trained to estimate the shared equalizer coefficients from the pixel intensities around the subject hole and the desired output. Least squares estimation is well suited for this purpose because its output minimizes the squared error and takes into account the coefficients of the effects of noise amplification.

[0180] When the intensity channel is on, the expected output is a pulse at the hole position (point source), and when the intensity channel is off, the expected output is the background level. In some implementations, the ground truth base call 112 is used to generate the expected output. In some implementations, the ground truth base call 112 is modified to take into account the per-hole DC offset, amplification factor, degree of polyclonality, and gain offset parameters included in the least squares estimate. In one implementation, during training, the DC offset, i.e., a fixed offset, is calculated as part of the least squares estimate. During inference, the DC offset is added as a bias to each equalizer calculation result.

[0181] In one specific implementation, Illumina's real-time analysis (RTA) base caller (which does not use an equalizer) is used to estimate the expected output. Details about RTA can be found in U.S. Patent Application No. 13 / 006,206, which is incorporated by reference as if fully set forth herein. Since RTA has a low base call error rate, the RTA base caller is used as the source of ground truth base calls 112. The base call errors are averaged across many training examples. In another specific implementation, ground truth base calls 112 are obtained using aligned genomic data, which has good quality because the aligned genomic data can use reference genomes and true value information, which are combined with knowledge obtained from multiple sequencing platforms and sequencing runs to average out noise.

[0182] The ground truth base calls 112 are base-specific intensity values ​​that reliably represent the intensity distributions of bases A, C, G, and T, respectively. A base caller (such as RTA) base calls clusters by processing the sequencing image 102 and generating a color-by-color intensity value / output for each base call. These color-by-color intensity values ​​can be considered as base-by-base intensity values ​​because, depending on the type of chemistry (e.g., 2-color chemistry or 4-color chemistry), a color is mapped to each of the bases A, C, G, and T. The base with the closest matching intensity distribution is called.

[0183] Fig.16 A specific implementation of base-by-base Gaussian fits containing base-by-base intensity targets at their centers that are used as ground truth for error calculations during training is shown. The base-by-base intensity outputs produced by the base callers for a large number of base calls in the training data (e.g., tens, hundreds, thousands, or millions of base calls) are used to produce the base-by-base intensity distribution. Fig.16A graph is shown with four Gaussian clouds, which are the probability distributions of the base-by-base intensity outputs for bases A, C, G, and T. The intensity values ​​at the centers of these four Gaussian clouds are used as ground truth intensity targets (assuming ground truth base calls 112 for bases A, C, G, and T, respectively), and are referred to herein as intensity targets.

[0184] Consider that during training, the input image data fed to the equalizer 104 is annotated with base "A" as the ground truth base call. Then, the target / desired output of the equalizer 104 is Fig.16 The intensity value at the center of the green cloud in , i.e., the intensity target for base A. Similarly, for the ground truth base call of base “C”, the expected output of equalizer 104 is Fig.16 , i.e., the intensity target for base C. Thus, the target or desired output during training of the equalizer 104 is the average intensity of the corresponding bases A, C, G, and T after averaging in the training data. In one implementation, the trainer 114 fits the coefficients of the equalizer 104 using least squares estimation to minimize the equalizer output error with respect to these intensity targets.

[0185] In one specific implementation, during training, the equalizer 104 applies the coefficients in a given lookup table (LUT) to the pixels marked with a given base in the sequencing image. This includes multiplying these coefficients element by element with the intensity value of the pixel, and generating a weighted sum of the intensity value, wherein these coefficients act as / as / as weights. The weighted sum then becomes the predicted output of the equalizer 104. Then, based on a cost / error function (e.g., sum of squared errors (SSE)), the error (e.g., minimum square error, minimum mean square error) between the weighted sum and the intensity target determined for a given base (e.g., from the center of the corresponding intensity Gaussian fit, as the average intensity observed for a given base) is calculated. The cost function (such as SSE) is a differentiable function for estimating the equalizer coefficients using an adaptive method, so the derivatives of the error with respect to the coefficients can be evaluated, and then these derivatives are used to update the coefficients with the value that minimizes the error. Repeat the process until the updated coefficients no longer reduce the error. In other specific implementations, a batch least squares method is used to train the equalizer 104.

[0186] In other specific implementations, Fig.16 The base-by-base intensity distribution / Gaussian cloud shown can be generated on a "well-by-well" basis and corrected for noise by adding DC offsets, amplification factors, and / or phase adjustment parameters. In this way, depending on the well position of a particular well, a corresponding base-by-base Gaussian cloud can be used to generate a target intensity value for that particular well.

[0187] In one implementation, a bias term is added to the dot product to produce the output of the equalizer 104. During training, the bias parameter can be estimated using a similar method used to learn the equalizer coefficients, i.e., the least squares method or the least mean square (LMS) method. In some implementations, the value of the bias parameter is a constant value equal to 1, i.e., a value that does not vary with the input pixel intensity. There is a bias for each set of equalizer coefficients. The bias is learned during training and then fixed for use during inference. The learned bias represents the DC offset used with the learned coefficients of each LUT in each equalizer calculation during inference. The bias takes into account random noise caused by different cluster sizes, different background intensities, varying stimulus responses, varying focus, varying sensor sensitivity, and varying lens aberrations.

[0188] In other decision-oriented implementations, the output of the equalizer 104 is assumed to be correct for training purposes.

[0189] In another implementation of training, the equalizer 104 generates only a single LUT (equalizer filter) for one bin, and then generates the remaining equalizer filters for the remaining bins using multiple per-bin interpolation filters 108. In this implementation, for each training example, the sensor pixels surrounding each hole are resampled / interpolated into a well-aligned space (i.e., the holes are centered in their respective pixel patches / local grids). The resampled pixels for each example are then aligned consistently across all holes.

[0190] However, in order to apply the single equalizer filter generated by the equalizer 104 to base calling in a real online system, we need to pre-process the raw sensor pixels of the sequencing image to return to a well-aligned space, that is, interpolation is performed on the raw pixels around each hole, and the interpolation parameters vary according to the sub-pixel position of the given hole. In order to avoid this interpolation process, we pre-calculate the total response of the sub-pixel position of the given hole. We calculate the well-aligned equalizer input value by interpolating the raw pixel intensity into the well-aligned pixel space. We convolve the interpolation response and the equalizer response together to reduce calculations. Since the interpolation filter varies with the sub-pixel hole position, this gives each sub-pixel hole position a different set of equalizer coefficients / equalizer filters, thereby generating the remaining LUTs for the remaining bins. Therefore, in this specific implementation of training, only the coefficients of a single equalizer filter are trained during training, but the pre-calculation process generates a library of LUT-based equalizers by applying the bin-specific interpolation filter 108 in conjunction with the single equalizer filter, where the LUT index is the sub-pixel hole position.

[0191] The trainer 114 can train the equalizer 104 and generate trained coefficients for the LUT 106 using a variety of training techniques. Examples of training techniques include least squares estimation, ordinary least squares, least mean square method, and recursive least squares. The least squares technique adjusts the parameters of the function to best fit the data set so that the sum of the squared residuals is minimized. Additional details about the least squares estimation algorithm can be found here—Least squares, https: / / en.wikipedia.org / w / index.php? title=Least_squares&oldid=951737821 (last accessed on April 28, 2020), which is incorporated by reference as if fully set forth herein. Ordinary least squares is a least squares method used for estimation in a linear regression model. Additional details about the ordinary least squares algorithm can be found here—Ordinary least squares, https: / / en.wikipedia.org / w / index.php? title=Ordinary_least_squares&oldid=951770366 (last visited on April 28, 2020), which is incorporated by reference as if fully set forth herein. In other specific implementations, other estimation algorithms and adaptive equalization algorithms may be used to train the equalizer 104.

[0192] The equalizer 104 can be trained in an offline mode. In offline mode, according to one specific implementation, the trained coefficients of the LUT 106 are generated using the following batch least squares equalization logic:

[0193]

[0194] In the above relationship, the LUT coefficients are βhat, the pixel intensity is X, and the target is y. A DC term is also added to the pixel intensity and the coefficients (e.g., an extra intensity term that is fixed to 1 for all cases). Then, as an example, consider that X is a matrix of size 82 (= 9×9 input intensities plus a constant DC term) × the number of training examples in the batch, and Y is the target output for each training example, i.e., each value is the intensity center of an on / off cloud that depends on the true value of the training example. βhat is then the set of coefficients that minimizes the sum of squared residuals, and also has size 82 (= 9×9 coefficients plus 1 DC term).

[0195] The equalizer 104 can also be trained in an online mode so that the coefficients of the LUT 106 are adapted to track changes in temperature (e.g., optical distortion), focus, chemistry, machine-specific variations, etc. on a "block-by-block" or sub-block basis while the sequencer is running and the sequencing run is performed cyclically. In the online mode, the trained coefficients of the LUT 106 are generated using adaptive equalization. The online mode uses the least mean squares as the training algorithm, which is a form of stochastic gradient descent. Additional details about the least mean squares algorithm can be found here - Least mean squares filter, https: / / en.wikipedia.org / w / index.php?title=Least_mean_squares_filter&oldid=941899198 (last visited on April 28, 2020), which is incorporated by reference as if fully set forth herein.

[0196] The least mean square technique uses the gradient of the square error relative to each coefficient to move these coefficients in the direction that minimizes the cost function, where the cost function is the expected value of the square error. This has extremely low computational cost, performing only the multiplication and accumulation operations of each coefficient. Except for the coefficients, no long-term storage is required. The least mean square technique is very suitable for processing huge amounts of data (e.g., parallel processing of data from billions of clusters). Extensions of the least mean square technique include normalized least mean square and frequency domain least mean square, which can also be used in this article. In some specific implementations, the least mean square technique can be applied in a decision-oriented manner, assuming that our decision is correct, that is, our error rate is very low, and a small Mu value will filter out any disturbed updates caused by incorrect base calls.

[0197] Fig.18 One specific implementation of an adaptive equalization technique that can be used to train the equalizer 104 is shown. Here, the equalization logic is y=x.h+d, where x is the input pixel intensity, h is the equalizer coefficient, and d is the DC offset. In one specific implementation, x and h are row and column vectors of length 81, respectively. The vector model is equivalent to the dot product of a 9×9 matrix representing the input pixels and coefficients. The cost is the expected value of the squared error. The gradient update moves each coefficient in the direction that reduces the expected value of the squared error. This results in the following update:

[0198]

[0199] For most systems, the expected function E{x(n)e * (n)} must be approximated. This can be done using the following unbiased estimator

[0200]

[0201] where N indicates the number of samples used for the estimation. The simplest case is N = 1

[0202]

[0203] For this simple case, the update algorithm is as follows

[0204]

[0205] In practice, this constitutes the update algorithm of the LMS filter.

[0206] In the above relationship, h is a vector of equalizer coefficients (e.g., 9×9 equalizer coefficients), x is a vector of equalizer input intensities (e.g., 9×9 pixels in a pixel patch), and e is the error of the equalizer calculation performed using the 81 values ​​in x, i.e., there is only 1 error term for each equalizer output.

[0207] Applying this update generates new estimates of the 9×9 equalizer coefficients that move these coefficients in a direction that (on average) reduces the mean square error (MSE). There are 81 updates, one for each equalizer coefficient. In some implementations, Mu is a small constant used to change the adaptation rate / convergence speed. The DC term update can be calculated in a similar manner. The gain term update can also be calculated in a similar manner.

[0208] The coefficient sets can be shared between, for example, patches, regions of patches, or flow cell surfaces. This is achieved by saving and restoring the coefficient sets when the input data changes.

[0209] In some implementations, since linear interpolation is applied to the set of coefficients, these updates are applied slightly differently as follows:

[0210] h(q,n+1)=h(q,n)+λ_q.mu.x(n).e(n)

[0211] In the above relationship, h(q,n) is the weight q at cycle n, λ_q is the linear interpolation weight for a particular set of coefficients, and may include four updates per equalizer output due to two-dimensional linear interpolation.

[0212] The recursive least squares technique extends the least squares technique to a recursive algorithm. Additional details about the recursive least squares algorithm can be found here—Recursive least squares filter, https: / / en.wikipedia.org / w / index.php?title=Recursive_least_squares_filter&oldid=916406502 (last accessed on April 28, 2020), which is incorporated by reference as if fully set forth herein.

[0213] In a multi-domain implementation, the LUT 106 and its trained coefficients may be generated along multiple domains. Examples of these domains include sequencers or sequencing instruments / machines (e.g., Illumina's NextSeq, MiSeq, HiSeq, and their respective models), sequencing protocols and chemistries (e.g., bridge amplification, exclusion amplification), sequencing runs (e.g., forward and reverse), sequencing illumination (e.g., structured, unstructured, angled), sequencing equipment (e.g., overhead CCD camera, underlying CMOS sensor, one laser, multiple lasers), imaging technology (single channel, dual channel, quad channel), flow cells (e.g., patterned, non-patterned, embedded on a CMOS chip, underlying CCD camera), and spatial resolution on the flow cell (e.g., Different areas or quadrants within a flow cell (e.g., different blocks on a flow cell (e.g., edge holes on a block that are closer to a laser or camera or fluid system)) and at different areas within a block (e.g., different channels on a block (e.g., edge holes on a channel that are closer to a laser or camera or fluid system)). Those skilled in the art will appreciate that other selectable domains and parameters typically associated with sequencing are similarly included (e.g., image processing algorithms, image registration algorithms, ground truth annotation schemes (e.g., continuous labels such as intensity values, hard labels such as one-hot encoding, soft labels such as softmax scores), temperature, focus, lenses, sequencing reagents, sequencing buffers).

[0214] Sequencing images generated using corresponding ones of these domains can be used to create discrete and different training sets for the corresponding domains. These discrete training sets can be used to train the equalizer 104 to generate LUTs with trained coefficients for the corresponding domains. Depending on which domain or combination of domains is used in the current or ongoing sequencing operation, the specially trained trained coefficients generated for the corresponding ones of the multiple domains can be stored and accessed accordingly during the online mode. For example, for a sequencing operation, a first set of coefficients that is more suitable for the edge holes of a circulation cell and a second set of coefficients that is more suitable for the center holes of the same circulation cell can be used.

[0215] In one implementation, a configuration file may specify different combinations of these domains, and may be analyzed during online mode to select different sets of coefficients specific to the domains identified by the configuration file.

[0216] In a multi-training implementation, the equalizer 104 is subjected to pre-training as well as training. That is, the LUT 106 and its coefficients are first trained using a first training technique during a pre-training phase, and then the LUT and its coefficients are retrained or further trained using a second training technique during a further training phase. The first training technique and the second training technique may be any of the training techniques listed above. The first training technique and the second training technique may be the same, or they may be different. For example, the pre-training phase may be an offline mode using a batch ordinary least squares training technique, while the training phase may be an online mode using an iterative stochastic least mean squares technique.

[0217] In some implementations, multi-domain implementations and multi-training implementations may be combined such that domain-specific coefficients are pre-trained and then further trained in a domain-specific manner. That is, further training (e.g., in an online mode) retrains the coefficients for a specific domain using only data representing the specific domain and similar to the data used in the pre-training phase. In other knowledge transfer implementations, pre-training and training may use training data from across domains, e.g., a set of coefficients is generated using images from a patterned flow cell during pre-training, but retrained using images from a non-patterned flow cell during a subsequent training phase.

[0218] Spatial Crosstalk Attenuator

[0219] Figure 2 Describes the use Figure 1A In one implementation, the trained LUT / equalizer filter 106 is used to attenuate spatial crosstalk from sensor pixels and base call clusters using the crosstalk-corrected sensor pixels. The trained equalizer base caller 104 operates during the inference phase when base calls occur. In some implementations, Figure 2 The actions shown are performed in a pre-processing stage prior to the base calling stage and generate crosstalk corrected image data that is used by the base caller for base calling.

[0220] In one implementation, the equalizer coefficients are applied to pixel patches 120 (image patches or local grids of sensor pixels) extracted from the sequencing image 116 on an imaging channel basis and a target cluster basis. With respect to the imaging channel basis, in some implementations, each sequencing image has image data for multiple imaging channels. Consider the optical system of an Illumina sequencer, which uses two different imaging channels: a red channel and a green channel. Then, in each sequencing cycle, the optical system produces a red image with red channel intensity and a green image with green channel intensity, which together form a single sequencing image (such as the RGB channels of a typical color image).

[0221] During training, the coefficients are trained / configured to maximize the signal-to-noise ratio (SNR) by minimizing the error between the predicted / estimated output and the expected / actual output. An example of an error is the mean square error (MSE) or mean square deviation (MSD). The signal for which the SNR is maximized is the intensity emission from the target cluster being base called (e.g., the cluster centered in the image patch), while the noise for which the SNR is minimized is the intensity emission from one or more adjacent clusters, i.e., spatial crosstalk, plus other noise sources (e.g., to account for background intensity emission). The trained coefficients are element-wise multiplied by the pixels of the image patch to calculate a weighted sum of the intensity values ​​of these pixels. This weighted sum is then used to base call the target cluster.

[0222] In one implementation, the patch extractor 118 extracts red pixel patches from the red channel and green pixel patches from the green channel from a single sequencing image. In other implementations, red pixel patches are extracted from the red sequencing image of the tested sequencing cycle, and green pixel patches are extracted from the green sequencing image of the tested sequencing cycle. The coefficients of LUT 106 are used to generate the red weighted sum of the red pixel patches and the green weighted sum of the green pixel patches. Then, both the red weighted sum and the green weighted sum are used to base call the target cluster. The size of the image patch 120 is w×h, where w (width) and h (height) are any numbers in the range of 1 to 10,000 (e.g., 3×3, 5×5, 7×7, 9×9, 15×15, 25×25). In some implementations, w and h are the same. In other implementations, w and h are different. Those skilled in the art will appreciate that data for one, two, three, four or more channels or images may be generated for each sequencing cycle of a target cluster, and one, two, three, four or more patches may be extracted to generate one, two, three, four or more weighted sums, respectively, for base calling of the target cluster.

[0223] With respect to extracting the target cluster basis of the pixel patches 120 from the sequencing image 116, the pixel extractor 118 extracts the pixel patches 120 based on the locations of the centers of these clusters / wells on the sequencing image 116, such that the center pixel of each extracted pixel patch contains the center of the target cluster / well. In some implementations, the patch extractor 118 locates the cluster / well centers on the sequencing image, identifies those pixels in the sequencing image that contain the cluster / well centers (i.e., the center pixels), and extracts the pixel patches of the contiguous neighboring pixel neighborhoods around the center pixels.

[0224] Figure 2 An example of a sequencing image 200 containing a central / point source of at least five clusters / wells on a flow cell is visualized. The pixels of the sequencing image 200 depict the intensity emission from the target cluster 1 (blue), and the intensity emission from the attached adjacent clusters 2 (purple), 3 (orange), 4 (brown), and 5 (green).

[0225] Figure 3 An example of extracting a pixel patch 300 (yellow) from a sequencing image 200 such that the center of target cluster 1 (blue) is contained in the center pixel 206 of the pixel patch 300 is visualized. Figure 3 Other pixels 202, 204, 214 and 216 are also shown, which contain the centers of adjacent clusters 2 (purple), 3 (orange), 4 (brown) and 5 (green), respectively.

[0226] Figure 4 An example of a cluster-to-pixel signal 400 is visualized. In one implementation, the sensor pixels (yellow) are in the pixel plane. The periodically distributed clusters 412 in the sample plane (e.g., a flow cell) cause spatial crosstalk. In one implementation, the target cluster and additional adjacent clusters are periodically distributed on the flow cell in a diamond shape and fixed to the holes of the flow cell. In another implementation, the target cluster and additional adjacent clusters are periodically distributed on the flow cell in a hexagonal shape and fixed to the holes of the flow cell. The signal cone 402 from the cluster is optically coupled to a local grid of sensor pixels (e.g., pixel patch 300) through at least one lens (e.g., one or more lenses of an overhead or adjacent CCD camera).

[0227] In addition to diamond and hexagonal shapes, the clusters can also be arranged in other regular shapes (such as squares, rhombuses, triangles, etc.). In yet other implementations, the clusters are arranged in a random, non-periodic arrangement on the sample plane. Those skilled in the art will appreciate that the clusters can be arranged in any arrangement on the sample plane, depending on the needs of a particular sequencing implementation.

[0228] Figure 5One example of cluster-to-pixel signal overlap 500 is visualized. Signal cones 402 overlap and impinge on sensor pixels, creating spatial crosstalk 502.

[0229] Figure 6 An example of a cluster signal pattern 600 is visualized. In one implementation, the cluster signal pattern 600 follows a decay pattern 602, where the cluster signal is strongest at the cluster center and decays as it propagates away from the cluster center.

[0230] Figure 6 Also shown is an example of equalizer coefficients 604 that are trained / configured to maximize the signal-to-noise ratio by computing a weighted sum of the intensity emission from target cluster 1 and the intensity emission from neighboring clusters 2, 3, 4, and 5. The equalizer coefficients 604 act as weights. The weighted sum is computed by element-wise multiplying a first matrix including the equalizer coefficients 604 with a second matrix including pixel intensity values, where each pixel intensity value is the sum of the emission from one or more of clusters 1, 2, 3, 4, and 5 plus other noise sources in the system measured by the pixel sensor.

[0231] Figure 7 An example of a sub-pixel LUT grid 700 for attenuating spatial crosstalk from a pixel patch 300 is visualized. Each pixel in the pixel patch 300 may be divided into a plurality of sub-pixels. Figure 7 , the pixel 206 containing the center of the target cluster 1 (blue) is divided into as many sub-pixels as the number of trained LUTs 106. That is, the pixel 206 is partitioned into the same number of sub-pixels as the number of bins for which the equalizer 104 generated the LUTs 106 during training. Therefore, each sub-pixel of the pixel 206 corresponds to a corresponding LUT in the LUT library generated by the equalizer 104 using decision-directed feedback and least squares estimation.

[0232] exist Figure 7In the example shown, the pixel 206 (center pixel) is divided into a 5×5 sub-pixel LUT grid 700 to produce 25 sub-pixels, which correspond to the 25 LUTs (equalizer filters) generated by the adaptive filter 104 as a training result. Each of the 25 LUTs includes coefficients configured to mix / combine the intensity values ​​of pixels in the pixel patch 300 in a manner that maximizes the signal-to-noise ratio, these pixels depicting the intensity emission from the target cluster 1 and the intensity emission from the neighboring clusters 2, 3, 4, and 5. The signal for which the signal-to-noise ratio is maximized is the intensity emission from the target cluster, and the noise for which the signal-to-noise ratio is minimized is the intensity emission from the neighboring clusters 2, 3, 4, and 5, i.e., spatial crosstalk, plus some random noise (e.g., to account for background intensity emission). These LUT coefficients are used as weights, and the mixing / combining includes performing element-by-element multiplication between the LUT coefficients and the intensity values ​​of the pixels in the pixel patch 300 to calculate the weighted sum of the intensity values ​​of these pixels.

[0233] The number of coefficients in each of the 25 LUTs is the same as the number of pixels in the pixel patch 300, i.e., there is a grid of 9×9 coefficients in each LUT for the 9×9 pixels in the pixel patch 300. This occurs because the coefficients are multiplied element-by-element with the pixels in the pixel patch 300.

[0234] In one embodiment, the pixel-to-subpixel converter ( Figure 1B The LUT grid 700 (not shown) divides the pixel 206 into sub-pixel LUT grids 700 based on a preset pixel divisor parameter (e.g., 1 / 5 pixel per sub-pixel to generate a 5×5 sub-pixel LUT grid 700). For example, the pixel may be divided into five sub-pixel bins with the following boundaries: -0.5, -0.3, -0.1, 0.1, 0.3, 0.5.

[0235] exist Figure 7, note that the center of target cluster 1 (blue) is substantially concentric with the center of transformed pixel 702. This occurs because the sequencing image 200, and therefore the pixel patch 300, is resampled in the following manner to make the center of target cluster 1 (blue) substantially concentric with the center of transformed pixel 702: (i) the sequencing image 200 is registered relative to the template image, and the affine transformation parameters and nonlinear transformation parameters are determined, (ii) the position coordinates of target cluster 1 (blue) are transformed into image coordinates of sequencing image 200 using these parameters, and (iii) interpolation is applied using the transformed position coordinates of target cluster 1 (blue) so that its center is substantially concentric with the center of transformed pixel 702. The positions of the holes in the sample plane are known and can be used to calculate the positions of the equalizer inputs for the particular holes in the original pixel space. We can then use interpolation to recover the intensities at those positions from the original image.

[0236] Figure 8 The selection of a LUT / equalizer filter from the LUT library 106 is shown based on the sub-pixel position of the cluster / hole center within the pixel. Since the center of the target cluster (blue) falls within a particular sub-pixel 12 of the sub-pixel LUT grid 700, and the particular sub-pixel 12 of the pixel 206 corresponds to a LUT 12 in the LUT library 106, the LUT selector 122 selects the LUT 12 and its coefficients from the LUT library 106 to be applied to the pixels of the pixel patch 300. The element-by-element multiplier 134 then multiplies the coefficients of the LUT 12 element-by-element by the intensity values ​​of the pixels in the pixel patch 300, and the products of the multiplications are summed to produce an output (e.g., a weighted sum 136). This output is used to base call the target cluster 1 (e.g., by feeding the output as an input to the base caller 138).

[0237] When the target cluster is substantially concentric with the center of the pixel, as described above for Figure 7 and Figure 8 As discussed, equalizer 104 implements the following equalization logic:

[0238]

[0239] In the above relationship, the hole center coordinates (m,n) are integers to ensure that the holes are substantially aligned with pixels; p(i,j) is the pixel intensity at position i,j; w(i,j) is the equalizer weight of the pixel at position i,j; i,j are summation limits acting on the pixel range around the hole centered on p(m,n), for example, -4<=i<=4, -4<=j<=4; and the output is the weighted average of the input pixels.

[0240] Fig. 9 A specific implementation is illustrated in which, because no implementation such as Figure 8 In the resampling discussed, the center of the target cluster 1 (blue) is substantially not concentric with the center of the pixel 206. In this specific implementation, interpolation occurs among the selected set of LUTs 124 to produce an interpolated LUT having interpolated coefficients. The interpolated LUT having interpolated coefficients is also referred to herein as a weight kernel 132.

[0241] First, as in Figure 8 As in , the first LUT corresponding to the particular sub-pixel in which the center of target cluster 1 (blue) falls is selected, namely LUT 12. Then, LUT selector 122 selects an additional sub-pixel lookup table from the bank of sub-pixel lookup tables 106, the additional sub-pixel lookup table corresponding to the sub-pixel that is most consecutively adjacent to the particular sub-pixel. Fig. 9 , the nearest consecutive neighboring sub-pixels adjoining the particular sub-pixel 12 are sub-pixels 7, 8, and 13, and therefore LUTs 7, 8, and 13 are selected from the LUT library 106, respectively.

[0242] Fig.10 One specific implementation of interpolating between a selected set of LUTs and generating corresponding LUT weights is depicted. Interpolator 126 is configured with interpolation logic (e.g., linear, bilinear, or bicubic interpolation) that uses the coefficients of the selected LUTs 12, 7, 8, and 13 and generates weights 128 for each of the LUTs 12, 7, 8, and 13.

[0243] Fig.13A , Fig. 13B , Fig. 13C , Fig.13D , Fig.13E and Fig.13F Examples of coefficients for LUTs 12, 7, 8, and 13 are shown. The figures also show examples 1312, 1322, and 1332 of interpolation logic used by interpolator 126 to calculate weights 128 for LUTs 12, 7, 8, and 13. The figures also show examples of weights 128 calculated for LUTs 12, 7, 8, and 13. The figures are snapshots of Excel spreadsheets, and the blue arrows and color coding in the figures are generated by Excel's Trace Priority feature to demonstrate the interpolation logic.

[0244] Fig.11 A weight core generator 130 is shown that generates a weight core 132 using the calculated weights 128 of the LUTs 12 , 7 , 8 , and 13 . Fig.14A One example of a weight kernel 132 is depicted. Fig. 14B and Fig. 14CAn example 1402 of weight kernel generation logic used by the weight kernel generator 130 to generate a weight kernel 132 from the calculated weights 128 of the LUTs 12, 7, 8, and 13 is illustrated. The weight kernel 132 includes interpolated pixel coefficients 1412 configured to mix / combine the intensity values ​​of pixels in the pixel patch 300 in a manner that maximizes the signal-to-noise ratio, the pixels depicting the intensity emission from the target cluster 1 and the intensity emission from the neighboring clusters 2, 3, 4, and 5. The signal for which the signal-to-noise ratio is maximized is the intensity emission from the target cluster, while the noise for which the signal-to-noise ratio is minimized is the intensity emission from the neighboring clusters 2, 3, 4, and 5, i.e., spatial crosstalk, plus some random noise (e.g., to account for background intensity emission). These interpolated pixel coefficients 1412 are used as weights, and the mixing / combining includes performing element-by-element multiplications between the LUT coefficients and the intensity values ​​of the pixels in the pixel patch 300 to calculate a weighted sum of the intensity values ​​of the pixels.

[0245] Fig.12 An element-wise multiplier 134 is shown multiplying the interpolated pixel coefficients 1412 of the weight kernel 132 element-wise with the intensity values ​​of the pixels in the pixel patch 300, and then summing the intermediate products 1202 of the multiplications to produce the weighted sum 136. For each hole, the optical system operates with a point spread function (the response of the optical system) on a point source (cluster intensity in the hole). In some implementations, bias is added to this operation to account for noise caused by different cluster sizes, different background intensities, varying stimulus responses, varying focus, varying sensor sensitivity, and varying lens aberrations. The captured image is a superposition of the responses from all holes. The selected LUT equalizes the system response around each hole to estimate the intensity of the point source from that hole, that is, it processes the PSF intensity on a local neighborhood / grid of sensor pixels to estimate the intensity of the point source that generates the local grid of sensor pixels. The equalizer operation is a dot product of the sensor pixels in the local grid with the equalizer coefficients.

[0246] When the target cluster is not substantially concentric with the center of the central pixel, as discussed above with respect to Fig. 9 , Fig.10 , Fig.11 and Fig.12 As discussed, the equalizer 104 implements the following equalization logic. When the hole is not centered in the pixel, the output of the equalizer 104 is calculated as a function of the virtual pixel intensity p'(i,j) derived from the actual pixel intensity of the pixel of the sequenced image:

[0247] (1)y m,n =∑ i,j p′(m+i,n+j).w(i,j)

[0248] In the above relationship, the hole center coordinates (m,n) may have a fractional part. Each 'virtual' equalizer input p'(i,j) is generated by applying an interpolation filter to a neighborhood of pixels. In one implementation, a windowed sinc low-pass filter h(x,y) is used for interpolation. In other implementations, some other filters may be used, such as a bilinear interpolation filter.

[0249] Use the interpolation filter to calculate the virtual pixel at position (i, j) as follows:

[0250] (2)p′(i,j)=∑ u,v p(u,v).h(iu,jv)

[0251] By combining relations (1) and (2), equalizer 104 uses only the original pixel intensities, as follows:

[0252]

[0253] In the above relationship, given sub-pixel offsets frac(m), frac(n), h is fixed; u,v specify the range of pixels used for interpolation to generate the equalizer input; and i,j specify the range of virtual pixels used as input to the equalizer 104.

[0254] For a given sub-pixel offset, it is the input pixels that change, not the filters or weights. Therefore, for the center of each binned sub-pixel offset, we compute a fixed set of interpolated equalizer coefficients. The output is then:

[0255]

[0256] In the above relationship, h fm,fn Represents the LUT equalizer coefficients for an aperture with binned fractional sub-pixel offsets fm,fn, where (fm,fn) is the LUT index.

[0257] Fig.15A and Fig. 15B It is shown how the interpolated pixel coefficients 1412 of the weight kernel maximize the signal-to-noise ratio and recover the underlying signal of target cluster 1 from the signal corrupted by the crosstalk from clusters 2, 3, 4, and 5.

[0258] The weighted sum 136 is fed as input to a base caller 138 to produce a base call 140. The base caller 138 can be a non-neural network based base caller or a neural network based base caller, examples of which are described in the applications incorporated by reference herein, such as U.S. Patent Application Nos. 62 / 821,766 and 16 / 826,168.

[0259] In yet other implementations, the need for interpolation is eliminated by having large LUTs, where each LUT has a large number of sub-pixel bins (e.g., each LUT has 50, 75, 100, 150, 200, 300, etc. sub-pixel bins).

[0260] Fig.19A A graph representing the base call error rate is shown using an image from a NovaSeq sequencer. The error rate is shown by cycle on the x-axis. 0.004 on the y-axis represents a base call error rate of 0.4%. The error rate is calculated here after the reads are mapped to a Phi-X reference (which is a high confidence ground truth set) and compared to it. The blue line is a traditional base caller. The red line is an improved equalizer-based base caller 104 disclosed herein. At the cost of limited additional computation, the total error rate is reduced by 57%. Due to additional noise in the system (e.g., prephasing / phasing, cluster dimming), the base error rate in subsequent cycles is higher. The performance gain in subsequent cycles increases and this is valuable because it shows that longer reads can be supported. The performance variation between cycles is also significantly reduced.

[0261] Fig.19B Another example of the performance results of the equalizer-based base caller 104 disclosed in the present invention on sequencing data from a NovaSeq sequencer and a Vega sequencer is shown. For the NovaSeq sequencer, the equalizer-based base caller 104 disclosed in the present invention reduces the base call error rate by more than 50%. For the Vega sequencer, the equalizer-based base caller 104 disclosed in the present invention reduces the base call error rate by more than 35%.

[0262] Fig.19C Another example of the performance results of the equalizer-based base caller 104 disclosed in the present invention on sequencing data from a NextSeq 2000 sequencer is shown. For the NextSeq 2000 sequencer, the equalizer-based base caller 104 disclosed in the present invention reduces the base call error rate by 10% on average without affecting the throughput.

[0263] Fig.19D A specific implementation of the computing resources required for the equalizer-based base caller 104 disclosed in the present invention is shown. As shown in the figure, the equalizer-based base caller 104 disclosed in the present invention can be run using a small number of CPU threads ranging from two to seven threads. Therefore, the equalizer-based base caller 104 disclosed in the present invention is a computationally efficient base caller that significantly reduces the base error rate and can therefore be integrated into most existing sequencers without requiring any additional computing resources or dedicated processors such as GPUs, FPGAs, ASICs, etc.

[0264] Dependence of base calling on flow cell tilt—additional implementation details

[0265] In some embodiments, the imager can determine the orientation of the surface plane of the sample being imaged, such as on the X-axis (sometimes referred to as the bevel), Y-axis (sometimes referred to as the tilt), and / or Z-axis (sometimes referred to as the twist). In some embodiments, the imager, such as in combination with an element associated with maintaining and / or moving the circulation pool relative to the imager, can use the determined orientation to reduce the portion of the image that would otherwise be out of focus. The reduction in the out-of-focus portion is performed, for example, by controlling the X-axis orientation, the Y-axis orientation, and / or the Z-axis orientation to increase the portion of the image that is within the DoF of the imager. Orientation control is achieved, for example, by one or more actuators and / or motor drives, depending on various embodiments. Further description is set forth in U.S. Provisional Patent Application No. 63 / 300,531 (Agent Docket No. IP-2205-PRV), entitled “Dynamic Detilt Focus Tracking,” filed on January 18, 2022. Notwithstanding the aforementioned reduction in out-of-focus image portions, some embodiments provide for measurements and / or determinations of tilt, focus, and / or distance that can be used to implement techniques described elsewhere herein to improve base calling accuracy via base calling that relies on flow cell tilt.

[0266] Determine tilt, focus and / or distance

[0267] Depending on the specific implementation, the measurement and / or determination of tilt, focus and / or distance (e.g., relative to a portion of the image being imaged) is performed via various techniques. The circulation pool and / or image area tilt can be determined by measuring the distance between the projection points, using a grid of resolution features and / or purposefully introduced optical aberrations. Multiple tilt values ​​across the circulation pool and / or one or more image areas can be determined via the creation and processing of a graph of surface heights. The image area focus can be determined by the conjugate lens beam spacing. The defocus of the image area can be measured using a multi-point focus tracker.

[0268] The foregoing techniques are described in more detail below.

[0269] Tilt determination based on point spacing measurements

[0270] In some embodiments, the tilt is determined by measuring the spacing between a pair of points projected onto a sample area being imaged. In some embodiments, the tilt is determined by measuring a first spacing between a first pair of points projected onto a sample image area (e.g., an area being imaged and / or to be imaged) and further by measuring a second spacing between a second pair of points projected onto the sample area. One or more pairs of points are projected by a light source. In some embodiments, the tilt determination is performed according to one dimension, and in some other embodiments, the tilt determination is performed according to multiple dimensions (e.g., two dimensions substantially perpendicular to each other). In some embodiments, the first spacing is used to determine a first sample height, the second spacing is used to determine a second sample height, and the first sample height and the second sample height are used to determine the corresponding sample tilt. In some embodiments, a tilt map is determined based on multiple spacing measurements of a pair of points and multiple images at multiple sample locations. Further description is set forth in U.S. Provisional Patent Application No. 63 / 300,531 (Agent Docket No. IP-2205-PRV), entitled “Dynamic Detilt Focus Tracking,” filed on January 18, 2022.

[0271] Focus determination based on conjugate lens beam spacing

[0272] In some embodiments, the degree of focus is determined by providing a pair of incident light beams to a conjugate lens. The conjugate lens directs the incident light beam toward a focus region. The incident light beam is reflected from a sample image region (e.g., a region being imaged and / or to be imaged). The reflected light beam returns to the conjugate lens and propagates through the conjugate lens. The relative spacing between the reflected light beams is measured, and the relative spacing is used to determine the degree of focus relative to the sample, the working distance, and / or the surface profile based on the relative spacing. Further description is set forth in U.S. non-provisional patent No. US8,422,031B2, entitled "Focusing Methods and Optical Systems and Assemblies Using the Same," filed on April 16, 2013.

[0273] Resolution feature grid

[0274] In some implementations, the tilt of the sample is determined by collecting full focal stacks of images of a grid of resolution features (such as a pinhole array and / or a plurality of isolated nanopores included in a flow cell) and analyzing the images to determine the tilt. For example, by performing multiple full focal stacks at different X coordinates and comparing the best focus Z position at each X coordinate, the tilt can be measured as an angle. Additionally or alternatively, by using an autofocus system to detect the Z position of an element (such as a cluster and / or a fiducial) observable with the imager at multiple X positions, the tilt can be measured as an angle. Further description is set forth in U.S. Non-Provisional Patent No. US10,830,700B2, entitled "Solid Inspection Apparatus and Method of Use," filed on March 1, 2019.

[0275] Multi-point focus tracker

[0276] A multi-point focus tracker measures defocus at multiple locations in the image plane. The defocus measurements are processed to determine tilt at the multiple locations.

[0277] Optical aberrations

[0278] Optical aberrations are introduced into the optical system of the imager (eg, using a phase mask) such that the point spread function is asymmetric between overfocus and underfocus, thereby enabling easy discrimination between overfocus and underfocus defocus. These discriminations are processed to determine tilt information.

[0279] Surface plot

[0280] The height of the flow cell is measured at multiple locations and a surface plot is created using the height. The surface plot is processed to determine the tilt at multiple locations.

[0281] Benchmarks and goals

[0282] An example of a fiducial is a distinguishable reference point in or on an object. For example, the reference point exists in an image of the object, in a data set obtained by detecting the object, or in any other representation of the object suitable for expressing information about a reference point relative to the object. The reference point may be specified by an x ​​and / or y coordinate in the plane of the object. Alternatively or additionally, the reference point may be specified by a z coordinate orthogonal to the xy plane, for example, defined by the relative position of the object and the detector. One or more coordinates of the reference point may be specified relative to one or more other features of the object or image or other data set obtained from the object.

[0283] Fig. 20ABenchmark examples are illustrated. The upper portion of the figure is a close-up of a single benchmark with four concentric bull's eye rings. The lower portion of the figure is an image of a block with six example bull's eye ring benchmarks in the image. In various implementations, each of the points in various places represents a corresponding oligonucleotide cluster, a corresponding nanopore of a patterned flow cell, or a corresponding nanopore having one or more oligonucleotide clusters therein. In some implementations, the bull's eye ring benchmark includes a light ring surrounded by a dark border to enhance contrast. The benchmark can be used as a reference point for aligning the imaged block, such as with other images of the same block (e.g., at various wavelengths). For example, the position of the benchmark in the image is determined via cross-correlation with the position of a reference virtual benchmark and determining the position as the position that maximizes the cross-correlation score. In some implementations, cross-correlation is performed using a cross-correlation relationship for discrete functions (see, e.g., Fig. 20C ).

[0284] Fig. 20B Example fiducials in various focus contexts are illustrated. The example fiducials are constructed in the form of a plus sign using a selective chrome layer so that areas with chrome appear dark and areas without chrome appear white. In some contexts, a fiducial implemented using chrome is referred to as an 'Uber Target', a chrome target, or simply a target. Illustrated from top to bottom, a camera (e.g., of an imager) focuses above the chrome layer, focuses at the chrome layer, and focuses below the chrome layer. When focusing at the chrome layer, the edges of the chrome appear sharp. When focusing above or below the chrome layer, the edges of the chrome appear blurred. In some implementations, the chrome target can be used to perform focus characterization.

[0285] According to various implementations as described elsewhere herein, a benchmark (e.g. Fig. 20A and / or Fig. 20B exemplified) can be used as reference image data (e.g., ground truth image data). In some implementations, a scoring relationship (see, e.g., Fig.20D ) to calculate a measure of the goodness of fit between the fiducials in the image and the virtual fiducials. In various implementations, various image alignment operations use a method based on evaluating one or more cross-correlation relationships (e.g., Fig. 20C exemplified) and / or one or more scoring relationships (e.g., Fig.20D In various implementations, various benchmark loss functions use information based on evaluating one or more cross-correlation relationships (e.g., Fig. 20C exemplified) and / or one or more scoring relationships (e.g., Fig.20D In various specific implementations, various baseline quality assessments use information based on evaluating one or more cross-correlation relationships (e.g., Fig. 20C exemplified) and / or one or more scoring relationships (e.g., Fig.20D exemplified).

[0286] Relative to Figure 1AA Example consideration of the dependence of base calling on flow cell tilt Fig. 20B As a first example, the camera is focused above the chrome layer, as in Fig. 20B The image is blurry. Figure 1AA In the training context of , in response to a determination that an image (or portion thereof) is out of focus, the '+base detector' is selected for training. For the out-of-focus context, the '+GT' element is used to train the '+LUT' element. Similarly, in Figure 1AA In the production context of , in response to a determination that an image (or portion thereof) is out of focus, '+base caller' is selected for production. For the out-of-focus context, base calling is performed using the '+LUT' element.

[0287] As a second example, the camera is focused at the chrome layer, as in Fig. 20B The image is clear. Figure 1AA In the training context of , in response to a determination that an image (or portion thereof) is in focus, a '=base caller' is selected for training. For the focus context, a '=GT' element is used to train a '=LUT' element. Similarly, in Figure 1AA In the production context, in response to a determination that an image (or portion thereof) is in focus, a '=base caller' is selected for production. For the focus context, base calling is performed using a '=LUT' element.

[0288] As a third example, the camera is focused below the chrome layer, as in Fig. 20B The image is blurry. Figure 1AA In the training context of , in response to a determination that an image (or portion thereof) is out of focus, the '-base detector' is selected for training. For the out-of-focus context, the '-GT' element is used to train the '-LUT' element. Similarly, in Figure 1AA In the production context of , in response to a determination that an image (or portion thereof) is out of focus, '-base caller' is selected for production. For the out-of-focus context, base calling is performed using the '-LUT' element.

[0289] Fig. 20C An example cross-correlation relation for a discrete function is illustrated. The example cross-correlation relation can be used, for example, using an example scoring relation (see, e.g. Fig.20D ) to determine the position of the reference (see e.g. Fig. 20A ).

[0290] Fig.20DAn example scoring relationship is illustrated. In the example scoring relationship, Minimum_CC is the maximum value of the cross-correlation, Maximum_CC is the maximum value of the cross-correlation, and RunnerUp_CC is the maximum cross-correlation value outside a radius of, for example, four pixels from the position of Maximum_CC.

[0291] Additional Technology

[0292] In various embodiments, one or more techniques related to the dependence of base calling on flow cell tilt are used in a system involving a self-learning base caller. For various examples, a neural network-based and / or non-neural network-based base caller is trained and used to perform base calling using one or more techniques related to the dependence of base calling on flow cell tilt. Additional descriptions are set forth in U.S. Provisional Patent Application No. 63 / 228,954, entitled “Self-Learned Base Caller,” filed on August 3, 2021.

[0293] In various embodiments, one or more sharpening masks are trained based on the flow cell tilt dependency. For example, the training (and subsequent use) of sharpening masks as described in U.S. non-provisional patent application No. 17 / 511,483, entitled "Intensity Extraction with Interpolation and Adaptation for Base Calling," filed on October 26, 2021 (Agent Docket No. ILLM 1053-1 / IP-2214-US) is suitable for training and using a sharpening mask set for each of a plurality of focus contexts. A first sharpening mask set is trained using image information and / or ground truth information associated with overfocus imaging. A second sharpening mask set is trained using image information and / or ground truth information associated with in-focus imaging. A third sharpening mask set is trained using image information and / or ground truth information associated with out-of-focus imaging. After training the three sharpening mask sets, the masks are used during base calling. During base calling, images determined to be out of focus are sharpened using a first set of sharpening masks, images determined to be in focus are sharpened using a second set of sharpening masks, and images determined to be out of focus are sharpened using a third set of sharpening masks.

[0294] Fig.21 An overview of a specific implementation of the RTA pipeline is illustrated. Images are collected in two channels (e.g., a first wavelength corresponding to image 1 and a second wavelength for image 2). Processing of the images is shown, starting with registration, and proceeding by fitting one or more Gaussians to determine the most likely base calls.

[0295] In some implementations, processing associated with image sharpening (e.g., using Laplace masks) is suitable for use with various techniques as described herein with respect to flow cell tilt dependence. For example, one or more Laplace masks are associated with a first tilt condition (e.g., substantially in focus), and one or more other Laplace masks are associated with a second tilt condition (e.g., substantially out of focus). Alternatively, a first one or more Laplace masks are associated with an overfocused tilt condition, a second one or more Laplace masks are associated with an in-focus tilt condition, and a third one or more Laplace masks are associated with an underfocused tilt condition.

[0296] In some implementations, the processing associated with spatially normalizing sub-block intensities is adapted to use various techniques as described herein with respect to flow cell tilt dependence. For example, a first spatial normalization of sub-blocks is selected for use with image regions determined to be overfocused, a second spatial normalization of sub-blocks is selected for use with image regions determined to be in focus, and a third spatial normalization of sub-blocks is selected for use with image regions determined to be underfocused. In some implementations, the spatial normalization is trained based on flow cell tilt, such as in the context of overfocused, in focus, and underfocused images, respectively. In some implementations, the spatial normalization uses techniques such as Figure 1AA The technique of the equalizer is illustrated, which is used in an over-focus base caller ('+base caller'), an in-focus base caller ('=base caller'), and an out-of-focus base caller ('-base caller').

[0297] In some implementations that use expectation maximization during base calling, flow cell tilt dependency is introduced by managing separate independent corresponding statistical models for each focus context. For example, processing associated with image regions that are out of focus is performed using an over-focus statistical model, processing associated with image regions that are in focus is performed using an in-focus statistical model, and processing associated with image regions that are out of focus is performed using an out-of-focus statistical model. Each of these statistical models has the same architecture, but is independently optimized for EM relative to the other statistical models.

[0298] Instead of being based on two focus categories (e.g., in focus / sharp vs. out of focus / blurry) or three focus categories (e.g., overfocus, in focus, and out of focus), some implementations are based on one, four, or more focus categories. For example, an implementation is based on five focus categories: largely overfocus, slightly overfocus, in focus, slightly out of focus, and largely out of focus. Other implementations are based on other numbers of focus categories.

[0299] Some implementations are not based on focus category, but rather on tilt, such as based on the magnitude of the tilt, the direction of the tilt, or both. Figure 1AA and Fig. 1AB An implementation similar to the one depicted. Instead of evaluating the tilt to determine over-focus, in-focus, and under-focus, the tilt assessment determines 'uphill', 'flat', and 'downhill' corresponding to tilt vectors that are upward, horizontal, and downward, respectively, relative to the scanning direction. The corresponding LUT in the corresponding base caller is trained with reference to the corresponding GT. Based on the classification of the tilt, base calling is performed with reference to the corresponding base caller (including the corresponding LUT). For another example, consider an implementation similar to the above, but with multiple upward categories (e.g., largely upward and slightly upward) and multiple downward categories (e.g., largely downward and slightly downward). For yet another example, consider an implementation similar to the above, but with multiple categories related to the tilt direction (e.g., with corresponding equal angular ranges centered at central angles of approximately 0 degrees, 90 degrees, 180 degrees, and 270 degrees relative to the scanning direction). For yet another example, consider an implementation that combines multiple upward and downward categories with categories for the tilt direction.

[0300] Some implementations measure tilt and / or height prior to starting a SBS run. Some implementations selectively measure tilt and / or height multiple times during a SBS run. Some implementations measure tilt and / or height once at the start of a SBS run. Some implementations (e.g., such as some implementations using an area sensor in combination with a flow cell) do not measure either tilt or height during a SBS run. Some implementations that do not measure either tilt or height during a SBS run use methods such as those described above. Figure 1AA and / or Fig. 1AB Describe the technology.

[0301] Some implementations adjust focus and / or tilt prior to starting a SBS run. Some implementations selectively adjust focus and / or tilt multiple times during a SBS run. Some implementations adjust focus and / or tilt once at the start of a SBS run. Some implementations (e.g., such as some implementations using an area sensor in combination with a flow cell) adjust neither focus nor tilt during a SBS run. Some implementations that do not adjust focus nor tilt during a SBS run use methods such as those described above. Figure 1AA and / or Fig. 1AB Describe the technology.

[0302] Compared to the specific implementation that performs focusing and / or tilt adjustment during the sequencing by synthesis operation, some specific implementations that neither perform focusing adjustment nor tilt adjustment during the sequencing by synthesis operation achieve enhanced throughput in some usage scenarios. Consider the first specific example of the specific implementation that neither performs focusing adjustment nor tilt adjustment during the sequencing by synthesis operation. During training, various base callers are selectively trained using images and associated tilt information (e.g., magnitude, direction, or both). The images and associated tilt information are collected without the benefit of focusing and tilt adjustment. The training is selective by using the tilt information associated with the image to select which base caller to train for the image. During production, various (trained) base callers are selectively used to perform base calls from the image and the associated tilt information. Base calls are selective by using the tilt information associated with the image to select which base caller to perform base calls for the image.

[0303] Consider a second specific example of an implementation in which neither focus adjustment nor tilt adjustment is performed during a sequencing-by-synthesis run. The second specific example is similar to the first specific example. However, rather than relying on tilt information to select a base caller, the tilt information is directly used as a parameter for training and production use of one or more base callers. Depending on the specific implementation, the tilt information used is variously the magnitude of the tilt, the direction of the tilt, or both.

[0304] In some specific implementations according to the foregoing specific examples and according to Figure 1AA and / or Fig. 1AB In some implementations of the described techniques, the base caller uses one or more artificial intelligence (AI) techniques, and training is performed in the context of at least some of the AI ​​techniques. Some implementations using one or more AI techniques are referred to as 'deepRTA' implementations. Additional information about DeepRTA can be found in U.S. patent applications 16 / 825,987; 16 / 825,991; 16 / 826,126; 16 / 826,134; 16 / 826,168; 62 / 979,412; 62 / 979,411; 17 / 179,395; 62 / 979,399; 17 / 180,480; 17 / 180,513; 62 / 979,414; 62 / 979,385; and 63 / 072,032.

[0305] In some specific implementations according to the foregoing specific examples and according to Figure 1AA and / or Fig. 1ABIn some implementations of the described techniques, the base callers use techniques other than AI. Some implementations that use one or more techniques other than AI are referred to as 'RTA' implementations.

[0306] In some implementations, for one or more sequencing-by-synthesis cycles of a sequencing-by-synthesis run, the tilt is measured in situ during fluorescence imaging of the sample or before and / or after fluorescence imaging. In some cases (e.g., the tilt of one or more regions of the circulation cell is relatively constant within a block and / or relatively constant over time), the tilt is measured once per sequencing-by-synthesis cycle or within a sequencing-by-synthesis cycle. In some cases (e.g., the tilt of one or more regions of the circulation cell changes over time, such as due to thermal fluctuations), the tilt is measured more than once per sequencing-by-synthesis cycle or within more than one sequencing-by-synthesis cycle.

[0307] Technical improvements and terminology

[0308] In the present application, the terms "cluster", "well", "sample" and "fluorescent sample" are used interchangeably because the well contains the corresponding cluster / sample / fluorescent sample. As defined herein, "sample" and its derivatives are used in their broadest sense and include any specimen, culture, etc. suspected of including a target. In some specific implementations, the sample includes DNA, RNA, PNA, LNA, chimeric or hybrid forms of nucleic acid. The sample may include any biological, clinical, surgical, agricultural, atmospheric or aquatic animal and plant specimens containing one or more nucleic acids. The term also includes any isolated nucleic acid sample, such as genomic DNA, fresh frozen or formalin-fixed paraffin-embedded nucleic acid specimens. It is also envisioned that the sample may come from: a single individual, a collection of nucleic acid samples from genetically related members, a nucleic acid sample from a genetically unrelated member, a nucleic acid sample (matched therewith) from a single individual (such as a tumor sample and a normal tissue sample), or a sample from a single source containing two different forms of genetic material (such as maternal DNA and fetal DNA obtained from a maternal subject), or contaminating bacterial DNA in a sample containing plant or animal DNA. In some implementations, the source of nucleic acid material can include nucleic acid obtained from a newborn, such as nucleic acid typically used in newborn screening.

[0309] Nucleic acid samples may include high molecular weight substances, such as genomic DNA (gDNA). Samples may include low molecular weight substances, such as nucleic acid molecules obtained from FFPE samples or archived DNA samples. In another specific implementation, low molecular weight substances include DNA of enzymatic fragmentation or mechanical fragmentation. Samples may include cell-free circulating DNA. In some specific implementations, the sample may include nucleic acid molecules obtained from biopsy tissue, tumor, scrapings, swabs, blood, mucus, urine, blood plasma, semen, hair, laser capture microdissection, surgical resection and other clinical or laboratory samples. In some specific implementations, the sample may be an epidemiological sample, an agricultural sample, a forensic sample or a pathogenic sample. In some specific implementations, the sample may include nucleic acid molecules obtained from animals such as humans or mammalian sources. In another specific implementation, the sample may include nucleic acid molecules obtained from non-mammalian sources (such as plants, bacteria, viruses or fungi). In some specific implementations, the source of nucleic acid molecules may be archived or extinct samples or species.

[0310] In addition, the methods and compositions disclosed herein can be used to amplify nucleic acid samples with low-quality nucleic acid molecules, such as degraded and / or fragmented genomic DNA from forensic samples. In a specific implementation, forensic samples may include nucleic acids obtained from crime scenes, nucleic acids obtained from missing persons DNA databases, nucleic acids obtained from laboratories associated with forensic investigations, or include forensic samples obtained by law enforcement agencies, one or more military services, or any such personnel. Nucleic acid samples can be purified samples or lysates containing crude DNA, such as from oral swabs, paper, fabrics, or other substrates that can be impregnated with saliva, blood, or other body fluids. Therefore, in some specific implementations, nucleic acid samples may include a small amount of DNA such as genomic DNA, or a fragmented portion of DNA. In some specific implementations, the target sequence may be present in one or more body fluids, wherein the body fluids include but are not limited to blood, sputum, plasma, semen, urine, and serum. In some specific implementations, the target sequence may be obtained from the victim's hair, skin, tissue samples, autopsy, or remains. In some specific implementations, nucleic acids including one or more target sequences may be obtained from dead animals or humans. In some implementations, the target sequence may include a nucleic acid obtained from non-human DNA (such as microbial, plant, or insect DNA). In some implementations, the target sequence or amplified target sequence is related to the purpose of human identification. In some implementations, the disclosure as a whole relates to methods for identifying characteristics of forensic samples. In some implementations, the disclosure as a whole relates to human identification methods using one or more target-specific primers disclosed herein or one or more target-specific primers designed using the primer design criteria outlined herein. In one implementation, a forensic sample or human identification sample comprising at least one target sequence can be amplified using any one or more target-specific primers disclosed herein or using the primer standards outlined herein.

[0311] As used herein, the term "adjacent" when used with respect to two reaction sites means that there are no other reaction sites between the two reaction sites. The term "adjacent" may have a similar meaning when used with respect to adjacent detection paths and adjacent photodetectors (e.g., there are no other photodetectors between adjacent photodetectors). In some cases, a reaction site may not be adjacent to another reaction site, but may still be within the immediate vicinity of the other reaction site. When a fluorescent emission signal from a first reaction site is detected by a photodetector associated with a second reaction site, the first reaction site may be in close proximity to the second reaction site. More specifically, when a photodetector associated with the second reaction site detects, for example, crosstalk from the first reaction site, the first reaction site may be in close proximity to the second reaction site. Adjacent reaction sites may be connected so that they are adjacent to each other; or adjacent sites may be non-connected, with spacing space between them.

[0312] All documents and similar materials cited in this application, including but not limited to patents, patent applications, articles, books, papers and web pages, are expressly incorporated by reference in their entirety, regardless of the format of these documents and similar materials. If one or more of the incorporated documents and similar materials differs from or contradicts this application, including but not limited to defined terms, term usage, described technology, etc., this application shall prevail. Additional information about terms can be found in U.S. non-provisional patent application No. 16 / 826,168 (Agent File No. ILLM 1008-20 / IP-1752-PRV), entitled "Artificial Intelligence-Based Sequencing", filed on March 21, 2020, and U.S. Provisional Patent Application No. 62 / 821,766 (Agent File No. ILLM 1008-9 / IP-1752-PRV), entitled "Artificial Intelligence-Based Sequencing", filed on March 21, 2019.

[0313] The disclosed technology uses neural networks to improve the quality and quantity of nucleic acid sequence information that can be obtained from nucleic acid samples, such as nucleic acid templates or their complementary sequences, e.g., DNA or RNA polynucleotides or other nucleic acid samples. Thus, relative to previously available methods, certain specific implementations of the disclosed technology provide higher throughput polynucleotide sequencing, e.g., higher DNA or RNA sequence data collection rates, higher sequence data collection efficiencies, and / or lower costs for obtaining such sequence data.

[0314] The disclosed techniques use a neural network to identify the centers of solid phase nucleic acid clusters and analyze the optical signals generated during sequencing of such clusters to unambiguously distinguish between adjacent, contiguous, or overlapping clusters so as to assign sequencing signals to a single discrete source cluster. Thus, these and related embodiments allow for the retrieval of meaningful information, such as sequence data, from regions of high density cluster arrays where usable information was previously unavailable due to the confounding effects of overlapping or very closely spaced adjacent clusters, including the effects of overlapping signals emanating therefrom (e.g., as used in nucleic acid sequencing).

[0315] As described in more detail below, in some specific implementations, provide a composition comprising a solid carrier, this solid carrier has one or more nucleic acid clusters as provided herein that are fixed thereon.Each cluster comprises a plurality of immobilized nucleic acids of the same sequence and has a markable center, this markable center has a detectable center mark as provided herein, and the immobilized nucleic acid in the surrounding area in the markable center and the cluster can be distinguished by this detectable center mark.This paper also describes the method for making and using this type of cluster with a markable center.

[0316] The disclosed embodiments will find use in many situations where advantage is gained from the ability to identify, determine, annotate, record, or otherwise assign the location of substantially central locations within a cluster, such as high-throughput nucleic acid sequencing, the development of image analysis algorithms for assigning optical or other signals to discrete source clusters, and other applications in which identifying the centers of immobilized nucleic acid clusters is desirable and beneficial.

[0317] In certain implementations, the present invention contemplates methods involving high-throughput nucleic acid analysis such as nucleic acid sequence determination (e.g., "sequencing"). Exemplary high-throughput nucleic acid analysis includes, but is not limited to, de novo sequencing, resequencing, whole genome sequencing, gene expression analysis, gene expression monitoring, epigenetic analysis, genomic methylation analysis, allele-specific primer extension (APSE), genetic diversity analysis, whole genome polymorphism discovery and analysis, single nucleotide polymorphism analysis, hybridization-based sequence determination methods, etc. Those skilled in the art will appreciate that a variety of different nucleic acids can be analyzed using the methods and compositions of the present invention.

[0318] Although the specific implementation of the present invention is described with respect to nucleic acid sequencing, they are applicable to any field of analyzing image data collected at different time points, spatial positions or other time or physical perspectives. For example, the methods and systems described herein can be used in the field of molecular and cell biology, wherein image data from microarrays, biological specimens, cells, organisms, etc. are collected and analyzed at different time points or perspectives. Images can be obtained using any number of techniques known in the art, including but not limited to fluorescence microscopy, optical microscopy, confocal microscopy, optical imaging, magnetic resonance imaging, tomography, etc. For another example, the methods and systems described herein can be applied, wherein image data obtained by monitoring, air or satellite imaging techniques, etc. are collected and analyzed at different time points or perspectives. The method and system are particularly useful for analyzing images obtained for a visual field, wherein the analytes being observed are kept in the same position relative to each other in the visual field. However, the analyte may have different features in a single image, for example, the analyte may look different in a single image of the visual field. For example, an analyte may appear different in terms of its color detected in different images, a change in signal intensity for a given analyte detected in different images, or even the appearance of a signal for a given analyte detected in one image and the disappearance of a signal for that analyte detected in another image.

[0319] As used herein, the term "analyte" is intended to mean a point or region in a pattern that can be distinguished from other points or regions based on relative position. A single analyte may include one or more specific types of molecules. For example, an analyte may include a single target nucleic acid molecule having a specific sequence, or an analyte may include several nucleic acid molecules having the same sequence (and / or its complementary sequence). Different molecules located at different analytes of a pattern can be distinguished from each other based on the position of the analyte in the pattern. Example analytes include, but are not limited to, holes in a substrate, beads (or other particles) in or on a substrate, a protrusion of a substrate, a ridge on a substrate, a gel material pad on a substrate, or a channel in a substrate.

[0320] Any of a variety of target analytes to be detected, characterized or identified can be used in the devices, systems or methods described herein. Exemplary analytes include, but are not limited to, nucleic acids (e.g., DNA, RNA or their analogs), proteins, polysaccharides, cells, antibodies, epitopes, receptors, ligands, enzymes (e.g., kinases, phosphatases or polymerases), small molecule drug candidates, cells, viruses, organisms, etc.

[0321] The terms "analyte", "nucleic acid", "nucleic acid molecule" and "polynucleotide" are used interchangeably herein. In various embodiments, nucleic acids can be used as templates as provided herein (e.g., nucleic acid templates, or nucleic acid complementary sequences complementary to nucleic acid templates) for specific types of nucleic acid analysis, including but not limited to nucleic acid amplification, nucleic acid expression analysis and / or nucleic acid sequence determination or suitable combinations thereof. In certain embodiments, nucleic acids include, for example, linear polymers of deoxyribonucleotides in 3'-5' phosphodiester or other bonds, such as deoxyribonucleic acid (DNA), such as single-stranded and double-stranded DNA, genomic DNA, copy DNA or complementary DNA (cDNA), recombinant DNA, or any form of synthetic or modified DNA. In other embodiments, nucleic acids include, for example, linear polymers of ribonucleotides in 3'-5' phosphodiester or other bonds, such as ribonucleic acids (RNA), such as single-stranded and double-stranded RNA, messenger (mRNA), copy RNA or complementary RNA (cRNA), alternatively spliced ​​mRNA, ribosomal RNA, small nucleolar RNA (snoRNA), microRNA (miRNA), small interfering RNA (sRNA), piwi RNA (piRNA), or any form of synthetic or modified RNA. The length of the nucleic acids used in the compositions and methods of the present invention can vary, and can be complete or full-length molecules or fragments or smaller portions of larger nucleic acid molecules. In specific embodiments, the nucleic acids can have one or more detectable labels, as described elsewhere herein.

[0322] The terms "analyte", "cluster", "nucleic acid cluster", "nucleic acid population" and "DNA cluster" are used interchangeably and refer to multiple copies of a nucleic acid template and / or its complementary sequence attached to a solid support. In some embodiments, a nucleic acid cluster comprises multiple copies of a template nucleic acid and / or its complementary sequence, which are attached to a solid support via their 5' ends. The copies of the nucleic acid chains that constitute the nucleic acid cluster may be in single-stranded or double-stranded form. The copies of the nucleic acid template present in the cluster may have nucleotides at corresponding positions that are different from each other, for example due to the presence of a labeling moiety. Corresponding positions may also include similar structures having different chemical structures but similar Watson-Crick base pairing properties, such as the case of uracil and thymine.

[0323] Nucleic acid populations may also be referred to as "nucleic acid clusters." Nucleic acid populations may optionally be generated by cluster amplification or bridge amplification techniques, as described in further detail elsewhere herein. Multiple repeats of a target sequence may be present in a single nucleic acid molecule, such as concatemers generated using a rolling circle amplification procedure.

[0324] Depending on the conditions used, the nucleic acid clusters of the present invention can have different shapes, sizes and densities. For example, the cluster can have a substantially circular, polygonal, donut-shaped or annular shape. The diameter of the nucleic acid cluster can be designed to be about 0.2 μm to about 6 μm, about 0.3 μm to about 4 μm, about 0.4 μm to about 3 μm, about 0.5 μm to about 2 μm, about 0.75 μm to about 1.5 μm or any diameter therebetween. In a specific implementation, the diameter of the nucleic acid cluster is about 0.5 μm, about 1 μm, about 1.5 μm, about 2 μm, about 2.5 μm, about 3 μm, about 4 μm, about 5 μm or about 6 μm. The diameter of the nucleic acid cluster can be affected by multiple parameters, including but not limited to the number of amplification cycles performed when generating the cluster, the length of the nucleic acid template or the density of the primers attached to the surface on which the cluster is formed. The density of the nucleic acid cluster can be designed to be usually between 0.1 / mm 2 , 1 / mm 2 , 10 / mm 2 , 100 / mm 2 、1,000 / mm 2 、10,000 / mm 2 Up to 100,000 / mm 2 The present invention also contemplates, in part, higher density nucleic acid clusters, such as 100,000 / mm 2 Up to 1,000,000 / mm 2 and 1,000,000 / mm 2 Up to 10,000,000 / mm 2 .

[0325] As used herein, "analyte" is an area of ​​interest within a specimen or field of view. When used in conjunction with a microarray device or other molecular analysis device, an analyte refers to an area occupied by similar or identical molecules. For example, an analyte can be an amplified oligonucleotide or any other group of polynucleotides or polypeptides having the same or similar sequence. In other embodiments, an analyte can be any element or group of elements that occupies a physical area on a specimen. For example, an analyte can be a piece of land, a body of water, etc. When imaging the analytes, each analyte will have a certain area. Therefore, in many embodiments, an analyte is more than just a pixel.

[0326] The distance between analytes can be described in any number of ways. In some implementations, the distance between analytes can be described from the center of one analyte to the center of another analyte. In other implementations, the distance can be described from the edge of one analyte to the edge of another analyte, or between the outermost identifiable points of each analyte. The edge of the analyte can be described as a theoretical or actual physical boundary on the chip, or a point within the boundary of the analyte. In other implementations, the distance can be described relative to a fixed point on the specimen or a fixed point in an image of the specimen.

[0327] Generally speaking, several specific implementations will be described herein with respect to the analysis methods. It should be understood that systems for performing the methods in an automated or semi-automated manner are also provided. Thus, the present disclosure provides neural network-based template generation and base calling systems, wherein these systems may include: a processor; a storage device; and a program for image analysis, the program including instructions for performing one or more of the methods described herein. Thus, the methods described herein may be performed, for example, on a computer having components described herein or known in the art.

[0328] The method and system shown herein can be used for analyzing any of a variety of objects. Particularly useful objects are solid supports or solid phase surfaces with connected analytes. When used for objects with a repetitive pattern of analytes in the xy plane, the method and system shown herein provide advantages. An example is a microarray with a connection set of cells, viruses, nucleic acids, proteins, antibodies, carbohydrates, small molecules (such as candidate drugs), bioactive molecules or other analytes of interest.

[0329] More and more applications have been developed for arrays of analytes with biomolecules such as nucleic acids and polypeptides. Such microarrays typically include deoxyribonucleic acid (DNA) or ribonucleic acid (RNA) probes. These probes are specific to nucleotide sequences present in humans and other organisms. In certain applications, for example, a single DNA or RNA probe may be attached to a single analyte of the array. A sample such as one from a known human or organism may be exposed to the array so that a target nucleic acid (e.g., a gene fragment, mRNA, or its amplicon) hybridizes with a complementary probe at the corresponding analyte in the array. The probe may be labeled in a target-specific process (e.g., due to a label present on the target nucleic acid or due to an enzyme label of the probe or target present in a hybridized form in the analyte). The array may then be checked by scanning the light of a specific frequency on the analyte to identify which target nucleic acids are present in the sample.

[0330] Biological microarray can be used for genetic sequencing and similar applications. Generally speaking, genetic sequencing includes determining the nucleotide sequence in the length of target nucleic acid (such as the fragment of DNA or RNA). Relatively short sequences are usually sequenced at each analyte, and the sequence information gained can be used in various bioinformatics methods to logically fit the sequence fragments together, thereby reliably determining the wider sequence of the length range of the genetic material derived from it. Developed an automated, computer-based algorithm for characteristic fragments, and recently used it for genome mapping, gene and its functional identification, etc. Microarray is particularly useful for characterizing genome content, because there are a large number of variants, and this replaces the selection scheme of many experiments carried out on single probes and targets. Microarray is an ideal form for carrying out such research in a practical manner.

[0331] Any of the various analyte arrays (also referred to as "microarrays") known in the art can be used in the methods or systems described herein. Typical arrays include analytes, each analyte having a separate probe or probe group. In the latter case, the probe group at each analyte is generally homogeneous, with a single type of probe. For example, with respect to nucleic acid arrays, each analyte may have a plurality of nucleic acid molecules, each of which has a common sequence. However, in some specific implementations, the probe group at each analyte of the array may be heterogeneous. Similarly, a protein array may have an analyte containing a single protein or protein group, which generally but not always has the same amino acid sequence. The probe may be attached to the surface of the array, for example, by covalent bonding of the probe to the surface or by non-covalent interaction of the probe with the surface. In some specific implementations, probes such as nucleic acid molecules may be attached to the surface via a gel layer, as described, for example, in U.S. Patent Application Serial No. 13 / 784,368 and U.S. Patent Application Publication No. 2011 / 0059865A1, each of which is incorporated herein by reference.

[0332] Example arrays include, but are not limited to, BeadChip arrays from Illumina, San Diego, Calif., or other arrays such as those in which probes are attached to beads present on a surface (e.g., beads in wells on a surface), such as U.S. Pat. Nos. 6,266,459; 6,355,431; 6,770,441; 6,859,570; or 7,622,294; or PCT Publication No. WO 00 / 63437, each of which is incorporated herein by reference. Other examples of commercially available microarrays that can be used include, for example, VLSIPS TM Other microarrays synthesized using the (very large scale immobilized polymer synthesis) technique. Spot microarrays can also be used in methods or systems according to some specific implementations of the present disclosure. An example spot microarray is a CodeLink TM Another microarray that can be used is one that uses an inkjet printing method (such as SurePrint from Agilent Technologies). TM Microarrays fabricated using PCR technology.

[0333] Other available arrays include arrays for nucleic acid sequencing applications. For example, arrays with genomic fragment amplicons (commonly referred to as clusters) are particularly useful, such as those described in Bentley et al., Nature 456:53-59 (2008), WO 4 / 018497; WO 91 / 06678; WO 7 / 123744; U.S. Patent No. 7,329,492; No. 7,211,414; No. 7,315,19; No. 7,405,281 or No. 7,057,26; or U.S. Patent Application Publication No. 2008 / 0108082A1, each of which is incorporated herein by reference. Another type of array that can be used for nucleic acid sequencing is an array of particles produced by emulsion PCR technology. Examples are described in Dressman et al., Proc. Natl. Acad. Sci. USA 100:8817-8822 (2003), WO 5 / 010145, U.S. Patent Application Publication No. 2005 / 0130173; or U.S. Patent Application Publication No. 2005 / 0064460, each of which is incorporated herein by reference in its entirety.

[0334] Arrays for nucleic acid sequencing generally have random spatial patterns of nucleic acid analytes. For example, HiSeq or MiSeq sequencing platforms from Illumina (San Diego, Calif.) utilize flow cells on which nucleic acid arrays are formed by random inoculation and then bridge amplification. However, patterned arrays may also be used for nucleic acid sequencing or other analytical applications. Example patterned arrays, methods for making them, and methods for using them are described in U.S. Serial No. 13 / 787,396; U.S. Serial No. 13 / 783,43; U.S. Serial No. 13 / 784,368; U.S. Patent Application Publication No. 2013 / 0116153A1; and U.S. Patent Application Publication No. 2012 / 0316086A1, each of which is incorporated herein by reference. Such patterned array analytes may be used to capture a single nucleic acid template molecule for inoculation, followed by, for example, a homogeneous population formed via bridge amplification. Such patterned arrays may be particularly useful for nucleic acid sequencing applications.

[0335] The size of the analytes on the array (or other objects used in the methods or systems herein) can be selected to suit a particular application. For example, in some implementations, the analytes of the array can have a size that only accommodates a single nucleic acid molecule. A surface having a plurality of analytes within this size range can be used to construct a molecular array, thereby detecting with single molecule resolution. Analytes within this size range can also be used in an array with analytes that each contain a population of nucleic acid molecules. Therefore, the analytes of the array can each have a size of no more than about 1 mm 2 , not more than about 500μm 2 , not more than about 100μm 2 , not more than about 10μm 2 , not more than about 1μm 2 , not more than about 500nm 2 , or not greater than about 100 nm 2 , not more than about 10nm 2 , not more than about 5nm 2 or not greater than about 1 nm 2 Alternatively or additionally, the analytes of the array will be no smaller than about 1 mm 2 , not less than about 500μm 2 , not less than about 100μm 2 , not less than about 10μm 2 , not less than about 1μm 2 , not less than about 500nm 2 , not less than about 100nm 2 , not less than about 10nm 2 , not less than about 5nm 2 or not less than about 1nm 2In practice, the analyte may have a size within a range selected from those upper and lower limits exemplified above. Although several size ranges of analytes on the surface have been exemplified with respect to nucleic acids and nucleic acid scales, it should be understood that analytes within these size ranges can be used for applications that do not include nucleic acids. It should also be understood that the size of the analyte is not necessarily limited to the scales used in nucleic acid applications.

[0336] For specific implementations including objects with multiple analytes (such as, arrays of analytes), the analytes can be discrete and spaced apart from each other. Arrays useful in the present invention can have analytes separated by edge-to-edge distances of up to 100 μm, 50 μm, 10 μm, 5 μm, 1 μm, 0.5 μm or less. Alternatively or additionally, the array can have analytes separated by edge-to-edge distances of at least 0.5 μm, 1 μm, 5 μm, 10 μm, 50 μm, 100 μm or more. These ranges can apply to the average edge-to-edge spacing of the analytes and the minimum or maximum spacing.

[0337] In some implementations, the analytes of the array need not be discrete, but rather adjacent analytes may be adjacent to each other. Whether or not the analytes are discrete, the size of the analytes and / or the spacing of the analytes may vary so that the array may have a desired density. For example, the average analyte spacing in a regular pattern may be at most 100 μm, 50 μm, 10 μm, 5 μm, 1 μm, 0.5 μm or less. Alternatively or additionally, the average analyte spacing in a regular pattern may be at least 0.5 μm, 1 μm, 5 μm, 10 μm, 50 μm, 100 μm or more. These ranges may also be applicable to the maximum or minimum spacing of a regular pattern. For example, the maximum analyte spacing of the regular pattern may be at most 100μm, 50μm, 10μm, 5μm, 1μm, 0.5μm or less; and / or the minimum analyte spacing in the regular pattern may be at least 0.5μm, 1μm, 5μm, 10μm, 50μm, 100μm or more.

[0338] The density of analytes in an array can also be understood in terms of the number of analytes present per unit area. For example, the average analyte density of an array can be at least about 1×10 3 Analytes / mm 2 , 1×10 4 Analytes / mm 2 , 1×10 5 Analytes / mm 2 , 1×10 6 Analytes / mm 2 , 1×10 7 Analytes / mm 2 , 1×10 8 Analytes / mm2 or 1×10 9 Analytes / mm 2 Alternatively or additionally, the average analyte density of the array may be at most about 1×10 9 Analytes / mm 2 , 1×10 8 Analytes / mm 2 , 1×10 7 Analytes / mm 2 , 1×10 6 Analytes / mm 2 , 1×10 5 Analytes / mm 2 , 1×10 4 Analytes / mm 2 or 1×10 3 Analytes / mm 2 or lower.

[0339] The above ranges may apply to all or part of a regular pattern, including, for example, all or part of an analyte array.

[0340] The analytes in the pattern can have any of a variety of shapes. For example, when viewed in a two-dimensional plane, such as on the surface of an array, the analytes may appear to be circular, annular, elliptical, rectangular, square, symmetrical, asymmetrical, triangular, polygonal, etc. The analytes can be arranged in a regular repeating pattern, including, for example, a hexagonal or linear pattern. The pattern can be selected to achieve a desired fill level. For example, a circular analyte is optimally filled in a hexagonal arrangement. Of course, other filling arrangements can also be used for circular analytes, and vice versa.

[0341] The pattern can be characterized by the number of analytes present in a subset of the smallest geometric units that form the pattern. The subset can include, for example, at least about 2, 3, 4, 5, 6, 10, or more analytes. Depending on the size and density of the analytes, a geometric unit can occupy less than 1 mm 2 , 500μm 2 , 100μm 2 , 50μm 2 , 10μm 2 , 1μm 2 , 500nm 2 , 100nm 2 , 50nm 2 , 10nm 2 Alternatively or additionally, the geometric unit may occupy an area greater than 10 nm. 2 , 50nm 2 , 100nm 2 , 500nm 2, 1μm 2 , 10μm 2 , 50μm 2 , 100μm 2 , 500μm 2 , 1mm 2 The features of the analytes in the geometric unit, such as shape, size, spacing, etc., may be selected from those set forth more generally herein for analytes in an array or pattern.

[0342] An array with a regular pattern of analytes may be ordered with respect to the relative positions of the analytes, but random with respect to one or more other features of each analyte. For example, with respect to a nucleic acid array, the nucleic acid analytes may be ordered with respect to their relative positions, but random with respect to one's knowledge of the sequence of the nucleic acid material present at any particular analyte. As a more specific example, a nucleic acid array formed by inoculating a repeating pattern of analytes with template nucleic acids, and amplifying the template at each analyte to form a copy of the template at the analyte (e.g., by cluster amplification or bridge amplification) will have a regular pattern of nucleic acid analytes, but the distribution of nucleic acid sequences across the array will be random. Therefore, typically detecting the presence of nucleic acid material on an array can produce a repeating pattern of analytes, while sequence-specific detection can produce a non-repetitive distribution of signals across the array.

[0343] It should be understood that the descriptions of patterns, sequences, randomness, etc. herein relate not only to analytes on an object, such as an analyte on an array, but also to analytes in an image. Thus, the patterns, sequences, randomness, etc. may be used in any of a variety of formats for storing, manipulating, or communicating image data, including but not limited to computer readable media or computer components, such as a graphical user interface or other output device.

[0344] As used herein, the term "image" is intended to mean a representation of all or part of an object. The representation may be a reproduction of an optical detection. For example, an image may be obtained from fluorescence, luminescence, scattering, or absorption signals. The portion of the object present in the image may be the surface or other xy plane of the object. Typically, an image is a 2-dimensional representation, but in some cases, the information in the image may be derived from 3 or more dimensions. An image need not include optically detected signals. Instead, non-optical signals may be present. An image may be provided in a computer-readable format or medium, such as one or more of those described elsewhere herein.

[0345] As used herein, an "image" is a reproduction or representation of at least a portion of a specimen or other object. In some implementations, the reproduction is an optical reproduction, such as produced by a camera or other optical detector. The reproduction may be a non-optical reproduction, such as a representation of an electrical signal obtained from a nanopore analyte array or a representation of an electrical signal obtained from an ion-sensitive CMOS detector. In a specific implementation, non-optical reproduction may be excluded from the methods or devices described herein. The image may have a resolution capable of distinguishing analytes of specimens present at any of a variety of spacings, including, for example, those spacings that separate less than 100 μm, 50 μm, 10 μm, 5 μm, 1 μm, or 0.5 μm.

[0346] As used herein, the terms "acquiring" or "acquisition" refer to any part of the process of obtaining an image file. In some embodiments, data acquisition may include generating an image of a specimen, searching for a signal in the specimen, instructing a detection device to search for or generate an image of the signal, giving instructions for further analysis or transformation of the image file, and any number of transformations or manipulations of the image file.

[0347] As used herein, the term "template" refers to a representation of the position or relationship between a signal or analyte. Thus, in some implementations, a template is a physical grid having a representation of a signal corresponding to an analyte in a specimen. In some implementations, a template can be a chart, table, text file, or other computer file indicating the position corresponding to the analyte. In the implementation presented herein, a template is generated to track the position of an analyte of a specimen across a set of images of the specimen captured at different reference points. For example, a template can be a set of x,y coordinates or a set of values ​​describing the direction and / or distance of an analyte relative to another analyte.

[0348] As used herein, the term "specimen" may refer to an object or region of an object that captures an image. For example, in a specific implementation of an image taken on the surface of the earth, a piece of land may be a specimen. In other specific implementations of the analysis of biomolecules in a circulation cell, the circulation cell may be divided into any number of sub-portions, each of which may be a specimen. For example, the circulation cell may be divided into various flow channels or channels, and each channel may be further divided into 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 140, 160, 180, 200, 400, 600, 800, 1000 or more separate imaging regions. An example of a circulation cell has 8 channels, wherein each channel is divided into 120 specimens or blocks. In another specific implementation, the specimen may be composed of a plurality of blocks or even the entire circulation cell. Therefore, the image of each specimen may represent an imaging region of a larger surface.

[0349] It should be understood that references to ranges and ordinal lists described herein include not only the enumerated numbers, but also all real numbers between the enumerated numbers.

[0350] As used herein, "reference point" refers to any temporal or physical distinction between images. In some implementations, the reference point is a time point. In other implementations, the reference point is a time point or cycle during a sequencing reaction. However, the term "reference point" can include other aspects that distinguish or separate images, such as angular aspects, rotational aspects, temporal aspects, or other aspects that can distinguish or separate images.

[0351] As used herein, "image subset" refers to an image set within a set. For example, a subset may include 1, 2, 3, 4, 6, 8, 10, 12, 14, 16, 18, 20, 30, 40, 50, 60 images, or any number of images selected from an image set. In a particular implementation, a subset may include no more than 1, 2, 3, 4, 6, 8, 10, 12, 14, 16, 18, 20, 30, 40, 50, 60 images, or any number of images selected from an image set. In some implementations, images are obtained from one or more sequencing cycles, wherein four images are associated with each cycle. Thus, for example, a subset may be a group of 16 images obtained by four cycles.

[0352] Base refers to a nucleotide base or nucleotide, A (adenine), C (cytosine), T (thymine), or G (guanine). "Base" and "nucleotide" are used interchangeably herein.

[0353] The term "chromosome" refers to the hereditary gene carrier of a living cell, which is derived from a chromatin chain containing DNA and protein components (especially histones). The conventional internationally recognized individual human genome chromosome numbering system is used herein.

[0354] The term "site" refers to a unique position on a reference genome (e.g., chromosome ID, chromosome position and orientation). In some implementations, a site can be the position of a residue, sequence tag, or fragment on a sequence. The term "gene locus" can be used to refer to a specific position of a nucleic acid sequence or polymorphism on a reference chromosome.

[0355] The term "sample" herein refers to a sample that is usually derived from a biological fluid, cell, tissue, organ or organism and contains a nucleic acid or a mixture of nucleic acids, which contains at least one nucleic acid sequence to be sequenced and / or phased. Such samples include, but are not limited to, sputum / oral fluid, amniotic fluid, blood, blood fractions, fine needle biopsy samples (e.g., surgical biopsy, fine needle biopsy, etc.), urine, peritoneal fluid, pleural fluid, tissue explants, organ cultures, and any other tissue or cell preparations, or fractions or derivatives thereof, or fractions or derivatives separated therefrom. Although samples are usually taken from human subjects (e.g., patients), samples can be taken from any organism with chromosomes, including but not limited to dogs, cats, horses, goats, sheep, cattle, pigs, etc. The sample can be used directly as obtained from a biological source, or after pretreatment to change the properties of the sample. For example, such pretreatment may include preparing plasma from blood, diluting viscous fluids, etc. Pretreatment methods may also involve, but are not limited to, filtration, precipitation, dilution, distillation, mixing, centrifugation, freezing, lyophilization, concentration, amplification, nucleic acid fragmentation, inactivation of interfering components, addition of reagents, lysis, and the like.

[0356] The term "sequence" includes or represents a nucleotide chain coupled to each other. Nucleotides can be based on DNA or RNA. It should be understood that a sequence may include multiple subsequences. For example, a single sequence (e.g., the sequence of a PCR amplicon) may have 350 nucleotides. The sample read may include multiple subsequences within these 350 nucleotides. For example, the sample read may include a first flanking subsequence and a second flanking subsequence having, for example, 20-50 nucleotides. The first flanking subsequence and the second flanking subsequence may be located on either side of a repeat fragment having a corresponding subsequence (e.g., 40-100 nucleotides). Each flanking subsequence may include a primer subsequence (e.g., 10-30 nucleotides) (or include a portion thereof). For ease of reading, the term "subsequence" will be referred to as a "sequence", but it should be understood that the two sequences are not necessarily separated from each other on a common chain. In order to distinguish the various sequences described herein, different labels (e.g., target sequence, primer sequence, flanking sequence, reference sequence, etc.) may be given to the sequence. Other terms such as "allele" may be given different labels to distinguish similar objects. This application uses "read" and "sequence read" interchangeably.

[0357] The term "double-end sequencing" refers to a sequencing method for sequencing both ends of a target fragment. Double-end sequencing can help detect genomic rearrangements and duplicate fragments, as well as gene fusions and new transcripts. The method for double-end sequencing is described in PCT Publication WO07010252, PCT Application Sequence No. PCTGB2007 / 003798, and U.S. Patent Application Publication US2009 / 0088327, each of which is incorporated herein by reference. In one example, a series of operations can be performed as follows: (a) generating a nucleic acid cluster; (b) linearizing the nucleic acid; (c) hybridizing a first sequencing primer, and performing repeated cycles of extension, scanning, and deblocking as described above; (d) "reversing" the target nucleic acid on the surface of the flow cell by synthesizing a complementary copy; (e) linearizing the resynthesized chain; and (f) hybridizing a second sequencing primer, and performing repeated cycles of extension, scanning, and deblocking as described above. The reversal operation can be performed by delivering the reagents described above for a single cycle of bridge amplification.

[0358] The term "reference genome" or "reference sequence" refers to any specific known genome sequence of any organism that can be used to reference the identified sequence from a subject, whether partial or complete. For example, reference genomes for human subjects and many other organisms can be found at the National Center for Biotechnology Information at ncbi.nlm.nih.gov. "Genome" refers to the complete genetic information of an organism or virus expressed in a nucleic acid sequence. The genome includes both genes and non-coding sequences of DNA. The reference sequence may be larger than the reads aligned therewith. For example, the reference sequence may be at least about 100 times larger, or at least about 1000 times larger, or at least about 10,000 times larger, or at least about 105 times larger, or at least about 106 times larger, or at least about 107 times larger than the aligned reads. In one example, the reference genome sequence is the sequence of the full-length human genome. In another example, the reference genome sequence is limited to a specific human chromosome, such as chromosome 13. In some specific implementations, the reference chromosome is a chromosome sequence from the human genome version hg19. Such sequences may be referred to as chromosome reference sequences, but the term reference genome is intended to encompass such sequences. Other examples of reference sequences include genomes of other species, as well as chromosomes, subchromosome regions (such as chains) of any species, and the like. In various implementations, a reference genome is a consensus sequence or other combination derived from multiple individuals. However, in some applications, a reference sequence may be taken from a specific individual. In other implementations, "genome" also encompasses a so-called "graph genome," which uses a specific storage format and representation of a genomic sequence. In one implementation, a graph genome stores data in a linear file. In another implementation, a graph genome refers to a representation in which alternative sequences (e.g., different copies of chromosomes with small differences) are stored in a graph as different paths. Additional information about the specific implementation of the graph genome can be found at https: / / www.biorxiv.org / content / biorxiv / early / 2018 / 3 / 20 / 194530.full.pdf, the contents of which are hereby incorporated by reference in their entirety.

[0359] The term "read" refers to a collection of sequence data describing a fragment of a nucleotide sample or reference. The term "read" may refer to a sample read and / or a reference read. Typically, although not necessarily, a read represents a short sequence of adjacent base pairs in a sample or reference. A read may be symbolically represented by a base pair sequence (ATCG format) of a sample or reference fragment. The read may be stored in a memory device and processed as appropriate to determine whether the read matches a reference sequence or meets other criteria. The read may be obtained directly from a sequencing device or indirectly from stored sequence information about a sample. In some cases, a read is a DNA sequence of sufficient length (e.g., at least about 25 bp) that can be used to identify a larger sequence or region, for example, it can be aligned and specifically assigned to a chromosome or genomic region or gene.

[0360] Next generation sequencing methods include, for example, sequencing by synthesis (Illumina), pyrophosphate sequencing (454), ion semiconductor technology (Ion Torrent sequencing), single molecule real-time sequencing (Pacific Biosciences), and sequencing by ligation (SOLiD sequencing). Depending on the sequencing method, the length of each read can vary from about 30bp to more than 10,000bp. For example, a DNA sequencing method using a SOLiD sequencer produces a nucleic acid read of about 50bp. For another example, Ion Torrent sequencing produces a nucleic acid read of up to 400bp and 454 pyrophosphate sequencing produces a nucleic acid read of about 700bp. For another example, a single molecule real-time sequencing method can produce a read of 10,000bp to 15,000bp. Therefore, in some specific implementations, the length of the nucleic acid sequence read is 30bp-100bp, 50bp-200bp, or 50bp-400bp.

[0361] The terms "sample read", "sample sequence" or "sample fragment" refer to sequence data of a genomic sequence of interest from a sample. For example, a sample read comprises sequence data from a PCR amplicon having forward and reverse primer sequences. The sequence data can be obtained from any selected sequence method. The sample read can, for example, come from a sequencing by synthesis (SBS) reaction, a sequencing by ligation reaction, or any other suitable sequencing method that requires determination of the length and / or identity of a repetitive element. The sample read can be a common (e.g., average or weighted) sequence derived from multiple sample reads. In some embodiments, providing a reference sequence includes identifying a gene locus of interest based on the primer sequence of the PCR amplicon.

[0362] The term "original fragment" refers to sequence data for a portion of a genomic sequence of interest that at least partially overlaps a specified position or secondary position of interest in a sample read or sample fragment. Non-limiting examples of original fragments include double spliced ​​fragments, single spliced ​​fragments, double unspliced ​​fragments, and single unspliced ​​fragments. The term "original" is used to indicate that the original fragment includes sequence data that has some relationship to the sequence data in the sample read, regardless of whether the original fragment exhibits supporting variations that correspond to and verify or confirm potential variations in the sample read. The term "original fragment" does not mean that the fragment necessarily includes supporting variations that verify the variation call in the sample read. For example, when a variation calling application determines that a sample read exhibits a first variation, the variation calling application may determine that one or more original fragments lack a corresponding type of "supporting" variation that might have been expected to occur given the variation in the sample read.

[0363] The terms "mapping", "aligned", "alignment" or "aligning" refer to the process of comparing a read or a tag to a reference sequence to determine whether the reference sequence contains the read sequence. If the reference sequence contains the read, the read can be mapped to the reference sequence, or in some specific implementations, mapped to a specific position in the reference sequence. In some cases, the alignment simply tells whether the read is a member of a specific reference sequence (i.e., whether the read exists in the reference sequence). For example, the alignment of the read with the reference sequence of human chromosome 13 will tell whether the read exists in the reference sequence of chromosome 13. Tools that provide this information may be referred to as set membership testers. In some cases, the alignment additionally indicates the position in the reference sequence to which the read or tag is mapped. For example, if the reference sequence is a human whole genome sequence, the alignment may indicate that the read exists on chromosome 13, and may also indicate that the read exists on a specific chain and / or site of chromosome 13.

[0364] The term "indel" refers to the insertion and / or deletion of bases in the DNA of an organism. Microindels represent indels that result in a net change of 1 to 50 nucleotides. In the coding region of a genome, frameshift mutations will occur unless the length of the indel is a multiple of 3. Indels can be contrasted with point mutations. Indels insert fragments that insert and delete nucleotides from a sequence, while point mutations are substitutions that replace one of the nucleotides without changing the total number of nucleotides in the DNA. Indels can also be contrasted with tandem base mutations (TBMs), which can be defined as substitutions at adjacent nucleotides (mainly substitutions at two adjacent nucleotides, but substitutions at three adjacent nucleotides have been observed).

[0365] The term "variation" refers to a nucleic acid sequence that is different from a nucleic acid reference. Typical nucleic acid sequence variations include, but are not limited to, single nucleotide polymorphisms (SNPs), short deletions and insertion polymorphisms (Indels), copy number variations (CNVs), microsatellite markers or short tandem repeats and structural variations. Somatic variation calls are the work of identifying variations that are present in DNA samples at low frequencies. Somatic variation calls are interesting in the context of cancer treatment. Cancer is caused by the accumulation of mutations in DNA. DNA samples from tumors are usually heterogeneous, including some normal cells, some cells in the early stages of cancer progression (with fewer mutations) and some late cells (with more mutations). Due to this heterogeneity, when sequencing a tumor (e.g., from FFPE samples), somatic mutations will usually occur at low frequencies. For example, SNVs can be seen in only 10% of the reads covering a given base. Variations to be classified as somatic or germline by a variation classifier are also referred to herein as "variations to be tested."

[0366] The term "noise" refers to erroneous variant calls caused by one or more errors in the sequencing process and / or variant calling application.

[0367] The term "variant frequency" refers to the relative frequency of an allele (genetic variant) at a specific locus in a population, expressed as a fraction or percentage. For example, the fraction or percentage can be the fraction of all chromosomes in a population that carry the allele. By way of example, the sample variant frequency represents the relative frequency of the allele / variant at a specific locus / position along the genomic sequence of interest relative to the "population", where the population corresponds to the number of reads and / or samples of the genomic sequence of interest obtained from an individual. As another example, the baseline variant frequency represents the relative frequency of the allele / variant at a specific locus / position along one or more baseline genomic sequences, where the "population" corresponds to the number of reads and / or samples of one or more baseline genomic sequences obtained from a population of normal individuals.

[0368] The term "variant allele frequency (VAF)" refers to the observed percentage of sequencing reads matching a variant divided by the overall coverage at the target position. VAF is a measure of the proportion of sequencing reads that carry a variant.

[0369] The terms "position", "designated position" and "gene locus" refer to the position or coordinates of one or more nucleotides within a nucleotide sequence. The terms "position", "designated position" and "gene locus" also refer to the position or coordinates of one or more base pairs in a nucleotide sequence.

[0370] The term "haplotype" refers to a combination of alleles inherited together at adjacent sites on a chromosome. A haplotype can be a locus, multiple loci, or an entire chromosome, depending on the number, if any, of recombination events that occur between a given set of loci.

[0371] The term "threshold" refers herein to a digital or non-digital value used as a cutoff value for characterizing a sample, nucleic acid, or a portion thereof (e.g., a read). The threshold may be changed based on empirical analysis. The threshold may be compared to a measured value or a calculated value to determine whether the source generating such a value should be classified in a particular manner. The threshold may be identified based on experience or analysis. The choice of the threshold depends on the confidence level that the user wishes to have to classify. The threshold may be selected for a specific purpose (e.g., to balance sensitivity and selectivity). As used herein, the term "threshold" indicates a point at which the analysis process may be changed and / or a point at which an action may be triggered. The threshold does not need to be a predetermined number. Instead, the threshold may be, for example, a function based on multiple factors. The threshold may be adjusted as appropriate. In addition, the threshold may indicate an upper limit, a lower limit, or a range between limits.

[0372] In some implementations, a metric or score based on sequencing data may be compared to a threshold. As used herein, the term "metric" or "score" may include a value or result determined by sequencing data, or may include a function based on a value or result determined by sequencing data. Like a threshold, a metric or score may be adjusted as appropriate. For example, a metric or score may be a normalized value. As an example of a score or metric, one or more implementations may use a count score when analyzing data. The count score may be based on the number of sample reads. The sample reads may have undergone one or more filtering stages so that the sample reads have at least one common feature or quality. For example, each sample read used to determine the count score may have been aligned with a reference sequence, or may be assigned as a potential allele. The number of sample reads with a common feature may be counted to determine the read count. The count score may be based on the read count. In some implementations, the count score may be a value equal to the read count. In other implementations, the count score may be based on the read count and other information. For example, the count score may be based on the read count of a specific allele of a genetic locus and the total number of reads of the locus. In some implementations, the count score may be based on the read counts of the genetic loci and previously obtained data. In some implementations, the count score may be a normalized score between predetermined values. The count score may also be a function of the read counts of other loci from the sample or a function of the read counts of other samples run in parallel with the sample of interest. For example, the count score may be a function of the read counts of a specific allele and the read counts of other loci in the sample and / or the read counts from other samples. For example, read counts from other loci and / or read counts from other samples may be used to normalize the count score of a specific allele.

[0373] The term "coverage" or "fragment coverage" refers to a count or other measure of multiple sample reads of the same fragment of a sequence. The read count may represent a count of the number of reads covering the corresponding fragment. Alternatively, coverage may be determined by multiplying the read count by a specified factor based on historical knowledge, sample knowledge, locus knowledge, etc.

[0374] The term "read depth" (usually a number followed by an "×") refers to the number of sequence reads with overlapping alignments at a target position. This is usually expressed as an average or percentage exceeding a cutoff value within a collection of intervals (such as exons, genes, or groups). For example, a clinical report might say that the group average coverage is 1,105×, with 98% of target bases covered >100×.

[0375] The term "base call quality score" or "Q score" refers to a PHRED scaled probability in the range of 0-50 that is inversely proportional to the probability that a single sequenced base is correct. For example, a T base call with a Q of 20 is considered very likely to be correct, with a probability of 99.99%. Any base call with a Q < 20 should be considered low quality, and any variant identified where a significant proportion of the sequenced reads supporting the variant are of low quality should be considered a potential false positive.

[0376] The term "variant read" or "number of variant reads" refers to the number of sequencing reads that support the presence of a variant.

[0377] Regarding "strand type" (or DNA strand type), the genetic information in DNA can be represented as a string of letters A, G, C, and T. For example, 5'-AGGACA-3'. Usually, the sequence is written in the direction shown here, that is, the 5' end is to the left and the 3' end is to the right. DNA can sometimes appear as a single-stranded molecule (such as in some viruses), but usually we find DNA as a double-stranded unit. It has a double helix structure, which has two antiparallel strands. In this case, the word "antiparallel" means that the two strands run parallel, but with opposite polarity. Double-stranded DNA is held together by pairing between bases, and the pairing is always such that adenine (A) pairs with thymine (T) and cytosine (C) pairs with guanine (G). This pairing is called complementarity, and one strand of DNA is called the complementary sequence of the other strand. Therefore, double-stranded DNA can be represented as two strings, like this: 5'-AGGACA-3' and 3'-TCCTGT-5'. Note that the two strands have opposite polarity. Thus, the strand types of the two DNA strands can be referred to as the reference strand and its complement, the forward strand and the reverse strand, the top strand and the bottom strand, the sense strand and the antisense strand, or the Watson strand and the Crick strand.

[0378] Read alignment (also called read mapping) is the process of finding the location of a sequence in the genome. Once aligned, the "mapping quality" or "mapping quality score (MAPQ)" of a given read quantifies the probability that its location on the genome is correct. Mapping quality is encoded on a phred scale, where P is the probability that the alignment is incorrect. The probability is calculated as: P = 10 (-MAQ / 10), where MAPQ is the mapping quality. For example, a mapping quality of 40 = 10 to the -4th power, which means that there is a 0.01% chance that the read is incorrectly aligned. Therefore, the mapping quality is associated with several alignment factors, such as the base quality of the read, the complexity of the reference genome, and the double-end information. Regarding the first factor, if the base quality of the read is low, it means that the observed sequence may be erroneous, and therefore its alignment is erroneous. Regarding the second factor, mappability refers to the complexity of the genome. Repeated regions are more difficult to map, and reads that fall into these regions usually obtain low mapping quality. In this case, MAPQ reflects the fact that the reads are not uniquely aligned and their true origin cannot be determined. Regarding the third factor, in the case of double-end sequencing data, consistent pairs are more likely to be well aligned. The higher the mapping quality, the better the alignment. Reads aligned with good mapping quality usually mean that the read sequence is good and are aligned with almost no mismatches in high mappability regions. MAPQ values ​​can be used as quality control for alignment results. The proportion of aligned reads with a MAPQ higher than 20 was typically used for downstream analysis.

[0379] As used herein, "signal" refers to a detectable event, such as, for example, an emission in an image, such as a light emission. Thus, in some implementations, a signal may represent any detectable light emission (i.e., a "spot") captured in an image. Thus, as used herein, a "signal" may refer to actual emission from an analyte of a specimen, and may refer to spurious emission that is not associated with the actual analyte. Thus, a signal may be generated by noise, and may subsequently be discarded as not representing the actual analyte of the specimen.

[0380] As used herein, the term "clusters" refers to a group of signals. In a specific implementation, the signals are derived from different analytes. In some implementations, a signal cluster is a group of signals gathered together. In other implementations, a signal cluster represents a physical area covered by an amplified oligonucleotide. Each signal cluster should be observed as several signals (one per template cycle, and possibly more due to crosstalk) in an ideal situation. Therefore, in the case of including two (or more) signals in a template from the same signal cluster, a repeated signal is detected.

[0381] As used herein, terms such as "minimum," "maximum," "minimize," "maximize," and grammatical variations thereof may include values ​​that are not absolute maximum or minimum values. In some implementations, these values ​​include near-maximum values ​​and near-minimum values. In other implementations, these values ​​may include local maximum values ​​and / or local minimum values. In some implementations, these values ​​include only absolute maximum or minimum values.

[0382] As used herein, "crosstalk" refers to a signal detected in one image that is also detected in a separate image. In some implementations, crosstalk may occur when an emitted signal is detected in two separate detection channels. For example, in the case where an emission signal appears in one color, the emission spectrum of the signal may overlap with another emission signal of another color. In some implementations, fluorescent molecules for indicating the presence of nucleotide bases A, C, G, and T are detected in separate channels. However, because the emission spectra of A and C overlap, some of the C color signals may be detected during detection using the A color channel. Therefore, the crosstalk between the A signal and the C signal allows the signal from one color image to appear in another color image. In some implementations, G and T crosstalk. In some implementations, the amount of crosstalk between channels is asymmetric. It should be understood that the amount of crosstalk between channels can be controlled by (among other things) selecting signal molecules with appropriate emission spectra and selecting the size and wavelength range of the detection channel.

[0383] As used herein, "register," "registering," "registration," and similar terms refer to any process of associating a signal from an image or data set at a first point in time or viewing angle with a signal from an image or data set at another point in time or viewing angle. For example, registration can be used to align signals from a set of images to form a template. As another example, registration can be used to align signals from other images with a template. One signal may be registered directly or indirectly to another signal. For example, a signal from image "S" may be directly registered to image "G." As another example, a signal from image "N" may be directly registered to image "G," or alternatively, a signal from image "N" may be registered to image "S," which has been previously registered to image "G." Therefore, the signal from image "N" is indirectly registered to image "G."

[0384] As used herein, the term "reference" is intended to mean a distinguishable reference point in or on an object. The reference point can be, for example, a mark, a second object, a shape, an edge, an area, an irregularity, a channel, a pit, a column, etc. The reference point may be present in an image of the object or in another data set obtained from the detected object. The reference point may be specified by an x ​​and / or y coordinate in the plane of the object. Alternatively or additionally, the reference point may be specified by a z coordinate orthogonal to the xy plane, for example defined by the relative position of the object and the detector. One or more coordinates of the reference point may be specified relative to one or more other analytes of the object or image or other data sets obtained from the object.

[0385] As used herein, the term "optical signal" is intended to include, for example, a fluorescent signal, a luminescent signal, a scattered signal, or an absorbed signal. The optical signal may be detected in the ultraviolet (UV) range (about 200 nanometers to 390 nanometers), the visible (VIS) range (about 391 nanometers to 770 nanometers), the infrared (IR) range (about 0.771 micrometers to 25 micrometers), or other ranges of the electromagnetic spectrum. The optical signal may be detected in a manner that excludes all or part of one or more of these ranges.

[0386] As used herein, the term "signal level" is intended to mean the amount or number of detected energy or encoded information having a desired or predefined characteristic. For example, an optical signal may be quantified by one or more of intensity, wavelength, energy, frequency, power, brightness, etc. Other signals may be quantified based on characteristics such as voltage, current, electric field strength, magnetic field strength, frequency, power, temperature, etc. Absence of signal is understood to be a signal level of zero or a signal level that is not significantly different from noise.

[0387] As used herein, the term "simulation" is intended to mean creating a representation or model of a physical thing or action that predicts the characteristics of the thing or action. In many cases, a representation or model can be distinguished from a thing or action. For example, a representation or model can be distinguished from a thing in terms of one or more features such as color, signal strength detected from all or part of a thing, size, or shape. In a specific implementation, a representation or model can be idealized, amplified, dimmed, or incomplete when compared to a thing or action. Therefore, in some implementations, for example, in terms of at least one of the above-mentioned features, the representation of a model can be distinguished from the thing or action it represents. The representation or model can be provided in a computer-readable format or medium, such as one or more of those described elsewhere herein.

[0388] As used herein, the term "specific signal" is intended to mean detected energy or encoded information that is selectively observed relative to other energy or information (such as background energy or information). For example, a specific signal can be a light signal detected at a specific intensity, wavelength, or color; an electrical signal detected at a specific frequency, power, or field strength; or other signals related to spectral and analytical detection known in the art.

[0389] As used herein, the term "scanning swath" is intended to mean a rectangular portion of an object. A scan swath may be an elongated strip that is scanned by relative movement between the object and the detector in a direction parallel to the longest dimension of the strip. Generally, the width of the rectangular portion or strip will be constant along its entire length. Multiple scan swaths of an object may be parallel to each other. Multiple scan swaths of an object may be adjacent to each other, overlap each other, abut each other, or be separated from each other by gap regions.

[0390] As used herein, the term "variance" is intended to refer to the difference between an expected value and an observed value or the difference between two or more observed values. For example, the variance can be the difference between an expected value and a measured value. The variance can be represented using a statistical function, such as the standard deviation, the square of the standard deviation, the coefficient of variation, etc.

[0391] As used herein, the term "xy coordinates" is intended to mean information that specifies a position, size, shape, and / or orientation in an xy plane. The information may be, for example, numerical coordinates in a Cartesian system. The coordinates may be provided relative to one or both of the x-axis and the y-axis, or may be provided relative to another position in the xy plane. For example, the coordinates of an analyte of an object may specify the position of the analyte relative to a reference or other analyte of the object.

[0392] As used herein, the term "xy plane" is intended to mean a 2-dimensional region defined by linear axes x and y. When used with reference to a detector and an object viewed by the detector, the region may be further specified as being orthogonal to the viewing direction between the detector and the detected object.

[0393] As used herein, the term "z-coordinate" is intended to mean information specifying the position of a point, line, or area along an axis orthogonal to the xy plane. In a particular implementation, the z-axis is orthogonal to the area of ​​the object observed by the detector. For example, the focus direction of an optical system may be specified along the z-axis.

[0394] In some specific implementations, affine transformation is used to transform the collected signal data. In some such specific implementations, template generation utilizes the fact that the affine transformation between color channels is consistent between runs. Due to this consistency, when determining the coordinates of the analyte in the specimen, a default offset set can be used. For example, a default offset file may include relative transformations (movement, scaling, skew) of different channels relative to a channel (such as the A channel). However, in other specific implementations, the offset between color channels drifts during operation and / or between operations, making it difficult to generate templates driven by offsets. In such specific implementations, the method and system provided herein can be generated using no offset templates, which will be further described below.

[0395] In some aspects of the above-mentioned specific implementation, the system may include a circulation cell. In some aspects, the circulation cell includes a channel or other configuration of a block, wherein at least some of the blocks include one or more analyte arrays. In some aspects, the analyte comprises a plurality of molecules such as nucleic acids. In some aspects, the circulation cell is configured to deliver the nucleotide bases of the label to the nucleic acid array, thereby extending the primers hybridized with the nucleic acid in the analyte, so as to generate a signal corresponding to the analyte comprising nucleic acid. In some specific implementations, the nucleic acids in the analyte are identical or substantially identical to each other.

[0396] In some image analysis systems described herein, each image in the image set includes a color signal, wherein different colors correspond to different nucleotide bases. In some aspects, each image in the image set includes a signal having a single color selected from at least four different colors. In some aspects, each image in the image set includes a signal having a single color selected from four different colors. In some systems described herein, nucleic acids can be sequenced by providing four different labeled nucleotide bases to a molecular array, thereby generating four different images, each image including a signal having a single color, wherein the signal color is different for each of the four different images, thereby generating a cycle of four color images corresponding to four possible nucleotides present at a specific position in the nucleic acid. In some aspects, the system includes a circulation cell, which is configured to deliver additional labeled nucleotide bases to the molecular array, thereby generating multiple cycles of color images.

[0397] In some specific implementations, the method provided herein may include determining whether the processor is actively collecting data or whether the processor is in a low activity state. Collecting and storing a large number of high-quality images generally requires a large amount of storage capacity. Additionally, once collected and stored, the analysis of image data may become resource-intensive and may interfere with the processing power of other functions, such as the collection and storage of additional image data in progress. Therefore, as used herein, the term low activity state refers to the processing power of the processor at a given time. In some specific implementations, the low activity state occurs when the processor is not collecting and / or storing data. In some specific implementations, the low activity state occurs when some data collection and / or storage is performed, but the additional processing power remains unchanged so that image analysis can occur simultaneously without interfering with other functions.

[0398] As used herein, "identifying a conflict" means identifying a situation where multiple processes are competing for a resource. In some such implementations, one process is given a higher priority than another process. In some implementations, the conflict may involve the need to prioritize time allocations, processing power, storage capacity, or any other resource that is prioritized. Thus, in some implementations, when processing time or power is to be distributed between two processes (such as analyzing a data set and acquiring and / or storing a data set), there is a conflict between the two processes, and the conflict may be resolved by prioritizing one of the processes.

[0399] Also provided herein is a system for performing image analysis. The system may include a processor; storage capacity; and a program for image analysis, the program including instructions for processing a first data set for storage and a second data set for analysis, wherein the processing includes acquiring and / or storing the first data set on a storage device, and analyzing the second data set when the processor does not acquire the first data set. In some aspects, the program includes instructions for the following operations: identifying at least one instance of a conflict between acquiring and / or storing the first data set and analyzing the second data set; and resolving the conflict in favor of acquiring and / or storing image data, so that acquiring and / or storing the first data set is given priority. In some aspects, the first data set includes an image file obtained from an optical imaging device. In some aspects, the system also includes an optical imaging device. In some aspects, the optical imaging device includes a light source and a detection device.

[0400] As used herein, the term "program" refers to instructions or commands that perform a task or process. The term "program" may be used interchangeably with the term module. In some implementations, a program may be a compilation of various instructions that execute under the same command set. In other implementations, a program may refer to discrete batches or files.

[0401] The following describes some surprising effects of the methods and systems for performing image analysis using the methods and systems described herein. In some sequencing implementations, an important measure of the effectiveness of a sequencing system is its overall efficiency. For example, the amount of mappable data generated every day and the total cost of installing and running the instrument are important aspects of an economic sequencing solution. In order to reduce the time of generating mappable data and improve the efficiency of the system, real-time base detection can be enabled on the instrument computer, and can be run synchronously with sequencing chemistry and imaging. This allows many data processing and analysis to be completed before the sequencing chemistry ends. Additionally, it can reduce the storage required for intermediate data and limit the amount of data that needs to be transmitted across the network.

[0402] Although the sequence output has increased, the data per run transmitted from the system provided herein to the network and the auxiliary analysis and processing hardware has been significantly reduced. By transforming the data on the instrument computer (acquisition computer), the network load is significantly reduced. Without these onboard, offline data reduction technologies, the image output of a group of DNA sequencing instruments would paralyze most networks.

[0403] The widespread adoption of high-throughput DNA sequencing instruments is due in part to ease of use, support for a wide range of applications, and suitability for almost any laboratory environment. The efficient algorithms presented herein allow important analytical capabilities to be added to a simple workstation that can control a sequencing instrument. This reduction in computing hardware requirements has several practical benefits that will become even more important as the level of sequencing output continues to increase. For example, by performing image analysis and base calling on a simple tower, heat generation, laboratory floor space, and power consumption are kept to a minimum. In contrast, other commercial sequencing technologies have recently expanded their computing infrastructure for primary analysis, with processing power up to more than five times, resulting in a corresponding increase in heat output and power consumption. Therefore, in some specific implementations, the computational efficiency of the method and system provided herein enables customers to increase their sequencing throughput while keeping server hardware costs to a minimum.

[0404] Thus, in some implementations, the methods and / or systems presented herein act as a state machine that tracks the individual state of each specimen, and when it detects that the specimen is ready to advance to the next state, it performs appropriate processing and advances the specimen to that state. A more detailed example of how the state machine monitors the file system to determine when the specimen is ready to advance to the next state, according to a particular implementation, is set forth in Example 1 below.

[0405] In some specific implementations, the method and system provided herein are multithreaded and can work with a configurable number of threads. Therefore, for example, in the case of nucleic acid sequencing, the method and system provided herein can work in the background during the real-time sequencing operation of real-time analysis, or it can use a pre-existing image data set to run for offline analysis. In some specific implementations, the method and system handle multithreading by giving each thread a subset of the specimen that it is responsible for. This minimizes the possibility of thread contention.

[0406] The method of the present disclosure may include the step of obtaining a target image of an object using a detection device, wherein the image includes a repeated pattern of an analyte on the object. A detection device capable of high-resolution imaging of a surface is particularly useful. In a specific implementation, the detection device will have sufficient resolution to distinguish analytes with the density, spacing and / or analyte size set forth herein. A detection device capable of obtaining an image or image data from a surface is particularly useful. An example detector is one that is configured to keep an object and a detector in a static relationship when obtaining an area image. A scanning device may also be used. For example, a device (e.g., a so-called "step-by-step shooting" detector) that obtains sequential area images may be used. It is also useful to continuously scan points or lines on the surface of an object to accumulate data to construct a device for a surface image. A point scanning detector may be configured to scan points (i.e., a small detection area) on the surface of an object via a grating motion in the xy plane of the surface. A line scanning detector may be configured to scan a line along the y dimension of the surface of the object, the longest dimension of the line occurring along the x dimension. It should be understood that a movable detection device, an object, or both may be used to achieve scanning detection. Detection devices particularly useful, for example, in nucleic acid sequencing applications, are described in U.S. Patent Application Publication Nos. 2012 / 0270305A1; 2013 / 0023422A1; and 2013 / 0260372A1; and U.S. Patent Nos. 5,528,050; 5,719,391; 8,158,926 and 8,241,573, each of which is incorporated herein by reference.

[0407] The specific implementation disclosed herein can be implemented as a method, device, system or product using programming or engineering techniques to produce software, firmware, hardware or any combination thereof. As used herein, the term "product" refers to a code or logic implemented in hardware or computer-readable media such as optical storage devices and volatile or non-volatile memory devices. Such hardware may include, but is not limited to, field programmable gate arrays (FPGAs), coarse-grained reconfigurable architectures (CGRAs), application-specific integrated circuits (ASICs), complex programmable logic devices (CPLDs), programmable logic arrays (PLAs), microprocessors or other similar processing devices. In a specific implementation, the information or algorithms set forth herein are present in non-transient storage media.

[0408] In a specific implementation, the computer-implemented method described herein can occur in real time when multiple images of an object are obtained. Such real-time analysis is particularly useful for nucleic acid sequencing applications, in which nucleic acid arrays are subjected to repeated cycles of fluids and detection steps. The analysis of sequencing data may generally be computationally intensive, so that it may be beneficial to perform the method described herein in real time or in the background while performing other data acquisition or analysis algorithms. Example real-time analysis methods that can be used with the inventive method are those methods for MiSeq and HiSeq sequencing devices available from Illumina (San Diego, Calif.) and / or described in U.S. Patent Application Publication No. 2012 / 0020537A1, which is incorporated herein by reference.

[0409] An example data analysis system formed by one or more programmed computers, wherein the programming is stored on one or more machine-readable media, wherein the code is executed to perform one or more steps of the method described herein. In one specific implementation, for example, the system includes an interface designed to allow the system to be networked to one or more detection systems (e.g., optical imaging systems) configured to collect data from a target object. The interface can receive and adjust data in appropriate places. In a specific implementation, the detection system will output digital image data, for example, image data representing individual picture elements or pixels, which together form an image of an array or other object. The processor processes the received detection data according to one or more routines defined by the processing code. The processing code can be stored in various types of memory circuits.

[0410] According to the specific implementation currently contemplated, the processing code performed on the detection data includes a data analysis routine designed to analyze the detection data to determine the locations and metadata of individual analytes visible or encoded in the data, as well as locations where no analytes were detected (i.e., locations where no analyte was present, or locations where no meaningful signal was detected from an existing analyte). In a specific implementation, analyte locations in the array will generally appear brighter than non-analyte locations due to the presence of fluorescent dyes attached to the imaged analytes. It should be understood that, for example, when the target of the probe at the analyte is not present in the array being detected, the analyte need not appear brighter than its surrounding area. The color in which the individual analytes appear may depend on the dye employed and the wavelength of light used by the imaging system for imaging purposes. Analytes to which the target is not bound or otherwise lacks a specific marker may be identified based on other features, such as their expected location in the microarray.

[0411] Once the data analysis routine has located individual analytes in the data, value assignment can be performed. Generally speaking, value assignment will be based on the characteristics of the data represented by the detector assembly (e.g., pixel) at the corresponding position, and digital values ​​will be assigned to each analyte. That is, for example, when processing imaging data, the value assignment routine can be designed to identify the light of a specific color or wavelength detected at a specific position, as indicated by the pixel group or cluster at the position. For example, in typical DNA imaging applications, four common nucleotides will be represented by four independent and distinguishable colors. Then, the value corresponding to the nucleotide can be assigned to each color.

[0412] As used herein, the terms "module," "system," or "system controller" may include hardware and / or software systems and circuits that operate to perform one or more functions. For example, a module, system, or system controller may include a computer processor, controller, or other logic-based device that performs operations based on instructions stored on a tangible and non-transitory computer-readable storage medium such as a computer memory. Alternatively, a module, system, or system controller may include a hard-wired device that performs operations based on hard-wired logic and circuits. The modules, systems, or system controllers shown in the accompanying drawings may represent hardware and circuits that operate based on software or hard-wired instructions, software that directs hardware to perform operations, or a combination thereof. A module, system, or system controller may include or represent a hardware circuit or circuit that includes one or more processors and / or is connected to the one or more processors, such as one or more computer microprocessors.

[0413] As used herein, the terms "software" and "firmware" are interchangeable and include any computer program stored in memory for execution by a computer, including RAM memory, ROM memory, EPROM memory, EEPROM memory, and non-volatile RAM (NVRAM) memory. The above memory types are only examples and thus do not limit the types of memory that can be used to store computer programs.

[0414] In the field of molecular biology, one of the methods used for nucleic acid sequencing is sequencing by synthesis. This technology can be applied to large-scale parallel sequencing projects. For example, by using an automated platform, tens of thousands of sequencing reactions can be performed simultaneously. Therefore, one embodiment of the present invention relates to an instrument and method for collecting, storing and analyzing image data generated during nucleic acid sequencing.

[0415] The huge gain of the amount of data that can be collected and stored makes the simplified image analysis method even more beneficial. For example, the image analysis method described herein allows designers and end users to effectively utilize existing computer hardware. Therefore, in the face of rapidly increasing data output, this paper presents a method and system that reduces the computational burden of processing data. For example, in the field of DNA sequencing, output has been scaled 15 times in the last year, and hundreds of billions of bases can be reached in a single run of a DNA sequencing device now. If the computing infrastructure requires proportional growth, most researchers will still be unable to achieve large genome-level experiments. Therefore, generating more original sequence data will increase the needs for secondary analysis and data storage, making the optimization of data transmission and storage extremely valuable. Some specific implementations of the method and system presented herein can reduce the time, hardware, network and laboratory infrastructure requirements required for generating available sequence data.

[0416] The present disclosure describes various methods and systems for performing these methods. Some examples of the methods are described as a series of steps. However, it should be understood that the specific implementation is not limited to the specific steps and / or step order described herein. Steps can be omitted, steps can be modified, and / or other steps can be added. In addition, the steps described herein can be combined, steps can be performed simultaneously, steps can be performed in parallel, steps can be divided into multiple sub-steps, steps can be performed in different orders, or steps (or a series of steps) can be re-executed in an iterative manner. In addition, although different methods are described herein, it should be understood that these different methods (or steps of these different methods) can be combined in other specific implementations.

[0417] In some implementations, a processing unit, processor, module, or computing system that is "configured to" perform a task or operation may be understood to be specifically constructed to perform the task or operation (e.g., such that one or more programs or instructions stored thereon or used in conjunction therewith are customized or intended to perform the task or operation, and / or such that the arrangement of processing circuitry is customized or intended to perform the task or operation). For purposes of clarity and avoidance of doubt, a general purpose computer (which, if appropriately programmed, may be "configured to" perform a task or operation) is not "configured to" perform a task or operation unless or until specifically programmed or structurally modified to perform the task or operation.

[0418] In addition, the operations of the methods described herein may be sufficiently complex that they cannot be practically performed by a person of ordinary skill or ordinary skill in the art within a commercially reasonable period of time. For example, the methods may rely on relatively complex calculations that make it impossible for a person to complete the methods within a commercially reasonable time.

[0419] Throughout this application, various publications, patents or patent applications have been cited. The disclosures of these publications are hereby incorporated by reference into this application in their entirety in order to more fully describe the prior art to which this invention pertains.

[0420] The term "comprising" is intended herein to be open ended and include not only the listed elements but also any additional elements.

[0421] As used herein, when used in reference to a collection of items, the term "each" is intended to identify a single item in the collection, but does not necessarily refer to every item in the collection. Exceptions may occur if explicitly disclosed or the context clearly dictates otherwise.

[0422] Although the invention has been described with reference to the examples provided above, it should be appreciated that various modifications can be made without departing from the invention.

[0423] Module in the present application can be realized in hardware or software, and need not be divided into identical frame accurately as shown in the figure.Some of these modules can also be realized on different processors or computers, or expand between multiple different processors or computers.In addition, it should be understood that, under the situation that does not affect the function realized, can be combined, synchronous operation or with some of the sequence operation modules different from those shown in the figure.As used herein, term "module" can include "submodule", and "submodule" itself can be considered as constituting module in this article.The frame that is assigned as module in the figure can also be considered as the flow chart step in the method.

[0424] As used herein, "identification" of an information item does not necessarily require direct designation of that information item. Information may be "identified" in a field by simply referencing the actual information using one or more levels of indirection, or by identifying one or more different information items that together are sufficient to determine the actual information item. Furthermore, the term "designate" is used herein to mean the same as "identify."

[0425] As used herein, a given signal, event, or value is "dependent on" a predecessor signal, an event or value of the predecessor signal, an event or value affected by a given signal, event, or value. If there are intervening processing elements, steps, or time periods, a given signal, event, or value may still be "dependent on" a predecessor signal, event, or value. If an intervening processing element or step combines more than one signal, event, or value, the signal output of the processing element or step is considered to be "dependent on" each of the signal, event, or value inputs. If a given signal, event, or value is the same as a predecessor signal, event, or value, this is merely a degenerate case in which a given signal, event, or value is still considered to be "dependent on" or "depending on" or "based on" a predecessor signal, event, or value. The "responsiveness" of a given signal, event, or value to another signal, event, or value is defined in a similar manner.

[0426] As used herein, "parallel" or "synchronous" does not require exact simultaneity. It is sufficient if the evaluation of one of the individuals begins before the evaluation of another of the individuals is completed.

[0427] Computer Systems

[0428] Fig.17A is a block diagram of an example computer system. The computer system includes a storage subsystem, a user interface input device, a CPU, a network interface, a user interface output device, and an optional deep learning processor (for simplicity, illustrated as a GPU, FPGA, CGRA) interconnected by a bus subsystem. The storage system includes a memory subsystem and a file storage subsystem. The memory subsystem includes a random access read / write memory (RAM) and a read-only memory (ROM). The ROM and file storage subsystem elements include, for example, non-transient computer-readable media capabilities for storing and executing programming instructions to implement all or any part of the RTA functions described elsewhere herein. According to various specific implementations, the deep learning processor can implement all or any part of the RTA functions described elsewhere herein. In various specific implementations, the deep learning processor element includes various combinations of CPU, GPU, FPGA, CGRAs, ASIC, ASIP, and DSP.

[0429] In general, the computer system 1700 can be used to implement the technology disclosed in the present invention. More specifically, the computer system 1700 includes at least one central processing unit (CPU) 1772 that can communicate with multiple peripheral devices via a bus subsystem 1755. The peripheral devices variously include a storage subsystem 1710, a user interface input device 1738, a user interface output device 1776, and a network interface subsystem 1774, which storage subsystem includes, for example, a memory device and a file storage subsystem 1736. The input and output devices enable user interaction with the computer system 1700. The network interface subsystem 1774 provides an interface to an external network, including an interface to corresponding interface devices in other computer systems.

[0430] User interface input devices 1738 include variously: keyboards; pointing devices, such as mice, trackballs, touch pads, and / or graphics tablets; scanners; touch screens incorporated into displays; audio input devices, such as voice recognition systems and microphones; and other types of input devices. In general, use of the term "input device" is intended to include all possible types of devices and ways of inputting information into computer system 1700.

[0431] User interface output devices 1776 variously include a display subsystem, a printer, fax capabilities, and / or a non-visual display (such as an audio output device). The display subsystem variously includes an LED display, a cathode ray tube (CRT), a flat panel device such as a liquid crystal display (LCD), a projection device, or some other mechanism for producing a visible image. The display subsystem optionally provides a non-visual display, such as an audio output device. In general, the use of the term "output device" is intended to include all possible types of devices and ways to output information from the computer system 1700 to a user or to another machine or computer system.

[0432] Storage subsystem 1710 can store software modules that include programming structures and data structures that provide the functionality of some or all of the techniques described herein. The software modules are typically executed by processor 1778.

[0433] Processor 1778 variously includes any combination of a graphics processing unit (GPU), a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), and / or a coarse-grained reconfigurable architecture (CGRA). Processor 1778 can be provided by a deep learning cloud platform such as Google Cloud Platform TM 、Xilinx TM and Cirrascale TM Managed. Examples of processors 1778 include Google's Tensor Processing Unit (TPU) TM , Rackmount solutions (such as GX4 Rackmount Series TM 、GX17 Rackmount Series TM ), NVIDIA DGX-1 TM 、Microsoft's Stratix V FPGA TM , Graphcore’s Intelligent Processor Unit (IPU) TM 、Qualcomm's Snapdragon processors TM Zeroth Platform TM 、NVIDIA's Volta TM 、NVIDIA's DRIVE PX TM , NVIDIA's JETSON TX1 / TX2 MODULE TM 、Intel's Nirvana TM 、Movidius VPU TM,Fujitsu DPI TM 、ARM's DynamicIQ TM 、IBM TrueNorth TM , with Testa V100s TM Lambda GPU servers, and more.

[0434] The memory subsystem 1722 of the storage subsystem 1710 variously includes multiple memories, including a main random access memory (RAM) 1732 for storing instructions and data during program execution and a read-only memory (ROM) 1734 for storing fixed information such as instructions and constants. The file storage subsystem 1736 can provide persistent storage for program files and data files, and variously includes a hard disk drive, a floppy disk drive and associated removable media, a CD-ROM drive, an optical drive, and / or a removable media tape disk. Modules that implement the functionality of certain specific implementations are variously stored by the file storage subsystem 1736 in the storage subsystem 1710, or stored in other machines accessible to the processor.

[0435] The bus subsystem 1755 enables communications between the various components and subsystems of the computer system 1700. Although the bus subsystem 1755 is schematically illustrated as a single bus, alternative implementations of the bus subsystem use multiple busses.

[0436] Computer system 1700 itself can be of different types depending on the specific implementation, including a personal computer, a portable computer, a workstation, a computer terminal, a network computer, a television, a mainframe, a server farm, a widely distributed collection of loosely networked computers, or any other data processing system or user device. Due to the ever-changing nature of computers and networks, the Fig.17A The depicted description of computer system 1700 is intended only as a specific example for illustrating various specific implementations of the present invention. Many other configurations of computer system 1700 are possible, with similar Fig.17A The computer systems depicted may have additional or fewer components than those shown.

[0437] In various implementations, the equalizer base caller 104 may be communicatively connected to the storage subsystem 1710 and / or the user interface input device 1738 .

[0438] In various implementations, one or more of the laboratory instruments and / or production instruments described elsewhere herein include one or more computer systems that are the same or similar to the example computer systems of the figure. In various implementations, any one or more of the training context and / or production context uses any one or more computer systems that are the same or similar to the example computer systems of the figure to perform RTA-related processing, such as operating as one or more servers associated with training data collection and / or processing and production data collection and / or processing.

[0439] In various specific implementations, the memory subsystem and / or the file storage subsystem can store information associated with the RTA, such as all or any portion of information associated with the GT and / or LUT elements of the various equalizers and / or base callers described elsewhere herein or associated with the various equalizers and / or base callers. For example, all or any portion of the stored information corresponds to variously initialization information for the equalizer used in the training context, trained information for the equalizer used in the training context, and / or any combination of trained information for the equalizer used in the production context. For another example, all or any portion of the stored information corresponds to one or more intermediate representations, such as those associated with information provided by the training context to the production context, as illustrated and described elsewhere herein.

[0440] Fig. 17B Illustrated are training and production elements for various aspects of base detection that rely on the tilt of the circulation cell. The upper part of the figure illustrates one or more training contexts, and the lower part illustrates one or more production contexts. Each training context in the training context includes one or more training data collection / processing capabilities, and each training data collection / processing capability has one or more corresponding training servers. Each training server in the training server can store corresponding training data, such as information generated by training performed via one or more RTA related activities. In some specific implementations, all or any part of a context in the training context corresponds to a laboratory instrument. Each production context in the production context includes one or more production instruments. Each production instrument in the production instrument can store production data.

[0441] In various implementations, the memory subsystem and / or the file storage subsystem can store image and tilt data and representations thereof, such as pixel intensities of one or more regions of an image and / or information that enables tilt determination. In various implementations, the computer system can process images in real time, including extracting the intensity of particular pixels in real time. In some implementations based on real-time pixel intensity extraction, all or any portion of the image data corresponding to the extracted region is not exclusively stored in the file storage subsystem. In various implementations, the computer system can process tilt measurements and / or information related to the determination of tilt measurements in real time. In various implementations, the computer system can process images and / or tilt information in real time to enable real-time base calling.

[0442] The training context of this figure represents various training contexts exemplified and described elsewhere herein. The production context of this figure represents various production contexts exemplified and described elsewhere herein. The training context uses collected and / or synthesized training data to train one or more RTA-related elements, such as an equalizer and / or a LUT associated with or included in an equalizer. Then, as shown by the dotted arrow 'deploy trained information', the results of the training are provided to the production context for us to, for example, provide tilt-dependent base calls.

[0443] As a first specific example, Fig. 17B A training context in the training context corresponds to Figure 1AA training context, and Fig. 17B The corresponding one or more production contexts in the production context correspond to Figure 1AA One or more instances of a production context. Fig. 17B The term 'deployment of trained information' corresponds to moving information from Fig. 17B Any one or more of the LUTs of the training context are provided to Fig. 17B The corresponding LUT of the production context is prepared for tilt-dependent production base calling.

[0444] As a second specific example, Fig. 17B A training context in the training context corresponds to Figure 1A system 100A, and Fig. 17B A production context in the production context corresponds to when it is used in production (e.g., performing base calling using information determined by training and stored in LUT 106). Figure 1A system 100A.

[0445] In some implementations, the same server is used in both the training context and the production context. Figure 1AAOne or more servers of the training context are also used to implement Figure 1AA The production context.

[0446] Specific implementation

[0447] The technology disclosed in the present invention uses an image processing technique based on equalization to attenuate spatial crosstalk from sensor pixels. The technology disclosed in the present invention can be practiced as a system, method, or article. One or more features of a specific implementation can be combined with a basic specific implementation. Specific implementations that are not mutually exclusive are taught to be combinable. One or more features of a specific implementation can be combined with other specific implementations. The present disclosure periodically reminds the user of these options. The omission of the statement repeating these options from some specific implementations should not be regarded as limiting the combinations taught in the preceding sections, and these statements will be incorporated by reference into each of the following specific implementations.

[0448] In one specific implementation, the disclosed technology provides a computer-implemented method for attenuating spatial crosstalk from sensor pixels.

[0449] The disclosed technology solves the spatial crosstalk on sensor pixels in the pixel plane caused by the periodically distributed fluorescent samples in the sample plane. The signal cones from the fluorescent samples are optically coupled to the local grid of sensor pixels through at least one lens. The signal cones overlap and impinge on the sensor pixels, thereby generating spatial crosstalk.

[0450] The disclosed technique captures in at least one sub-pixel lookup table the characteristic extension of a characteristic signal cone projected through a lens and the resulting contribution of the characteristic signal cone to fluorescence detected by sensor pixels in a local grid of sensor pixels that is substantially concentric with the center of the characteristic signal cone.

[0451] The disclosed technique interpolates between a set of sub-pixel lookup tables that represent feature extensions at sub-pixel resolution to generate an interpolated lookup table based on the center of a target fluorescent sample.

[0452] The disclosed technique separates the signal from a target fluorescence sample by convolving an interpolated lookup table with sensor pixels in a target local grid of sensor pixels to project the center of the signal cone onto the approximate center of the target local grid of sensor pixels.

[0453] The disclosed technique uses the sum of the convolution contributions of the separated signals as the fluorescence intensity from the target fluorescent sample.

[0454] The technology disclosed in the present invention then uses the fluorescence intensity to perform base calling on the first target fluorescent sample. The fluorescence intensity of the first target fluorescent sample is determined for each imaging channel in a plurality of imaging channels. Consider a four-channel chemistry that uses four imaging channels to generate four images in each sequencing cycle. Then, for the first target fluorescent sample, four fluorescence intensities are measured and determined as described above using the technology disclosed in the present invention. These four fluorescence intensities are then processed by a base caller to perform base calling on the first target fluorescent sample. Similarly, for a two-channel chemistry, two fluorescence intensities are used to perform base calling on the first target fluorescent sample.

[0455] The methods described in this and other parts of the disclosed technology may include one or more of the following features and / or features described in conjunction with the additional methods disclosed herein. For the sake of brevity, the combinations of features disclosed in this application are not listed separately and are not repeated with each basic feature set. The reader will understand how the features identified in the method can be easily combined with the basic feature sets identified as specific implementations in other parts of the application.

[0456] In some implementations, the periodically distributed fluorescent samples are arranged in a diamond shape. In other implementations, the periodically distributed fluorescent samples are arranged in a hexagonal shape.

[0457] Other implementations of the methods described in this section may include a non-transitory computer-readable storage medium storing instructions that are executable by a processor to perform any of the above methods. Yet another implementation of the methods described in this section may include a system that includes a memory and one or more processors that are operable to execute instructions stored in the memory to perform any of the above methods.

[0458] In another specific implementation, the technology disclosed in the present invention provides a computer-implemented method for base calling.

[0459] The disclosed technique accesses an image whose pixels depict intensity emissions from a target cluster and intensity emissions from additional adjacent clusters. These pixels include a center pixel that contains the center of the target cluster. Each of these pixels can be divided into multiple sub-pixels.

[0460] According to a particular sub-pixel, among a plurality of sub-pixels of a central pixel including a center pixel of a target cluster center, the disclosed technology selects a sub-pixel lookup table corresponding to the particular sub-pixel from a sub-pixel lookup table library. The selected sub-pixel lookup table contains pixel coefficients configured to accept intensity emissions from the target cluster and reject intensity emissions from neighboring clusters.

[0461] The disclosed technique multiplies pixel coefficients element-wise by intensity values ​​of pixels in an image and sums the products of the multiplications to produce an output.

[0462] The disclosed technology uses this output to perform base calling on the target cluster.

[0463] Each feature discussed in this particular implementation section for other implementations is also applicable to this method implementation. As indicated above, all method features are not repeated here and should be considered repeated by reference.

[0464] In some specific implementations, the technology disclosed in the present invention also includes: (i) selecting additional sub-pixel lookup tables from a sub-pixel lookup table library, which additional sub-pixel lookup tables correspond to sub-pixels that are most continuously adjacent to a specific sub-pixel, (ii) interpolating between pixel coefficients of the selected sub-pixel lookup table and pixel coefficients of the selected additional sub-pixel lookup table, and generating interpolated pixel coefficients, which are configured to accept intensity emissions from a target cluster and reject intensity emissions from adjacent clusters, (iii) multiplying the interpolated pixel coefficients element-by-element by the intensity values ​​of pixels in the image, and summing the products of the multiplications to produce an output, and (iv) using the output to perform base detection on the target cluster.

[0465] In some specific implementations, the target cluster and the additional adjacent clusters are periodically distributed on the flow cell in a diamond shape and fixed on the holes of the flow cell. In other specific implementations, the target cluster and the additional adjacent clusters are periodically distributed on the flow cell in a hexagonal shape and fixed on the holes of the flow cell.

[0466] In some implementations, the interpolation is based on at least one of linear interpolation, bilinear interpolation, and bicubic interpolation.

[0467] In some implementations, pixel coefficients of sub-pixel lookup tables in a sub-pixel lookup table library are learned as a result of training an equalizer using decision-directed equalization. In one implementation, decision-directed equalization uses least squares estimation as a loss function. In one implementation, least squares estimation uses ground truth base calls to minimize squared error. In one implementation, ground truth base calls are modified to account for DC offset, amplification factor, and degree of polyclonality.

[0468] In some implementations, pixel coefficients of a sub-pixel lookup table in the sub-pixel lookup table bank are obtained from a combination of: (i) a single sub-pixel lookup table whose pixel coefficients are learned as a result of training an equalizer using decision-directed equalization, and (iii) a pre-computed interpolation filter set. Each interpolation filter in the interpolation filter set corresponds to each sub-pixel in the plurality of sub-pixels.

[0469] The technology disclosed in the present invention also includes making the center of the target cluster substantially concentric with the center of the central pixel by: (i) aligning the image relative to the template image and determining affine transformation parameters and nonlinear transformation parameters, (ii) using these parameters to transform the position coordinates of the target cluster and additional adjacent clusters into image coordinates of the image, and generating a transformed image with transformed pixels, and (iii) using the transformed position coordinates of the target cluster and additional adjacent clusters to apply interpolation so that their respective cluster centers are substantially concentric with the center of the corresponding transformed pixel containing the cluster center.

[0470] The disclosed technology also includes generating an output for each of a plurality of images captured using a corresponding imaging channel in a particular sequencing cycle, and using the output generated for each image to perform base calling on a target cluster.

[0471] Other implementations of the methods described in this section may include a non-transitory computer-readable storage medium storing instructions that are executable by a processor to perform any of the above methods. Yet another implementation of the methods described in this section may include a system that includes a memory and one or more processors that are operable to execute instructions stored in the memory to perform any of the above methods.

[0472] Although the present invention is disclosed by referring to the specific implementation and examples described in detail above, it should be understood that these examples are intended to be illustrative rather than restrictive. It is contemplated that modifications and combinations will be easily conceived by those skilled in the art, and these modifications and combinations will be within the spirit of the present invention and the scope of the following claims.

Claims

1. A method for selectively performing base calling on in-focus elements and out-of-focus elements of an image collected during sequencing, the method comprising: capturing an image of a portion of the flow cell using a sensor having a depth of field (DoF); classifying corresponding elements of the image as in focus or out of focus with respect to a spatial relationship to the DoF based at least in part on one or more tilt measurements of the flow cell portion; selecting one or more respective base callers for each of the in-focus class and the out-of-focus class based at least in part on a scalar tilt measurement; as well as Base calling of the image is performed using each of the one or more selected base callers. 2 . The method of claim 1 , wherein the out-of-focus categories include an over-focus category for out-of-focus elements above the DoF and an under-focus category for out-of-focus elements below the DoF.

3. The method of claim 2, wherein selecting one or more base callers for the focus category comprises: Select a base caller appropriate for processing the in-focus image.

4. The method of claim 2, wherein selecting one or more base callers for the out-of-focus category comprises: A base caller is selected that is suitable for processing the out-of-focus image of at least a portion of the corresponding element in the out-of-focus category.

5. The method of claim 2, wherein selecting one or more base callers for the out-of-focus category comprises: A base caller is selected that is suitable for processing the out-of-focus image of at least a portion of the corresponding element in the out-of-focus class.

6. The method of claim 1, wherein the image is collected during sequencing by synthesis.

7. The method of claim 1 , wherein each of the one or more base callers for the defocus class comprises an equalizer adapted to apply defocus correction based on a set of trained coefficients in each of a plurality of lookup tables (LUTs).

8. The method of claim 7, wherein at least one equalizer of the one or more base callers is trained at least in part using ground truth (GT) corresponding to an out-of-focus context.

9. The method of claim 7, wherein at least one equalizer of the one or more base callers is trained at least in part using ground truth (GT) corresponding to an over-focus context.

10. A method of selectively performing equalizer-based base calling on in-focus elements and out-of-focus elements of an image collected during sequencing, the method comprising: capturing an image of a portion of the flow cell using a sensor having a depth of field (DoF); classifying corresponding elements of the image as in focus or out of focus with respect to a spatial relationship to the DoF based at least in part on one or more tilt measurements of the flow cell portion; selecting one or more corresponding equalizer-based base callers for each of the in-focus class and the out-of-focus class based at least in part on a scalar tilt measurement; as well as performing base calling of the image using the one or more selected equalizer-based base callers, wherein performing base calling comprises: Apply the set of coefficients in the corresponding LUT to the intensity values ​​of the corresponding set of image pixels of the target base, determining a weighted sum of intensity values ​​of pixels of the image based on application of the set of coefficients, and The weighted sum is output as the base call prediction. The method according to claim 10 , wherein the out-of-focus category comprises an over-focus category and an under-focus category.

12. The method of claim 11, wherein selecting one or more base callers for the focus category comprises: Select a base caller appropriate for processing the in-focus image.

13. The method of claim 11, wherein selecting one or more base callers for the overfocus category comprises: Select a base caller that is appropriate for handling out-of-focus images.

14. The method of claim 11, wherein selecting one or more base callers for the out-of-focus category comprises: Select an equalizer-based base caller that is suitable for handling out-of-focus images.

15. The method of claim 10, wherein the image is collected during sequencing by synthesis.

16. A method of training an equalizer to perform equalizer-based base calling on in-focus elements and out-of-focus elements of an image collected during sequencing, the method comprising: obtaining a training data set of images of respective portions of one or more flow cells, wherein each image includes a known out-of-focus region based on a tilt measurement of the respective flow cell portion; Inputting the training data set into the equalizer; Causing the equalizer to perform base calling on known out-of-focus regions in the training data set, wherein performing base calling comprises: applying a set of coefficients in a corresponding LUT of the equalizer to the intensity values ​​of a corresponding set of image pixels of the target base, determining a weighted sum of intensity values ​​of pixels of the image based on application of the set of coefficients, calculating an error in the weighted sum based on an intensity target determined for the target base, using a derivative of the error to update the set of coefficients with a value that reduces the error, The equalizer is repeatedly caused to perform base calling on known out-of-focus regions until the updated set of coefficients no longer reduces the error.

17. The method of claim 16, wherein the equalizer is trained in a training context to perform equalizer-based base calling, and wherein the method further comprises: The updated set of coefficients is derived for use in an equalizer in a production context.

18. The method of claim 16, wherein the equalizer is trained in the production context.

19. A system for selectively performing base calling on in-focus elements and out-of-focus elements of an image collected during sequencing, the method comprising: at least one processor; and at least one memory system having computer-executable instructions stored thereon that, when executed by the processor, cause the processor to: receiving an image of a portion of the flow cell captured using a sensor having a depth of field (DoF); determining a classification of a corresponding element of the image as being in focus or out of focus with respect to a spatial relationship to the DoF based at least in part on one or more tilt measurements of the flow cell portion; selecting one or more respective base callers for each of the in-focus class and the out-of-focus class based at least in part on a scalar tilt measurement; as well as Each of the one or more selected base callers is caused to selectively perform base calling on corresponding in-focus and out-of-focus elements of the image.

20. The method of claim 19, wherein the out-of-focus categories include an over-focus category for out-of-focus elements above the DoF and an under-focus category for out-of-focus elements below the DoF.

21. The method of claim 20, wherein the computer executable instructions, when executed by the processor, cause the processor to select one or more base callers for the in-focus category, wherein the base callers are suitable for processing in-focus images.

22. The method of claim 20, wherein the computer executable instructions, when executed by the processor, cause the processor to select one or more base callers for the out-of-focus category, wherein the base callers are adapted to process out-of-focus images.

23. The method of claim 20, wherein the computer executable instructions, when executed by the processor, cause the processor to select one or more base callers for the out-of-focus category, wherein the base callers are adapted to handle out-of-focus images.

24. The method of claim 19, wherein the image of a portion of a flow cell is collected during sequencing by synthesis.

25. The method of claim 19, wherein each of the one or more base callers for the defocus class comprises an equalizer adapted to apply defocus correction based on a set of trained coefficients in each of a plurality of lookup tables (LUTs).

26. The method of claim 25, wherein at least one equalizer of the one or more base callers is trained at least in part using ground truth (GT) corresponding to an out-of-focus context.

27. The method of claim 25, wherein at least one equalizer of the one or more base callers is trained at least in part using ground truth (GT) corresponding to an out-of-focus context.

Images