Systems and methods for detecting disease subtypes

Abstract

Systems and methods for detecting a subtype of a disease state and for determining the development of a resistance mechanism in a disease are disclosed. A method may include receiving, at an input component of the system, a set of sequence reads associated with a nucleic acid sample; generating, using a processor of the system and by analyzing the set of sequence reads, methylation data; and analyzing, using the processor, the methylation data to identify a subtype of the disease state. Another method may include obtaining methylation data from a targeted methylation sequencing assay, applying the methylation data to a trained machine learning model, and receiving an output indicative of whether MRD is present in a test subject and / or whether a disease develops a resistance mechanism. Other aspects are described and claimed.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of priority to U.S. Provisional Patent Application No. 63 / 370,014 filed on August 1, 2022, U.S. Provisional Patent Application No. 63 / 382,664 filed on November 7, 2022, and U.S. Provisional Patent Application No. 63 / 385,590 filed on November 30, 2022, each of which is hereby incorporated by reference in its entirety. Technical Field

[0002] The present disclosure relates generally to the field of model-based characterization and classifiers for predicting disease states and subtypes from nucleic acid samples. Background Art

[0003] It has been observed that deoxyribonucleic acid (DNA) methylation plays an important role in regulating gene expression. Abnormal DNA methylation is always related to many disease processes (including cancer). Research and technological progress have led to the development of new technologies for detecting various disease states at early stages. For example, DNA methylation spectra analysis using methylation sequencing (e.g., whole genome bisulfite sequencing (WGBS)) is increasingly considered to be an important diagnostic tool for detecting, diagnosing and / or monitoring cancer. More specifically, the specific pattern and / or allele-specific methylation pattern of differentially methylated regions can be used as a molecular marker for non-invasive diagnosis using circulating free (cf) DNA. However, this area still needs to be used to analyze methylation sequencing data from free DNA to detect, diagnose and / or monitor improved methods for diseases (such as cancer).

[0004] The present disclosure is intended to address one or more of these aforementioned challenges. The background description provided herein is intended to generally introduce the context of the present disclosure. Unless otherwise indicated herein, the materials described in this section are not prior art to the claims of this application and will not be identified as prior art or inspiration of prior art by inclusion in this section. Summary of the invention

[0005] In summary, one aspect provides a method for using a system to detect subtypes of a disease state, the method comprising: receiving a set of sequence reads associated with a nucleic acid sample at an input component of the system; using a processor of the system and generating methylation data by analyzing the set of sequence reads; and using the processor to analyze the methylation data to identify subtypes of the disease state.

[0006] On the other hand, a method for training a machine learning model to detect the development of resistance mechanisms in a cancer being treated is provided, the method comprising: obtaining a set of training data from a source, wherein the training data comprises methylation data derived from a targeted methylation sequencing assay; after the obtaining, annotating the set of training data by assigning a histological label to each training data in the set; applying the set of annotated training data to the machine learning model; and optimizing the pattern recognition capability of an algorithm associated with the machine learning model based on the application.

[0007] On the other hand, a method for detecting the development of resistance mechanisms in a cancer being treated using a trained machine learning model associated with a computer system is provided, the method comprising: receiving methylation data derived from a targeted methylation sequencing assay from a biological sample associated with a test subject; after the receiving, applying the methylation data to the trained machine learning model; and after the applying, receiving an output from the trained machine learning model, the output comprising: A) a first indication indicating whether minimal residual disease is present in the test subject after administering the treatment to the cancer; and B) a second indication providing a finding of the minimal residual disease in the test subject in response to the first indication, indicating whether at least a portion of the cancer cells in the minimal residual disease have transformed from a first cancer type to a second cancer type due to the development of the resistance mechanism.

[0008] The foregoing is a summary and thus may contain simplifications, generalizations, and omissions of detail; consequently, those skilled in the art will appreciate that the summary is illustrative only and is not intended to be in any way limiting.

[0009] For a better understanding of the embodiments, together with other and further features and advantages thereof, reference is made to the following description taken in conjunction with the accompanying drawings.The scope of the invention will be pointed out in the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS The patent file contains at least one color drawing / photograph. Copies of this patent with color drawing / photograph will be provided by the Office upon request and payment of the necessary fee.

[0011] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate several embodiments and, together with the description, serve to explain the principles of the present disclosure.

[0012] Figure 1A An exemplary computer system for performing the techniques described herein is depicted.

[0013] Figure 1B An exemplary software platform for performing the techniques described herein is depicted.

[0014] Figure 2 Depicted is a confusion matrix according to one aspect of the present disclosure.

[0015] Figure 3 A table according to one aspect of the present disclosure is depicted.

[0016] Figure 4 Another table according to one aspect of the present disclosure is depicted.

[0017] Figure 5 A heat map according to one aspect of the present disclosure is depicted.

[0018] Fig. 6A Depicted is a graph of data according to one aspect of the present disclosure.

[0019] Figure 6B Depicted is a confusion matrix according to one aspect of the present disclosure.

[0020] Fig. 7A A heat map according to one aspect of the present disclosure is depicted.

[0021] Figure 7B Another heat map according to an aspect of the present disclosure is depicted.

[0022] Figure 8 Depicted is a bar graph according to one aspect of the present disclosure.

[0023] Fig.9A Depicted is a schematic representation of a gene body of a transcription factor according to one aspect of the present disclosure.

[0024] Fig. 9B Another gene body representation of a transcription factor motif according to one aspect of the present disclosure is depicted.

[0025] Fig. 10A A testing process flow according to one aspect of the present disclosure is depicted.

[0026] Fig. 10B Another testing process flow according to one aspect of the present disclosure is depicted.

[0027] Fig.11 Depicted is a diagram of how to calculate β values ​​according to one aspect of the present disclosure.

[0028] Fig.12 Another heat map according to an aspect of the present disclosure is depicted.

[0029] Fig.13A -F depicts a cluster graph pattern according to one aspect of the present disclosure.

[0030] Fig.14A Depicted is a line diagram of a transcription factor according to one aspect of the disclosure.

[0031] Fig. 14B A heat map according to one aspect of the present disclosure is depicted.

[0032] Fig.15 Another string diagram of a transcription factor according to one aspect of the present disclosure is depicted.

[0033] Fig.16 Another string diagram of a transcription factor according to one aspect of the present disclosure is depicted.

[0034] Fig.17 Another string diagram of a transcription factor according to one aspect of the present disclosure is depicted.

[0035] Fig.18 Another string diagram of a transcription factor according to one aspect of the present disclosure is depicted.

[0036] Fig.19 Another heat map according to an aspect of the present disclosure is depicted.

[0037] Fig. 20A -F depicts a cluster graph pattern according to one aspect of the present disclosure.

[0038] Fig.21 A diagram depicting sample size considerations according to one aspect of the present disclosure.

[0039] Fig. 22 Illustrative information according to one aspect of the present disclosure is depicted.

[0040] Fig.23 Illustrative information according to one aspect of the present disclosure is depicted.

[0041] Fig.24 Depicted is an exemplary method of training a machine learning model to determine whether a resistance mechanism has developed in a cancer according to one aspect of the present disclosure.

[0042] Fig.25 Depicted is an exemplary method of utilizing a trained machine learning model to determine whether a resistance mechanism has developed in a cancer according to one aspect of the present disclosure.

[0043] Fig.26 Depicted is a confusion matrix according to one aspect of the present disclosure.

[0044] Fig. 27 Depicted is an exemplary method of training a machine learning model to generate a prognostic score from patterns in DNA methylation data according to one aspect of the present disclosure.

[0045] Fig.28 Illustrative information according to one aspect of the present disclosure is depicted.

[0046] Fig.29 Depicted is an exemplary method for generating a final patient prognostic score by combining a machine-generated prognostic score from a trained machine learning model with a ctDNA score, according to one aspect of the present disclosure.

[0047] Fig.30 Depicted is a graph presenting patient prognostic scores generated from different methods according to one aspect of the present disclosure. DETAILED DESCRIPTION The terms used hereinafter are to be interpreted in their broadest reasonable manner, even though it is used in conjunction with the detailed description of certain specific examples of the present disclosure. In fact, certain terms may even be emphasized hereinafter; however, any term intended to be interpreted in any limited manner will be clearly and specifically defined in this "Detailed Description" section. Both the foregoing general description and the following detailed description are exemplary and explanatory only and do not limit the claimed features.

[0049] In the present disclosure, the term "based on" means "based at least in part on". Unless the context indicates otherwise, the singular forms "a / an" and "the" include plural referents. The term "exemplary" is used in the sense of "example" rather than "ideal". The terms "comprises", "comprising", "includes", "including" or other variations thereof are intended to cover non-exclusive inclusions, so that a process, method or product that includes a list of elements does not necessarily include only those elements, but may include other elements that are not explicitly listed or inherent to such a process, method, article or device. Relative terms, such as "about", "approximately", "substantially" and "generally", are used to indicate a possible variation of ±10% of the stated value or understanding of a value. In addition, the term "between" used to describe a range of values ​​is intended to include the minimum and maximum values ​​described herein. In the claims and description, the use of the term "or" is used to mean "and / or", unless it is explicitly stated that only alternatives are referred to or the alternatives are mutually exclusive, but the present disclosure supports definitions that only refer to alternatives and "and / or". As used herein, "another" may mean at least a second or more.

[0050] As used herein, the term "user" generally encompasses any person or entity, such as a researcher and / or a care provider (e.g., a physician, etc.), who may require information, problem resolution, or engage in any other type of interaction with a provider of the systems and methods described herein (e.g., via an application program interface resident on their electronic device, etc.). The term "electronic application" or "application" may be used interchangeably with other terms (such as "program," etc.), and generally encompasses software that is configured to interact with, modify, overlay, supplement, or operate in conjunction with other software.

[0051] As used herein, a "machine learning model" generally encompasses instructions, data, and / or models configured to receive inputs and apply one or more of weights, biases, classifications, or analyses to the inputs to generate outputs. The outputs may include, for example, analyses based on the inputs, predictions, suggestions, or recommendations associated with the inputs, dynamic actions performed by the system, or any other suitable type of output. Machine learning models are typically trained using training data (e.g., empirical data, and / or samples of input data) that are fed into the model in order to establish, adjust, or modify one or more aspects of the model, such as weights, biases, criteria for forming classifications or clusters, and the like. Aspects of a machine learning model may operate on the inputs linearly, in parallel, via a network (e.g., a neural network), or via any suitable configuration.

[0052] DNA methylation has long been recognized as a hallmark of cancer and holds great promise for early cancer detection. In particular, by utilizing targeted whole genome bisulfite sequencing ("WGBS"), combined with the processing power associated with machine learning techniques, methylated DNA sequences can be efficiently read and abnormally methylated sequences (i.e., those that may be indicative of cancer) can be identified. Thus, targeted DNA methylation assays performed on cell-free DNA ("cfDNA") fragments may be able to detect multiple cancers at all stages, including at early stages when treatment may be more effective.

[0053] GRAIL's Multi-Cancer Early Detection (MCED) test uses targeted methylation assays from plasma samples along with trained classifiers to detect the presence of invasive cancers and identify the origin of cancer signals in patients (i.e., "tissue of origin"). In addition, GRAIL's post-diagnostic procedures provide tissue-informed estimates of circulating methylated variant allele frequencies (mVaF) (i.e., circulating tumor fraction) to quantify disease burden and detect minimal residual disease (MRD). In addition, GRAIL has demonstrated the ability to distinguish subtypes of different cancer types from cfDNA. More specifically, technology has been developed that can use cfDNA methylation patterns to roughly distinguish cancers by their embryological origin, and more specifically to classify epithelial solid cancers into histological types or subtypes (e.g., adenocarcinoma, HPV-associated squamous cell carcinoma, non-HPV-associated squamous cell carcinoma, Müllerian carcinoma, and transitional cell carcinoma). GRAIL was further able to use cfDNA methylation patterns to differentiate ovarian cancer type II and uterine high-grade serous carcinoma (more aggressive subtypes) from ovarian cancer type I, endometrioid carcinoma, and endometrial carcinoma (less aggressive subtypes).

[0054] Embodiments of the present application can utilize existing systems and processes to identify subtypes of cancer, as further described herein. Embodiments of the present disclosure generally relate to using methylation data to define molecular subtypes of cancer, and the present disclosure provides an in-depth case study that defines subtypes of small cell lung cancer (SCLC) as an example cancer.

[0055] SCLC is an aggressive form of lung cancer that is often associated with a poor prognosis. In contrast to non-small cell lung cancer (NSCLC), for which therapies targeting genomic mutations in various oncogenes have been effectively used, SCLC has not benefited from targeted therapies, in part because SCLC malignancies are almost universally driven by loss-of-function mutations in the tumor suppressor genes TP53 and RB1.

[0056] Subtypes of SCLC have been identified using gene expression profiling, with four major subtypes defined based on the expression levels of three key transcription factors: ASCL1, NEUROD1, and POU2F3.

[0057] Patients with advanced SCLC do not often undergo surgical resection, and SCLC is generally not discovered until it is at an advanced stage. Therefore, access to tumor tissue may be limited. However, blood-based methylation assays may have advantages in both research and development as well as in clinical settings. Methylation-based approaches in cfDNA have been envisioned for classifying lung cancer into broader types, namely NSCLC (and its subtypes: lung adenocarcinoma, lung squamous cell carcinoma, and large cell carcinoma) versus SCLC. However, this approach may not identify methods for more specifically identifying subtypes of SCLC into their component subtypes. In addition, this approach focuses on methylation levels derived from quantitative methylation-specific PCR in only four pre-specified genes (APC, HOXA9, RARB2, and RASSF1A), which do not include the three transcription factors that define SCLC subtypes recently defined in the literature.

[0058] Therefore, some aspects of the present disclosure relate to the use of a single targeted, blood-based methylation assay for early detection of cancer, quantification of the disease burden of patients on treatment, and identification of subtypes of cancer (e.g., SCLC). For example, identification of cancer molecular subtypes may help understand cancer, determine patient prognosis, or guide treatment decisions. More specifically, machine learning techniques can be used for targeted methylation data from blood-derived cfDNA to identify molecular cancer subtypes and predict the sensitivity of individual cancers to therapeutic interventions. The embodiments disclosed herein suggest understanding the methylation characteristics of cancer subtypes, which span different genomic loci and may cover more genomic loci than traditional methods. In addition, cancer subtype identification with information for therapy selection can be performed with the same blood draw and sample, which is used to detect the presence of invasive cancer in a screening environment. Although some embodiments of the present disclosure relate to SCLC, the methods described herein can be used to identify molecular subtypes of multiple cancers using methylation data.

[0059] Another situation addressed by some embodiments of the present disclosure relates to detecting resistance mechanisms developed by some cancers under treatment. More specifically, there are a variety of targeted therapies, and they can be used to treat many common types of cancers, such as adenocarcinomas. For example, lung adenocarcinomas with epidermal growth factor receptor (EGFR) mutations may be treated with EGFR inhibition, while prostate cancer may be treated with hormone deprivation therapy or androgen receptor signaling inhibition.

[0060] Cancers under treatment-related selection pressure may develop certain resistance mechanisms. For example, lung adenocarcinomas under EGFR inhibition may show resistance after a period of time (e.g., about 12 months). As another example, prostate cancer may escape treatment and transform into castration-resistant and potentially metastatic cancer. One mechanism of acquired treatment resistance is the transdifferentiation of the tumor, that is, the original adenocarcinoma transforms into a small cell neuroendocrine carcinoma. After the transformation, the neuroendocrine carcinoma may no longer respond to the original therapy, and a new treatment regimen may need to be selected and implemented. This type of treatment escape has been reported in approximately 15%-20% of advanced prostate cancers and 5%-14% of adenocarcinomas that are resistant to EGFR inhibition.

[0061] Traditionally, transdifferentiation is typically detected only after initial treatment failure (i.e., as observed by tumor recurrence) and by rebiopsy of a metastatic or recurrent or growing primary cancer. Rebiopsy can be a clinically intensive process, and by the time a rebiopsy is performed, it can often be too late for successful treatment adjustments. Solutions have been proposed to detect transdifferentiation from ctDNA using genomic profiling and targeted small variant assays. However, there may be too many shared genomic aberrations between the original adenocarcinoma and the transformed small cell carcinoma to identify a clear distinction between the corresponding genomic variants. For example, mutations on the p53 and Rb1 pathways that are almost ubiquitous in small cell carcinomas are also found in adenocarcinomas and are not specific for neuroendocrine transformation.

[0062] Treatment monitoring typically includes an assessment of disease burden or detection of MRD, which corresponds to a small amount of cancer cells that may remain in the body after treatment (e.g., tumor removal, therapy administration, etc.). During or after cancer treatment, any remaining cancer cells may become active and begin to multiply, which may lead to disease recurrence. Therefore, the detection of MRD may indicate that the treatment is not completely effective or the treatment is incomplete. MRD may exist, for example, because some cancer cells become resistant to the drugs used, for example, by transdifferentiation. MRD can be detected, for example, by liquid biopsy. However, even if there is a liquid biopsy assay to detect transdifferentiation to small cell carcinoma, a second assay and analysis is required to obtain information about MRD and disease burden and possible transdifferentiation for disease monitoring, which may be time-consuming and / or cumbersome.

[0063] Thus, in view of the foregoing, some aspects of the present disclosure relate to concurrently utilizing a single, targeted, liquid biopsy (e.g., blood-based) methylation assay for MRD assessment, as previously described above, for concurrent disease monitoring, and identifying the development of recurrence and / or resistance mechanisms in cancer, e.g., through transdifferentiation. In this regard, machine learning techniques can be employed to identify and differentiate cancer types, e.g., adenocarcinoma and neuroendocrine carcinoma, by training models to identify cancer signal origins (CSOs) based on histological type markers rather than markers based on anatomical site. One or more downstream actions can thereafter be performed, e.g., alternative treatment recommendations, and administration of different therapeutic therapies.

[0064] Another situation addressed by some embodiments of the present disclosure relates to patient prognosis. More specifically, patient prognosis can provide an estimate of the course and / or outcome of a cancer by assessing the risk of death of the patient and / or disease progression of the cancer. Knowing patient prognosis may be valuable for clinical trial design and / or therapy selection, for example, more aggressive treatment prescriptions may be prescribed for patients with a poor prognosis, and conversely, more conservative treatment prescriptions may be prescribed for patients with a less severe prognosis.

[0065] Prognostic biomarkers are desirable for managing subjects identified as having cancer. Understanding the prognosis of subjects at the cancer detection stage (particularly early cancer detection) may affect one or more of the treatment options (including surgery or non-surgical), follow-up screening schedules, or other aspects of cancer and patient management. For example, in the case of phase I NSCLC, prognostic biomarkers can identify subjects with poor prognosis and therefore increased risk profiles, which can then prove that neoadjuvant therapy is reasonable for the subject. A variety of biomarkers, including proteomics and genomic biomarkers, can be used for patient prognosis. For example, tumor score (mVAF) is a biomarker that may be an important prognostic feature. More specifically, the increased tumor score may be associated with a series of biological events known to indicate poor patient prognosis, for example, increased tumor size, presence of lymph nodes involved by tumors, presence of distant metastases, increased metabolism and mitotic activity of tumor cells, and tumor invasion of various biological structures (e.g., blood vessels, lymphatic vessels, adjacent structures, etc.).

[0066] Existing prognostic information obtained from liquid biopsies compresses the presence of ctDNA into a single metric or ctDNA score, namely tumor fraction, mVAF, or ctDNA concentration. Although the score can provide valuable insights into the subject's disease status, it is largely binary in nature, for example, a large tumor fraction corresponds to a poor prognosis, while a low tumor fraction corresponds to a good prognosis. However, studies have shown that some subjects with relatively large tumor fractions still perform relatively well, and sometimes better than other patients with relatively low tumor fractions. Therefore, even if ctDNA scores are similar, it may be advantageous to utilize additional patient-specific information to better capture the extensive heterogeneity of cancer and its impact on prognosis.

[0067] Therefore, in view of the foregoing, some aspects of the present disclosure relate to using DNA methylation features to improve the accuracy of subject prognosis. More specifically, in one embodiment, a supervised machine learning classifier can be trained on DNA methylation data, and the DNA methylation data is labeled with known outcomes for each subject (e.g., wherein each label provides an indication of subject survival after X years and an indication of disease progression, etc.). The prognostic score can be obtained from the aforementioned classifier and can thereafter be combined with the ctDNA score to generate a final prognostic score for the subject. Compared with using cTF, VAF or ctDNA concentration alone, the final prognostic score can provide a more accurate indication of the subject's prognosis.

[0068] More specifically, for the foregoing, the circulating tumor fraction cTF or methylation variant allele frequency mVAF measures how many fractions of DNA molecules in a plasma sample come from tumor cells (cTF) and carry methylation signals (mVAF) that are almost unique to tumor cells. Circulating tumor DNA molecules bring methylation signals that are independent of the total amount captured in cTF or mVAF, and these circulating tumor DNA molecules can contain different methylation patterns that characterize tumors and therefore allow prognosis. Classifiers can be trained to understand and identify methylation signals that indicate a good or poor prognosis. Therefore, once two independent prognostic metrics are obtained (i.e., the amount of ctDNA (cTF or mVAF) and the methylation pattern carried by ctDNA), these metrics can be combined into a final metric (e.g., by multiplying the two metrics by each other).

[0069] The subject matter of the present disclosure will now be described more fully below with reference to the accompanying drawings, which form a part of the present disclosure and show specific exemplary embodiments by way of illustration. An embodiment or implementation described herein as "exemplary" should not be interpreted as, for example, being preferred or advantageous over other embodiments or implementations; rather, it is intended to reflect or indicate that the embodiment is an "example" embodiment. The subject matter may be embodied in a variety of different forms, and therefore, the subject matter covered or claimed is intended to be interpreted as not being limited to any exemplary embodiment set forth herein; the exemplary embodiments are provided merely illustrative. Similarly, a reasonably broad range of the subject matter claimed or covered is also expected. For example, among others, the subject matter may be embodied as a method, device, component, or system. Therefore, the embodiment may, for example, take the form of hardware, software, firmware, or any combination thereof. Therefore, the following detailed description is not intended to be considered as having a limiting meaning.

[0070] Throughout the specification and claims, terms may have subtle meanings that are implied or connoted in context, beyond the meaning explicitly indicated. Likewise, the phrases "in one embodiment" or "in some embodiments" as used herein do not necessarily refer to the same embodiment, and the phrase "in another embodiment" as used herein does not necessarily refer to a different embodiment. For example, it is intended that the claimed subject matter includes combinations of the exemplary embodiments in whole or in part.

[0071] Figure 1A An exemplary system for identifying cancer subtypes using targeted methylation assays is depicted. The exemplary system 100 includes a data collection component 10, a database 20, and a device data intelligence component 30 connected to each other via a network 40. Alternatively or additionally, one or more of these components can be locally connected to another component without relying on a network connection; for example, via a wired connection. Sequencing data of free nucleic acids is used to illustrate the concept. However, those skilled in the art will appreciate that the current method can also be applied to sequencing data or non-sequencing data of other materials.

[0072] As disclosed herein, the data collection component 10 may include a device or machine that can be used to generate sequencing data. In some embodiments, the data collection component 10 may include a sequencing machine or a facility that uses a sequencing machine to generate nucleic acid sequence data for a biological sample. Any applicable biological sample can be used. In some embodiments, the biological sample is cell-based; for example, one or more types of tissue. In some embodiments, the biological sample is a sample containing free nucleic acid fragments. Examples of biological samples include, but are not limited to, blood samples, serum samples, plasma samples, urine samples, saliva samples, etc.

[0073] Examples of sequencing data may include, but are not limited to, sequence read data of targeted genomic locations, partial or whole genome sequencing data of a genome represented by nucleic acid fragments in free or cell-based samples, partial or whole genome sequencing data comprising one or more types of epigenetic modifications (e.g., methylation), or combinations thereof.

[0074] The data acquired by the data collection component 10 may be transmitted to the database 20 via the network 40. In some embodiments, the collected data may be analyzed by the data intelligence component 30 via a local or network connection. Figure 1B Exemplary functional modules that may be implemented to perform the tasks of data intelligence component 30 are depicted.

[0075] Figure 1B An exemplary computer system 110 for processing sequencing data is depicted. The exemplary embodiment 110 implements such functions by implementing a user input and output (I / O) module 120, a memory or database 130, a data processing module 140, a data analysis module 150, a classification module 160, a network communication module 170, and any other functional modules (e.g., error correction or compensation modules, data compression modules, etc.) that may be required to perform a specific task on one or more computer devices. As disclosed herein, the user I / O module 120 may further include input submodules such as a keyboard and output submodules such as a display (e.g., a printer, monitor, or touch pad). In some embodiments, all functions are performed by one computer system. In some embodiments, functions are performed by more than one computer.

[0076] It is also disclosed herein that a specific task can be performed by implementing one or more functional modules. In particular, each enumerated module itself can include multiple submodules. For example, the data processing module 140 can include a submodule for data quality assessment (e.g., for discarding very short sequence reads or sequence reads including obvious errors), a submodule for normalizing the number of sequence reads compared with different regions of the reference genome, a submodule for compensating / correcting GC bias, etc.

[0077] In some embodiments, the user can use I / O module 120 to manipulate data available on the local device, or data obtained from a remote service device or another user device via a network connection. For example, I / O module 120 can allow a user to perform data analysis through a graphical user interface (GUI), such as via a keyboard, mouse, or touchpad. In some embodiments, the user can manipulate the data via voice control. In some embodiments, user authentication may be required before the user is authorized to access the requested data.

[0078] In some embodiments, the user I / O module 120 can be used to manage various functional modules. For example, when an existing data processing session is ongoing, a user can request input data via the user I / O module 120. The user can do this by discretely selecting menu options or typing commands without interrupting the existing process.

[0079] As disclosed herein, a user may use any type of input to direct and control data processing and analysis via I / O module 120 .

[0080] In some embodiments, system 110 further includes a memory or database 130. In some embodiments, database 130 comprises a local database accessible via user I / O module 120. In some embodiments, database 130 comprises a remote database accessible by user I / O module 120 via a network connection. In some embodiments, database 130 is a local database that stores data retrieved from another device (e.g., a user device or a server). In some embodiments, memory or database 130 can store data retrieved in real time from an Internet search.

[0081] In some embodiments, the database 130 can send data to and receive data from one or more other functional modules, including but not limited to a data collection module (not shown), a data processing module 140, a data analysis module 150, a classification module 160, a network communication module 170, etc.

[0082] In some embodiments, database 130 may be a local database of other functional modules. In some embodiments, database 130 may be a remote database that can be accessed by other functional modules via a wired or wireless network connection (e.g., via network communication module 170). In some embodiments, database 130 may include a local portion and a remote portion.

[0083] In some embodiments, the system 110 includes a data processing module 140. The data processing module 140 can receive real-time data from the I / O module 120 or the database 130. In some embodiments, the data processing module 140 can perform standard data processing algorithms, such as one or more of noise reduction, signal enhancement, normalization of counts of sequence reads, correction of GC bias, etc. In some embodiments, the data processing module 140 can identify global or local systematic errors. For example, sequencing data can be compared with regions within a reference genome. For the same subject, the number of sequence reads compared to different genomic regions may be different. The number of sequence reads compared to the same genomic region may vary from subject to subject. Some of these differences, especially those observed in healthy subjects (i.e., including all types of organisms, not just humans), may be caused by systematic errors rather than by association with one or more disease conditions. For example, if the sequencing data corresponding to a particular genomic region shows a wide range of variations between healthy subjects, the data processing module 140 can classify the particular genomic region as a high noise region, and the corresponding data can be excluded from further analysis. In some embodiments, identification and processing of possible systematic errors may be performed by the data analysis module 140 as follows.

[0084] In some embodiments, system 110 includes a data analysis module 150. In some embodiments, data analysis module 150 includes identifying and processing systematic errors in sequencing data, as described in conjunction with data processing module 140.

[0085] In some embodiments, system 110 includes a classification module 160, which analyzes data from a test subject whose status in terms of a medical condition is unknown, and then classifies the subject based on the probability that the unknown test subject is suitable for a particular category. In some embodiments, one or more parameters include a binomial probability score calculated based on a logistic regression analysis. As disclosed herein, the binomial probability score can correspond to the probability that the subject suffers from a certain medical condition (such as cancer). For example, a score exceeding a predefined threshold can indicate the probability of suffering from a subtype of cancer of interest. In some embodiments, one or more parameters may include a sequencing data distribution pattern associated with the presence of a certain type or subtype of cancer. A subject with a pattern similar to a cancer pattern may be diagnosed as suffering from the type of cancer. In some embodiments, sequencing data distribution patterns may be identified in conjunction with more than one specific type of cancer, thereby allowing unknown subjects to be classified with further details.

[0086] As disclosed herein, the network communication module 170 can be used to facilitate communication between a user device, one or more databases, and any other suitable system or device via a wired or wireless network connection. Any communication protocol / device can be used, including but not limited to a modem, an Ethernet connection, a network card (wireless or wired), an infrared communication device, a wireless communication device, and / or a chipset (such as Bluetooth TM Devices, 802.11 devices, WiFi devices, WiMax devices, cellular communication facilities, etc.), near field communication (NFC), Zigbee communication, radio frequency (RF) or radio frequency identification (RFID) communication, PLC protocol, 3G / 4G / 5G / LTE-based communication, etc. For example, a user device with a user interface platform for processing / analyzing low-coverage sequencing data can communicate with another user device with the same platform, a conventional user device without the same platform (e.g., a conventional smart phone), a remote server, a physical device of a remote IoT local network, a wearable device, a user device communicatively connected to a remote server, etc.

[0087] The functional modules described herein are provided by way of example. It should be understood that different functional modules can be combined to create different utilities. It will also be understood that additional functional modules or submodules can be created to implement a certain utility.

[0088] The following Figure 2 -6 illustrates how a trained cancer classifier can utilize cfDNA methylation data to identify cancer subtypes.

[0089] Reference now Figure 2 , a confusion matrix is ​​provided, which presents the data generated by the machine learning model (i.e., trained cancer classifier), for being identified as a sample with a specific histological type or cancer subtype. In one embodiment, the classifier using cfDNA methylation as input is trained to distinguish the embryonic lineage of cancer cells (epithelial cells, nervous system cells, lymphocytes, myeloid cells, plasma cells, mesenchymal cells, melanocytes, neuroendocrine cells and germ cells), and further distinguishes the histological type of all epithelial cancers (which are the most common solid cancer types), i.e. adenocarcinoma (ADC), HPV-related squamous cell carcinoma (HPV), non-HPV-related squamous cell carcinoma (SCC), Mullerian carcinoma (mullerian) and transitional cell carcinoma. In cross validation, the classifier is evaluated for the collection of 2920 plasma samples from patients with cancer and 3704 plasma samples from patients without cancer. For 1619 samples, plasma samples contain circulating tumor DNA (ctDNA) with detectable levels.

[0090] exist Figure 2In the confusion matrix shown, each row contains samples from one true embryonic lineage and histological type of cancer. Each column represents samples with detectable ctDNA that the trained classifier identified as having that histological type or cancer subtype. Samples counted on the diagonal represent cases where the classifier correctly identified cancer by embryonic origin and histological type based on methylation signals in cfDNA.

[0091] Reference now Figure 3 , presents a table listing the accuracy (number of cases detected and number of results with correct cancer signal origin) broken down by each truth class.

[0092] Figure 3 and 4 A CSO accuracy of 84% and a CSO precision of 86.3% were shown to identify neuroendocrine carcinomas and distinguish them from epithelial carcinomas. For lung cancer, this distinction allowed the classifier to reliably separate non-small cell lung cancer NSCLC from small cell lung cancer SCLC using only plasma samples obtained from a blood draw, without obtaining tissue samples from lung cancer.

[0093] Reference now Figure 4 , presents a table listing Figure 2 The precision of the classifier discussed in . For each class, the frequency that the class returned by the classifier is correct and the sample actually has that class is listed. The numbers are presented once for all cases detected by the classifier (cso_precision) and once only for cases that are actually invasive cancer (cso_precision_true positive (cso_precision_tp)). In the latter column, results for samples where the patient does not have cancer (false positive detections) are not included.

[0094] Reference now Figure 5 , presenting a heat map showing classifier results for cancers of the ovary and uterus with different histological types and anatomical locations of primary cancer. In one embodiment, the instant classifier is trained according to categories defined by anatomical origin, and two different subtypes of ovarian cancer and uterine cancer are also trained. More specifically, the cells of origin of type 1 ovarian cancer are in the epithelial cells of the ovary. This is usually a slower growing cancer with a relatively good prognosis. Type 2 ovarian cancer is a cancer in which the cells of origin may be in the fallopian tube. These tumors in the fallopian tube spread tumor cells very early (probably starting with tumor growth of only a few hundred cells) and colonize the ovaries. These cancers are usually invasive and are almost only detected when metastases appear in the ovaries. These tumor metastases originating from the fallopian tube are also found in the uterus and peritoneal lining. In summary, type II ovarian cancer and these similar cancers of fallopian tube origin are described as high-grade serous carcinomas.

[0095] exist Figure 5 In the classifier shown, all high-grade serous carcinomas are trained as one class (ie, named hgs_pelvis), and epithelial grade 1 carcinomas of the ovary and uterus or endometrial carcinomas are trained as a second class (ie, named ovary_uterus). Figure 5 Classifier results for all cancers of the ovary and uterus are shown. More specifically, the colors in the heat map change from low classifier signal (blue) to high classifier signal (yellow). The classifier results for each sample are displayed as column labels, and the scores are displayed as row labels. High-grade serous type II carcinomas can be distinguished from low-grade epithelial type I carcinomas. One of the cases was misclassified as pancreatic cancer or gallbladder cancer, and not all plasma samples had detectable ctDNA (classifier result was non-cancer).

[0096] Now common reference Fig. 6A and 6B , provides a data plot and confusion matrix showing the Figure 5 Classifier results for , broken down by true column labels, i.e., high-grade serous type II carcinoma of the ovary, epithelial type I carcinoma of the ovary, high-grade serous carcinoma of the uterus, and epithelial endometrial carcinoma of the uterus. Type II high-grade serous carcinoma of the ovary was reliably identified by the classifier (20 cases with ctDNA detected and correctly classified, 2 cases without detectable ctDNA, and 2 cases with detectable ctDNA but misclassified as type I cancer).

[0097] Above Figure 2 -6 may be applicable to the identification of SCLC subtypes, as further described herein and shown in Figures 7-26. In addition, the entire disclosure of co-owned U.S. Patent Application Publication No. 2020 / 0365229 and U.S. Patent Application Publication No. 2021 / 0313006 is incorporated herein by reference, except for any definitions, subject disclaimers or disclaimers, and except where the incorporated material is inconsistent with the content explicitly disclosed herein, in which case the language of this disclosure shall prevail. Some or all of U.S. Patent Application Publication No. 2020 / 0365229 and U.S. Patent Application Publication No. 2021 / 0313006 may also be applicable to the identification of cancer subtypes.

[0098] Reference now Fig. 7A and 7B, provides a heat map of the different known subtypes of SCLC that have been previously documented in the literature. These subtypes are currently defined by expression signatures in tissues, and more specifically, by three transcription factors (i.e., expression levels of three genes): ASCL1, NEUROD1, and POU2F3. The expression level of a fourth transcription factor, YAP1, may also have a subtype effect. RNA sequencing of tissue samples and cell lines was used to identify Fig. 7A and 7B For the management of patients diagnosed with SCLC, cancer tissue samples are often not available. Therefore, it is desirable to identify these or other subtypes using only a plasma sample obtained by a simple blood draw. Fig. 7A and 7B Inspection of the heatmaps in Figure 2 shows that the expression levels of these genes are generally mutually exclusive in the tissue and cell line data presented in the figure. However, data generated from tissue samples may not reflect the intratumor heterogeneity that can be detected from plasma samples, which may provide a more complete understanding of SCLC subtype heterogeneity and therapeutic implications.

[0099] Reference now Figure 8 , provides a bar graph illustrating different types of samples obtained for methylation signature analysis for identification of SCLC subtypes. More specifically, samples were collected as part of the Circulating Cell-Free Genome Atlas (CCGA) study, in which 15,000 cancer and non-cancer participants participated in a prospective, observational case-control study for training and validation of the Galleri classifier. Figure 8 As shown, fifty-five patients with SCLC were selected for the analysis described herein, subdivided by patient sex and clinical stage, and 57 non-cancer samples matched by age, sex, and smoking status.

[0100] Reference now Fig.9A and 9B , provides a putative gene body illustration of a transcription factor gene body and motifs. One such putative transcription factor may be, for example, ASCL1. Although there is currently no information on the response of sampled patients to therapy (e.g., Figure 8), but methylation patterns can be examined in different regions of the genome that are biologically relevant to the transcription factors that define the subtypes. More specifically, the transcription factors that define the subtypes encode proteins that bind to specific genomic sequences (i.e., motifs) that can be found throughout the genome. Through the examinations described herein, it can be found that there are stratifications in regions that are biologically associated with known subtype identification biomarkers. Specifically, substructures can be observed in the methylation signals that allow the above-mentioned subtypes to be generated using methylation data. It may eventually be found that one or all of these candidate subtypes can predict the prognosis of therapy response or clinical outcome.

[0101] Reference now Fig. 10A , a general flow chart of the cfDNA methylation testing process is provided. First, a sample, such as a blood sample, can be collected from a subject. The cfDNA in the sample can be evaluated by using bisulfite conversion, which can allow the distinction between methylated cytosines and unmethylated cytosines in the cfDNA. Library preparation can then be performed by appropriate means, and an enrichment step can be performed. Sequencing can then be performed, followed by demultiplexing and alignment, and then methylation calls can be made. Based on the methylation data, classification of cancer type, histological type, or subtype can be performed. Finally, quality control can be performed and reported to a healthcare provider or patient. Embodiments of the present disclosure may provide more data for classification, or may allow additional data to be output during the classification step.

[0102] Fig. 10B A flowchart of how to incorporate exploratory analysis of methylation data into the current cfDNA methylation testing process is provided in . Fig. 10B As shown, after methylation calling, the methylation data can be analyzed to identify cancer subtypes, and analysis can be performed based on the identified cancer subtypes. The information can be used to improve the classification step, or to broaden the range of clinical or biological predictions that can be provided by the classifier. The methylation analysis step, in Fig. 10B In the paper, called the Methylation Toolbox, it allows exploration of methylation signature substructures across samples. In some aspects, analysis of methylation signature substructures can be performed very quickly for thousands of samples.

[0103] Reference now Fig.11 , provides an illustration of how to calculate a methylation beta value or a methylation-like beta value. The beta value refers to the fraction of molecules identified at a particular CpG site that is methylated. The fraction can be calculated by dividing the number of methylated molecules by the total number of molecules. For illustration purposes, red represents methylated molecules and blue represents unmethylated molecules.

[0104] Reference now Fig.12, a heat map 1200 is provided that illustrates the methylation beta value of each CpG in certain regions of the gene bodies and promoters of SCLC-associated transcription factors. The selected CpGs are those located in the gene bodies and promoter regions of transcription factors that define SCLC subtypes, but alternative regions can be selected, including one or more of the following: regions identified as potential transcription factor binding sites, regions or single CpG loci identified as differentially methylated between known transcriptionally defined SCLC subtypes, regions or single CpG loci identified as differentially methylated between responders and non-responders to a particular therapy, and / or regions or single CpG loci associated with any of those defined above.

[0105] With reference to heat map 1200, CpG is displayed on the row, and sample is displayed on the column. Row has been annotated, to illustrate in which gene each CpG is located and whether it is located in the gene body or its promoter. Inspection of heat map 1200 can reveal that in the region where the sample is hypermethylated or hypomethylated, there is a "blockiness" (blockiness) in the heat map structure. For example, heat map 1200 reveals that the cluster of sample 122 has hypermethylation along the gene body 1222 and promoter region 1224 of NEUROD1 1226. For another example, sample 124 has hypermethylation along the gene body 1242 and promoter region 1244 of ASCL1 1246. In another example, sample 126 has hypomethylation along the gene body 1262 of NEUROD1 1264.

[0106] Embodiments of the present disclosure are directed to using the beta values ​​in a matrix like this as feature values ​​to create a cancer subtype classifier (in this case, an SCLC subtype classifier), and using the feature values ​​in a labeled training sample set to create a machine learning model (e.g., a penalized regression model, a support vector machine, a shallow neural network, etc.) to identify various SCLC subtypes, as defined by traditional expression profiles or by response to treatment.

[0107] In addition to the beta value as the feature value of the classifier, the machine learning model of this embodiment can also utilize: the total counts of methylated and unmethylated molecules at each region or CpG locus and / or the counts of abnormally methylated molecules at each region of the CpG locus. Therefore, a single run of GRAIL's current targeted methylation assay can be applied to early cancer detection or disease burden monitoring, and can also be used in parallel to detect cancer subtypes, such as SCLC subtypes, to provide information for prognosis, to provide information for treatment decisions, or to identify the rise of different subtypes of treatment resistance.

[0108] Reference now Figures 13A-13F , dimensionality reduction is performed to reduce the number of variables in the data. Figures 13A-13FThe figure shown in illustrates the Fig.12 How do the samples of be clustered? Figures 13A-13F The data in are subjected to principal component analysis (PCA) dimensionality reduction, and Figures 13C-13D The data in the dataset are subjected to nonlinear dimensionality reduction using Uniform Manifold Approximation and Projection (UMAP). Fig.13A , 13B and the colors of the participant samples in 13E correspond to Fig.12 Colors assigned to tree clusters below the top dendrogram to enable visualization Fig.12 The participant sample shown in the Fig.13A , 13B , 13E and 13F on the two-dimensional map. Fig.12 In , about 6 tree clusters were identified based on the methylation signatures of the samples. These 6 tree clusters represent the subtypes imposed on the SCLC data based on the methylation signatures of each sample. Figures 13A-13F This can indicate whether subtypes inferred using methylation signatures correspond to meaningful differences in participant samples or to other clinical or patient information, such as sex or smoking status, that is not indicative of cancer subtype.

[0109] Fig.13A Indicates that from Fig.12 The participant samples in the tree cluster row appear to be grouped in similar clusters in a two-dimensional space defined by principal component 1 (PC1, explaining 73.1% of the variance) and principal component 2 (PC2, explaining 6.36% of the variance). Fig. 13C and 13D Data sets stratified by sex and smoking status are shown separately. Given that the data points for each participant sample appear to be evenly distributed across the clusters for sex and smoking status, Fig. 13C and 13D It may be suggested that the clusters depicted in the figure do not seem to be attributable to these other covariates, which may induce spurious signals when evaluating cancer data. Therefore, the SCLC subtypes identified using methylation status may be attributable to meaningful differences in methylation signatures between samples rather than confounding variables. Fig.13E Again, it shows that samples from the same tree cluster (or subtype) tend to cluster together. Fig.13F The samples are generally shown clustered according to cancer status, although there are some outliers. Fig.12 The "Stage" row at the top indicates the stage of lung cancer, with blue samples representing non-cancerous samples and increasingly darker purple samples representing increasingly advanced cancers. Fig.12 These sample colors in correspond to Fig.13F The data points shown in . Fig.13FThe cancer samples found to be clustered with non-cancer samples in may have low tumor scores, and thus no significant signal may be seen in these samples, causing them to cluster with non-cancer samples rather than cancer samples.

[0110] Reference now Fig.14A ,based on Fig.12 The heat map 1200 data shown in FIG. 1 provides a line graph 1400. Such a line graph 1400 can provide a clearer visual indication of the genomic context of the methylation state of each sample cluster. In this figure, Fig.12 The data shown is limited to one specific gene, namely NEUROD1. The beta values ​​are plotted on the y-axis of the line graph 1400, and the genomic locations are plotted on the x-axis of the line graph 1400. As a general rule of thumb, higher beta values ​​indicate more methylation on the DNA fragment in that region. Therefore, as a representative example of the information that can be gleaned from such a graph, one can focus on sample cluster 4, which corresponds to the sample cluster previously described in Fig.12 As can be seen in line graph 142, samples associated with sample cluster 4 are all highly methylated in NEUROD1, which is consistent with Fig.12 It can also be seen that NEUROD1 is highly methylated both far from the transcription start site ("TSS") and close to the TSS.

[0111] Reference now Fig. 14B , presents a heat map previously documented in the literature. See https: / / pubmed.ncbi.nlm.nih.gov / 33482121 / . The heat map also shows beta values, where more yellow areas indicate greater methylation and more blue areas indicate less methylation. The heat map is derived from cell lines and shows that subtypes defined by expression values ​​of different transcription factors are associated with methylation values ​​in cell lines at different distances from the TSS of the gene NEUROD1. The heat map data shows methylation beta values ​​for expression-defined SCLC subtypes (referred to as "ASCL1 only," "NEUROD1 only," and "both"), which are defined by whether one or both of the genes ASCL1 or NEUROD1 are expressed; the subtypes are defined by Fig. 14B The “ASCL1 only” samples in the heat map tend to have high methylation at about 200 base pairs near the TSS and high methylation at about 1,500 base pairs away from the TSS. Fig. 14B This hypermethylation pattern relative to the TSS of NEUROD1 in the “ASCL1 only” sample in is also seen in samples from cluster 4, as Fig.14AAs shown, these samples showed high methylation at both 200 bp from NEUROD1 TSS and 1500 bp from NEUROD1 TSS. This correspondence in the data may suggest that the subtype identified in sample cluster 4 may correspond to the ASCL1 high subtype. Therefore, this correlation may suggest that cancer subtypes can be identified using methylation signatures as described herein. Fig.14A It also suggests that there may be additional layers of stratification that can be seen by observing methylation patterns in the gene body of NEUROD1, for example, samples from red cluster 3 exhibited mid-range β values ​​more than 5000 bp from the NEUROD1 TSS, and this is evident from Fig. 14B Therefore, using targeted methylation assays, methylation signatures as described herein can be analyzed to identify subtypes previously defined in the literature using conventional methods and to reproduce these patterns using methylation signatures. Embodiments of the present disclosure can utilize targeted methylation assays and pattern recognition to reproduce existing subtypes or discover new subtypes that may be of interest, for example, in identifying cancer, determining prognosis, or informing treatment options.

[0112] A similar pattern is shown in Figure 15-18 Each of these plots looks at a different region of the genome. Fig.15 , based on Fig.12 The heat map data shown in 1500 provides a line graph of methylation values ​​within the gene body of the transcription factor ASCL1. Fig.15 The tree cluster 1 shown in is separated from the rest of the samples and is related to Fig.12 corresponds to the sample 124 shown in FIG.

[0113] Reference now Figure 16-18 ,based on Fig.12 The heat map data shown provides a line graph of methylation values ​​within the gene body of POU2F3 (i.e., 1600-1800). Fig.14A and 15 As stated, Figure 16-18 The methylation patterns seen in different genomic regions were called.

[0114] Reference now Fig.19 , another heat map 1900 illustrating methylation beta values ​​is provided. The same method as previously described (e.g., reference ) can be used using CpGs within transcription factor motifs or binding sites as starting points. Fig.12) The same kind of analysis was performed using the HOCOMOCO [citation needed] approach. More specifically, rather than looking at CpGs in the gene bodies of the 4 transcription factors, we could look at CpGs within regions identified as carrying motifs or binding sites for SCLC-associated transcription factors (e.g., motifs for ASCL1, NEUROD1, and POU2F3). To perform this analysis, motifs were identified from HOCOMOCO [citation needed], the locations of these motifs in the genome were identified using PWMTool [citation needed], and CpGs that were within 50 bp on either side of these motifs (called “motif windows”) were identified. Additionally, we filtered CpGs from these motif locations to require that they were significantly differentially methylated when comparing methylation values ​​from SCLC patients to those from non-cancer patients (differential methylation was identified using a beta-binomial model). We also removed CpGs that were present in more than one motif window or that overlapped with the CTCF motif window. Fig.19 , the structure can also be seen in the heat map, and the structure is different from that from Fig.12 In other words, since an alternative set of genomic regions has been used, the clusters learned from the heat map may differ from Fig.12 The clusters learned in the example are different and may provide different information.

[0115] Reference now Fig. 20A -F, provides several figures that illustrate the Fig.19 How the samples of Fig.13A - The cluster diagram shown in F.

[0116] Reference now Fig.21 , provides a graph 2100 of sample size considerations. In one embodiment, the number of samples in plasma should be arranged to be twice the number of samples required to build a classifier based on tissue data (a smaller number may be sufficient in SCLC).

[0117] Reference now Fig. 22 , presents a diagram indicating how targeted methylation probes can improve test efficiency. More specifically, cfDNA fragments may have different methylation states, such as methylated states and unmethylated states. Two types of probes, hyper-probes and hypo-probes, can be designed to target cfDNA fragments with the same methylation state at CpG. However, this does not mean that each captured cfDNA fragment is fully methylated or unmethylated. In one embodiment, each type of probe can select a type of methylation state (after bisulfite conversion). For example, a high probe can select a methylated fragment, while a low probe can select an unmethylated fragment.

[0118] Reference now Fig.23 , presents additional illustrations indicating how targeted methylation probes can improve test efficiency. More specifically, as Fig. 22 As shown, cfDNA fragments may have different methylation states. These fragments can be designated as binary targets or semi-binary targets. With respect to the former, binary targets can be targeted by the two types of probes (i.e., high probes and low probes) described above. In contrast, with respect to the latter, semi-binary targets can be targeted by only one probe (e.g., a high probe for methylated fragments and a low probe for unmethylated fragments).

[0119] As described above, some embodiments of the present disclosure may involve detecting resistance mechanisms developed by some cancers under treatment. In this case, methylation assays can be used to identify transdifferentiation. To this end, machine learning techniques can be used to identify and differentiate cancer types, e.g., adenocarcinoma and neuroendocrine carcinoma, by training models to identify cancer signal origins (CSOs) based on histological type markers rather than markers based on anatomical site.

[0120] Typically, after cancer is detected in a patient (e.g., by utilizing one or more cancer detection technologies), an appropriate treatment prescription can be prescribed. After treatment begins, regular testing for cancer (i.e., in the form of MRD for disease monitoring) can be performed at predetermined intervals, the length of which can be determined by one or more factors (e.g., the severity of the cancer, the type or intensity of treatment, biological characteristics associated with the subject, etc.). Appropriate time intervals can be about weeks, months, or years, for example, every 2-6 months (e.g., 2, 3, 4, 5, or 6 months), every year, every 2 years, etc. If cancer recurrence is detected, one or more imaging scans (e.g., bone scans, whole body scans, etc.) can be performed, and a new course of treatment can be determined based on the results of the scans. Additionally or alternatively, identifying that transdifferentiation has occurred can provide additional indicative metrics, indicating that the prescribed first-line treatment or current treatment has failed. In this way, targeted methylation assays can be used for disease monitoring. Further, if transdifferentiation has been detected and second-line treatment or further therapy has been started, targeted methylation assays can be regularly continued to be used to determine whether the disease burden changes over time. Detecting changes in disease burden can help determine whether second-line therapy is working and / or if or when second-line therapy fails.

[0121] Reference now Fig.24 , machine learning models can be trained to determine whether MRD is present within a biological sample and, if so, whether cancer cells associated with MRD have developed resistance to treatment.

[0122] At step 2405, a set of training data may be obtained from a source. In one embodiment, the source may be an accessible database, and the training data may include DNA methylation data derived from targeted methylation sequencing assays performed on biological samples obtained from training subjects. In one embodiment, the accessible database may be continuously or periodically updated with new training data.

[0123] In step 2410, the set of training data can be annotated so that each piece of training data in the set is assigned a histological label rather than a label based on anatomical site. In clinical practice, the histological type of cancer is often determined by pathological examination of tissue or other specimens. The training data can be labeled to indicate whether the DNA methylation data is obtained from a source with a given type of cancer (e.g., lung cancer, prostate cancer, etc.). The label can also include age, gender, race, or other information related to the individual source of the DNA methylation data (e.g., the patient's responsiveness to a particular course of treatment, etc.).

[0124] At step 2415, the annotated training data may be applied to a machine learning model. In one embodiment, the machine learning model may be virtually any type of supervised, unsupervised, or hybrid machine learning model selected by the user.

[0125] In step 2420, applying the annotated training data set to the machine learning model can optimize the pattern recognition capabilities of the algorithm associated with the machine learning model. More specifically, the algorithm of the machine learning model can be trained to: A) identify whether MRD is present in the test subject (i.e., by determining whether there is any cancer signal in the biological sample, or if residual disease has been detected by imaging or other clinical tests, then by using the annotation information); and, if present, B) determine whether the cancer cells associated with MRD have developed resistance mechanisms to previously administered or ongoing treatments. In this regard, the algorithm can be trained to identify whether transdifferentiation has occurred, specifically by identifying that at least some cancer cells have transformed from a first cancer type (e.g., adenocarcinoma) to a second cancer type (e.g., small cell neuroendocrine carcinoma) based on histological labels. In an optional embodiment, the machine learning model can be further trained to output a new treatment recommendation in response to identifying that a resistance mechanism has developed and / or identifying the type of resistance mechanism. In one embodiment, the new treatment recommendation may include a new drug therapy recommended for the patient, a surgical procedure performed on the patient, an adjustment to the patient's liquid biopsy collection (e.g., blood draw) schedule, and the like.

[0126] In one embodiment, a new treatment recommendation may take into account the test subject's degree of response to various types of potential treatments. More specifically, the machine learning model may be able to identify biological patterns in a blood draw of a test subject and predict (e.g., by utilizing data obtained through one or more specialized training phases that identify the degree of response of other test subjects with similar biological patterns to different types of treatments) the test subject's degree of response to a potential type of treatment. In one embodiment, a new treatment recommendation may include a recommendation for a disease monitoring schedule based on determining the subject's degree of response to a potential treatment (e.g., if the test subject is expected to respond well to treatment, a 6-month testing interval is recommended, while if the test subject is expected to respond poorly to treatment, a 2-month testing interval is recommended, etc.).

[0127] Reference now Fig.25 , a trained machine learning model can be used to determine whether MRD is present and, if so, whether the cancer cells have developed resistance to ongoing treatment.

[0128] At step 2510, methylation data can be obtained, the methylation data being derived from a targeted methylation sequencing assay performed on a biological sample obtained from a test subject. At step 2515, the methylation data can be applied to a trained machine learning model (e.g., as described above with reference to Fig.24The machine learning model discussed). The methylation data can be processed by the algorithm of the trained machine learning model, and the output result can be received. More specifically, at step 2520, the machine learning model can determine whether MRD exists in the test subject. This determination can be facilitated by identifying whether there are any cancer signals in the methylation data. In response to determining that there is no cancer signal in the methylation data at step 2520, the machine learning model can output a result at step 2525 indicating that cancer is not detected. Alternatively, in response to determining that MRD exists at step 2520, the machine learning model can further determine at step 2530 whether the cancer cells have developed resistance mechanisms in response to ongoing treatment. This determination can be facilitated by identifying whether at least a portion of the detected cancer cells have been transformed from a first cancer type (i.e., the cancer type initially identified and the cancer type initially targeted by the current treatment) to another cancer type. The algorithm of the trained machine learning model may be able to perform the identification based on a training data set annotated with histological labels for each cancer type. More specifically, when predicting the histological type based on ctDNA detected in a liquid biopsy (e.g., plasma) sample, a robust separation between adenocarcinoma and neuroendocrine carcinoma can be identified. In response to determining at step 2530 that the cancer cells associated with MRD are equivalent to the first cancer type, the machine learning model can output a result at step 2535 indicating that the first cancer type has recurred in the test subject. Alternatively, in response to determining at step 2530 that at least a portion of the cancer cells associated with MRD have been transformed from the first cancer type to the second cancer type, the machine learning model can output a result at step 2540 indicating that the cancer cells of the first cancer type have developed resistance to ongoing treatment. In an optional embodiment, the trained machine learning model can further output a treatment recommendation based on determining that the cancer has developed resistance to treatment. For example, the treatment recommendation can suggest an alternative therapy that can be administered to the test subject, a potential surgery that can be performed on the test subject, an adjustment to the liquid biopsy collection (e.g., blood draw) schedule for the test subject, and the like.

[0129] Reference now Fig.26, presents a confusion matrix 2600, which is the result of applying methylation data from a set of test subjects to a trained machine learning model. In one embodiment, the machine learning model is trained by using histological cancer types as CSO labels. For clarity, only samples with detected cancer signals are shown. The column labels correspond to the true value categories of the histological types, and the row labels indicate the classifier results. For explanation purposes, "adc" corresponds to adenocarcinoma; "hpv" corresponds to HPV-positive squamous cell carcinoma; "scc" corresponds to HPV-negative squamous cell carcinoma, and "epithelial_unclear type (epithelial_nos)" corresponds to epithelial carcinoma of unclear histological type. Inspection of the confusion matrix can reveal the accuracy of the prediction results of the machine learning model. The confusion matrix indicates the frequency of calls that the classifier generates from the individual's blood draw that match the diagnosis determined for the individual based on clinical examination. The experimental results show that targeted methylation sequencing assays can be used for disease monitoring and can be used in parallel for disease burden quantification or MRD detection and detection of cancer transdifferentiation.

[0130] Reference now Fig. 27 , the machine learning model can be trained to generate a subject prognostic score. In step 2705, a collection of training data can be obtained from a source. In one embodiment, the source can be an accessible database, and the training data can include DNA methylation data, which is derived from a targeted methylation sequencing assay performed on a biological sample (e.g., a liquid sample, such as a blood sample) obtained from one or more training subjects. In one embodiment, the accessible database can be continuously or periodically updated with new training data. In one embodiment, the subject methylation data on which the final prognostic score is based can be the same methylation data used in other cancer classification determinations. More specifically, both prognosis and cancer classification determinations can be performed from the same blood sample, which can avoid the subject having to undergo additional liquid sample collection, such as another blood draw.

[0131] Aspects of the embodiments may employ filtering cfDNA fragments based on the methylation status of the cfDNA fragments. For example, embodiments of the present disclosure may identify regions in the genome in which the presence of some abnormally methylated fragments can separate cancer subjects with good prognosis from those with poor prognosis, or separate cancer subjects from non-cancer subjects. In one embodiment, the DNA methylation data used to train a machine learning model may contain only those abnormally methylated cfDNA fragments. More specifically, when the methylation pattern of a DNA fragment is unlikely to be observed in the cfDNA fragments seen in healthy non-cancer study participants, the DNA fragment may be included in subsequent training and scoring.

[0132] In addition to the foregoing, the training data can be based on a methylation feature set substantially identical to the methylation feature set used in the aforementioned Galleri classifier (i.e., the feature set can represent those CpG sites of abnormal methylation that are identified for cancer detection or for distinguishing the origins of different cancer signals). For example, the P_prognostic_CSO (P_prognostic_CSO) labeling method and the P_prognostic_Galleri_fragment (P_prognostic_Galleri_fragment) labeling method can both utilize methylation patterns that are commonly used to distinguish different cancer types. More specifically, each of the aforementioned methods utilizes the same set of methylation sites for training the Galleri classifier (i.e., those CpG sites that cancer detection and / or cancer signal origins in the genome rely on). Alternatively, the training data can be based on different feature sets, which correspond to methylation sites that are considered to be particularly valuable for prognosis of the subject. For example, the P_prognostic_fragment method can be a de novo characterization in which a methylation panel is examined for unique methylation sites that may be particularly relevant to the prognosis of a subject (i.e., more relevant to the prognosis of a subject). There may or may not be overlap between the traditional Galleri methylation pattern and the methylation pattern identified as being particularly relevant to determining the prognosis of a subject.

[0133] In step 2710, the set of training data may be annotated so that each piece of training data in the set (i.e., each set of DNA methylation data associated with a training subject) is assigned a label indicating the outcome of the disease on the training subject. In one embodiment, possible labels may include: subject death after X years, subject disease progression within X years, disease progression and subject death within X years, subject metastasis or recurrence after X years, and subject disease-free and alive after X years. For example, referring to Fig.28 , providing a plurality of sample labels utilized on the training data set. The training labels in set 2805 provide an indication of whether the training subject survived or died after a predetermined number of years and / or whether the cancer disappeared, stagnated, or progressed (e.g., even after surgery or after treatment, the cancer continued to grow or began to grow again).

[0134] Return to reference Fig. 27, in step 2715, the annotated training data can be applied to the machine learning model. In one embodiment, the machine learning model can actually be any type of supervised, unsupervised, or hybrid machine learning model selected by the user. In one embodiment, the application of the annotated training data can optimize the pattern recognition capabilities of the algorithm associated with the machine learning model to understand which methylation patterns in cancer subjects indicate subject prognosis. For example, the machine learning model can learn that the presence of a group of abnormally methylated CpG sites (Group A) leads to a poor subject prognosis (e.g., disease progression and death within 2 years), while a different group of abnormally methylated CpG sites (Group B) leads to a more favorable subject prognosis (no disease progression and subject survival after 2 years).

[0135] Reference now Fig.29 , discloses a method for generating a final prognostic score for a patient based on a combination of a model-generated prognostic score and a ctDNA score.

[0136] At step 2905, methylation data can be obtained, the methylation data being derived from a targeted methylation sequencing assay performed on a biological sample obtained from a test subject. The biological sample can be a liquid biological sample, such as a blood draw. At step 2910, the methylation data derived from the targeted methylation sequencing assay can be applied to a trained machine learning model (e.g., as described above with reference to Fig.24The machine learning model discussed). The methylation data can be processed by the algorithm of the trained machine learning model, and the output result in the form of a prognostic score can be received in step 2915. In one embodiment, the prognostic score can have a value in the range of 0 to 1, where 0 indicates that the methylation pattern is highly similar to the methylation pattern of patients in the training population with a good prognosis, and 1 indicates that the methylation pattern is highly similar to the methylation pattern of patients in the training population with a poor prognosis. In step 2920, a ctDNA score can be determined, which represents the score of cfDNA in a biological sample that originates from a tumor rather than non-cancer tissue (i.e., a tumor score). In one embodiment, one or more computational tools and / or methods known in the art can be used to estimate the tumor score. In one embodiment, the ctDNA score can be obtained by taking the log10 of the calculated tumor score value, i.e., log10 (mVAF). In step 2925, a final patient prognostic score can be generated by combining the machine-generated prognostic score received in step 2915 with the ctDNA score identified in step 2920. In one embodiment, the ctDNA score and the machine-generated prognostic score can be combined in one or more different ways. For example, one non-limiting way to combine a ctDNA score and a machine-generated prognostic score can be to multiply the two metrics. Other ways to combine a ctDNA score and a machine-generated prognostic score include adding the two metrics or combining them together by almost any other mathematical means.

[0137] In one embodiment, the final patient prognosis score can be used as the basis for one or more downstream actions and / or determinations. For example, the range of values ​​into which the final patient prognosis score falls can provide an indication of how good or poor the prognosis is. For example, if the final patient prognosis score falls into a first value range (e.g., 0 to 0.3), a good prognosis indication can be provided, if the final patient prognosis score falls into a second value range (e.g., 0.3-0.6), a medium prognosis indication can be provided, and if the final patient prognosis score falls into a third value range (e.g., 0.6-1), a poor prognosis indication can be provided. These ranges are exemplary only, and different groupings or grouping quantities can be utilized according to clinically relevant ranges. In other instances, a threshold value can be set, and the final patient prognosis score can be compared with the threshold value. A value below a threshold value can indicate a good prognosis score, and a value above a threshold value can indicate a poor prognosis indication.

[0138] Additionally or alternatively, various recommendations (e.g., recommendations for the type of clinical research to be participated in, recommendations for disease management processes, recommendations for follow-up plans during / after treatment, etc.) can be dynamically provided to subjects, clinicians, and / or clinical trial supervisors based on the range of values ​​into which the final patient prognosis score falls or whether the value is above or below a threshold. For example, if the final patient prognosis score falls within the above-mentioned third value range or is above a threshold, a recommendation can be created, suggesting that the subject be recruited into a more active clinical trial in an attempt to improve the poor predicted prognosis. In other aspects, a more active treatment regimen can be recommended, or a more frequent follow-up schedule during / after treatment can be recommended (e.g., more frequent testing, scanning, biopsy collection (e.g., blood drawing) or other follow-up plans can be recommended). A recommendation can be created, suggesting that the biopsy collection (e.g., blood drawing) schedule of the subject be modified after treatment, and blood drawing can occur more frequently through the schedule. In yet other aspects, if a patient identified as having a poor prognosis is recruited into a clinical trial, the clinical trial recruitment or the resulting data analysis can be modified so that prognosis is taken into account as a relevant variable that may be controlled.

[0139] For another example, if the final patient prognosis score falls within the above-mentioned first value range or is below the threshold, a less aggressive treatment regimen can be recommended, or a less frequent follow-up regimen during / after treatment can be recommended. For example, more frequent testing, scanning, biopsy collection (e.g., blood draw) or other follow-up regimens can be recommended. A recommendation can be created to suggest modifying the subject's biopsy collection (e.g., blood draw) schedule after treatment, through which blood draws can occur less frequently. In yet other aspects, if patients identified as having a better prognosis are recruited into a clinical trial, the clinical trial recruitment or the resulting data analysis can be modified so that prognosis is taken into account as a relevant variable that may be controlled.

[0140] In addition to the above, the final patient prognostic score can be utilized in a variety of ways in the context of a clinical trial. More specifically, in the context of clinical trial design, knowledge of patient prognosis can be valuable for patient stratification and inclusion criteria in order to increase event rates and power. For example, given a particular clinical trial, the patient population can be divided into different subgroups based on prognostic indications derived from the final patient prognostic score.

[0141] Reference now Fig.30, presenting Figure 3000, the figure illustrates a plurality of patient prognosis predictions for a type of cancer (i.e., lung squamous cell carcinoma). In Figure 3000, multiple prognosis prediction sets are presented with different specificity levels, and each set contains prognosis predictions generated using different methods. For the purpose of this discussion, two specific prognosis prediction metrics (3005, 3010) in a prognosis prediction set 3015 in Figure 3000 can be checked. More specifically, the related set 3015 can contain the first prognosis prediction 3005 corresponding to a single tumor score value (i.e., mVAF) and another prognosis prediction 3010 (i.e., P_prognosis_Galleri_fragment) corresponding to the combination of the prognosis score generated by the machine and the ctDNA score, as previously disclosed above. Inspection of Figure 3000 shows that the prognosis prediction 3010 associated with the P_prognosis_Galleri_fragment generally has greater specificity and sensitivity than the tumor score value 3005, which indicates that the former provides a better patient prognosis than only a single mVAF value. An additional observation that can be made from graph 3000 is that the prognostic prediction 3010 associated with P_prognosis_Galleri_fragments has similar specificity in each set as the P_prognosis_CSO metric and P_prognosis_fragments, but outperforms both in terms of sensitivity.

[0142] In some embodiments, the methods, systems and / or classifiers disclosed herein can be used to detect the presence (or absence) of cancer, monitor cancer progression or recurrence, monitor treatment response or effectiveness, monitor transdifferentiation of cancer, determine the presence or monitor minimal residual disease (MRD), quantify disease burden, generate patient prognostic scores or any combination thereof. In some embodiments, the system and / or classifier can be used to identify the histological type or molecular subtype of cancer. For example, the system and / or classifier can be used to identify cancer as any of the following cancer types: epithelial cancer, nervous system cancer, lymphocyte cancer, medullary cancer, plasma cell cancer, mesenchymal cell cancer, melanocyte cancer, neuroendocrine cancer and germ cell cancer, and all epithelial cancers are further distinguished into histological types, i.e., adenocarcinoma (ADC), HPV-related squamous cell carcinoma (HPV), non-HPV-related squamous cell carcinoma (SCC), Mullerian duct-derived carcinoma and transitional cell carcinoma. In some embodiments, subtypes of known cancer types can be identified, such as type I epithelial ovarian cancer and type II high-grade serous ovarian cancer, or different molecular subtypes of SCLC that can be identified by dominant transcription factors. In some embodiments, a test report can be generated to provide the patient with their test results, including, for example, a probability score that the patient has a disease state (e.g., cancer), a disease type (e.g., a cancer type), and / or a histological type or molecular subtype (i.e., when a report is generated, a CSO label can be generated for a specific person based on auxiliary factors such as gender, smoker status, etc.). In some embodiments, the methods and / or classifiers disclosed herein are used to identify cancer types, histological types, or molecular subtypes in suspects who have been diagnosed with cancer. In some embodiments, this identification of cancer types, histological types, or molecular subtypes can use the same blood draws, plasma samples, and sequencing results that have been used to detect cancer in subjects participating in early detection or screening programs for cancer. According to various aspects of the present disclosure, the methods and systems disclosed herein can be trained to detect or classify a variety of cancer indications. For example, the methods, systems, and classifiers disclosed herein can be used to detect the presence of one or more, two or more, three or more, five or more, or ten or more different types of cancer, histological types, or molecular subtypes. In some embodiments, the cancer is one or more of adenocarcinoma, HPV-associated squamous cell carcinoma, non-HPV-associated squamous cell carcinoma, type I Müllerian carcinoma, type II Müllerian carcinoma, transitional cell carcinoma, or a neuroendocrine carcinoma (e.g., small cell carcinoma), whose molecular subtypes are identified by different transcription factors.

[0143] Data sets can be generated for sequence data obtained from cancer patients within any desired time point set, and analyzed according to the method disclosed herein, to monitor the cancer status of patients. In certain embodiments, the first time point is before cancer treatment (for example, before resection surgery or therapeutic intervention), and the second time point is after cancer treatment (for example, after resection surgery or therapeutic intervention), and the method is used to monitor the effectiveness of treatment. In other embodiments, the first time point and the second time point are both before cancer treatment (for example, before resection surgery or therapeutic intervention). In other embodiments, the first time point and the second time point are both after cancer treatment (for example, before resection surgery or therapeutic intervention), and the method is used to monitor the effectiveness of treatment or the loss of therapeutic effectiveness. In other embodiments, data sets can be generated for sequence data obtained from cancer patients at the first time point and the second time point, and analyzed, for example, to monitor cancer progression, to determine whether cancer is in remission (for example, after treatment), to monitor or detect residual disease or disease recurrence, or to monitor treatment (for example, therapeutic) efficacy or whether cancer changes its histological type or molecular subtype while developing resistance to cancer treatment.

[0144] In some embodiments, the time interval between the first time point and the second time point ranges from about 15 minutes to about 30 years, such as about 30 minutes, such as about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, or about 24 hours, such as about 1, 2, 3, 4, 5, 10, 15, 20, 25, or about 30 days, or such as about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 months, or such as about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 months. 5, 4, 4.5, 5, 5.5, 6, 6.5, 7, 7.5, 8, 8.5, 9, 9.5, 10, 10.5, 11, 11.5, 12, 12.5, 13, 13.5, 14, 14.5, 15, 15.5, 16, 16.5, 17, 17.5, 18, 18.5, 19, 19.5, 20, 20.5, 21, 21.5, 22, 22.5, 23, 23.5, 24, 24.5, 25, 25.5, 26, 26.5, 27, 27.5, 28, 28.5, 29, 29.5 or about 30 years. In other embodiments, the frequency of generating a data set for sequence data obtained from a patient may be: at least once every 3 months, at least once every 6 months, at least once a year, at least once every 2 years, at least once every 3 years, at least once every 4 years, or at least once every 5 years.

[0145] In yet another embodiment, the information obtained from any of the methods described herein can be used to make or influence clinical decisions (e.g., cancer diagnosis, treatment selection, evaluation of treatment effectiveness, etc.). For example, in one embodiment, a doctor can prescribe appropriate treatment (e.g., resection, radiation therapy, chemotherapy, and / or immunotherapy) based on the use of the data set in the classification process. In some embodiments, information such as the classification based on the data set can be provided as a readout to the doctor or subject. In some embodiments, the classification based on the data set can indicate the effectiveness of the cancer treatment.

[0146] In some embodiments, treatment is one or more cancer therapeutic agents selected from the group including: chemotherapeutic agents, targeted cancer therapeutic agents, differentiation therapeutic agents, hormone therapeutic agents and immunotherapeutic agents. For example, treatment can be one or more chemotherapeutic agents selected from the group including: alkylating agents, antimetabolites, anthracyclines, antitumor antibiotics, cell scaffold disruptors (taxanes), topoisomerase inhibitors, nuclear division blocking substances, corticosteroids, kinase inhibitors, nucleotide analogs, platinum-based agents and any combination thereof. In some embodiments, treatment is one or more targeted cancer therapeutic agents selected from the group including: signal transduction inhibitors (e.g., tyrosine kinase and growth factor receptor inhibitors), histone deacetylase (HDAC) inhibitors, retinoic acid receptor agonists, proteosome inhibitors, angiogenesis inhibitors and monoclonal antibody conjugates. In some embodiments, treatment is one or more differentiation therapeutic agents including retinoids, such as retinoic acid, alitretinoin and bexarotene. In some embodiments, treatment is one or more hormone therapeutic agents selected from the group including: antiestrogens, aromatase inhibitors, progestins, estrogens, antiandrogens and GnRH agonists or analogs. In one embodiment, treatment is one or more immunotherapeutics selected from the group including: monoclonal antibody therapy (such as rituximab (RITUXAN (Rituxan)) and alemtuzumab (CAMPATH)), nonspecific immunotherapy and adjuvants (such as BCG, interleukin-2 (IL-2) and interferon-α), immunomodulatory drugs (e.g., thalidomide and lenalidomide (REVLIMID (RevliMID))). Appropriate cancer therapeutic agents can be selected based on characteristics such as tumor type, histological type, molecular subtype, cancer stage, previous cancer treatment or therapeutic agent exposure, and other characteristics of the cancer.

[0147] In addition to standard desktops or servers, any computer system capable of achieving the required storage and processing requirements will be suitable for practicing the embodiments of the present disclosure and is fully within the scope of the present disclosure. This can include tablet devices, smart phones, pin pad devices, and any other computer devices, whether mobile devices or even devices distributed on a network (i.e., cloud-based devices).

[0148] Unless otherwise specifically stated, it will be apparent from the following discussion that it should be understood that throughout the specification, discussions using terms such as "processing," "computing," "calculating," "determining," "analyzing," etc. refer to the actions and / or processes of a computer or computing system or similar electronic computing device that manipulate and / or transform data represented as physical (such as electronic) quantities into other data similarly represented as physical quantities.

[0149] In a similar manner, the term "processor" may refer to any device or portion of a device that processes electronic data, such as from registers and / or memory, to transform the electronic data into other electronic data, such as that which may be stored in registers and / or memory. A "computer," "computing machine," "computing platform," "computing device," or "server" may include one or more processors.

[0150] According to various embodiments of the present disclosure, the methods described herein may be implemented by a software program executable by a computer system. Further, in an exemplary non-limiting embodiment, the implementation may include distributed processing, component / object distributed processing, and parallel payment. Alternatively, a virtual computer system process may be constructed to implement one or more of the methods or functions as described herein.

[0151] Although this specification describes components and functions that can be implemented in specific embodiments with reference to specific standards and protocols, the present disclosure is not limited to such standards and protocols. For example, standards for Internet and other packet-switched network transmissions (e.g., TCP / IP, UDP / IP, HTML, HTTP, etc.) represent examples of the prior art. Such standards are periodically replaced by faster or more efficient equivalents having substantially the same functionality. Therefore, alternative standards and protocols having the same or similar functionality as those disclosed herein are considered to be equivalents thereof.

[0152] It will be appreciated that in one embodiment, the steps of the method discussed are performed by an appropriate processor (or processors) of a processing (i.e., computer) system executing instructions (computer readable code) stored in a storage device. It will also be appreciated that the disclosed embodiments are not limited to any particular implementation or programming technique, and that the disclosed embodiments may be implemented using any suitable technique for implementing the functionality described herein. The disclosed embodiments are not limited to any particular programming language or operating system.

[0153] It should be understood that in the above description of exemplary embodiments, different features of the embodiments are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more different inventive aspects. However, this method of disclosure should not be interpreted as reflecting an intention that the claimed embodiments require more features than those expressly recited in the claims. On the contrary, as reflected in the following claims, aspects of the invention lie in less than all the features of a single previously disclosed embodiment. Therefore, the claims following the "Detailed Description" are hereby expressly incorporated into this "Detailed Description", with each claim acting as a separate embodiment by itself.

[0154] In addition, as will be appreciated by those skilled in the art, although some embodiments described herein include some features but do not include other features included in other embodiments, the combination of features of different embodiments is intended to be within the scope of this disclosure and form different embodiments. For example, in the following claims, any of the required embodiments may be used in any combination.

[0155] In addition, some embodiments are described herein as methods or combinations of method elements that can be implemented by a processor of a computer system or by other components that perform functions. Therefore, a processor with necessary instructions for performing such methods or method elements forms a component for performing the method or method elements. In addition, the elements of the device embodiments described herein are examples of components for performing functions performed by the elements for the purpose of performing functions.

[0156] In the description provided herein, numerous specific details are listed. However, it should be understood that the embodiments of the present disclosure can be practiced without these specific details. In other cases, well-known methods, structures and techniques are not shown in detail in order not to obscure the understanding of this specification.

[0157] Similarly, it should be noted that the term coupling when used in the claims should not be interpreted as being limited to only direct connections. The terms "coupled" and "connected" and their derivatives may be used. It should be understood that these terms are not intended to be synonymous with each other. Therefore, the scope of the statement that device A is coupled to device B should not be limited to devices or systems in which the output of device A is directly connected to the input of device B. This means that there is a path between the output of A and the input of B, which path may be a path that includes other devices or components. "Coupled" can mean that two or more elements are in direct physical or electrical contact, or that two or more elements are not in direct contact with each other but still cooperate or interact with each other.

[0158] Therefore, although embodiments that are considered to be preferred embodiments of the present disclosure have been described, those skilled in the art will recognize that other and further modifications may be made thereto without departing from the spirit of the present disclosure, and it is intended that all such changes and modifications within the scope of the present disclosure be claimed. For example, any rules given above merely represent procedures that may be used. Functionality may be added or deleted from the block diagrams, and operations may be interchanged between functional blocks. Steps may be added or deleted from the methods described within the scope of the present disclosure.

[0159] The subject matter disclosed above should be considered illustrative rather than restrictive, and the appended claims are intended to cover all such modifications, enhancements, and other embodiments that fall within the true spirit and scope of the present disclosure. Therefore, in order to be allowed to the greatest extent by law, the scope of the present disclosure will be determined by the most widely permitted interpretation of the following claims and their equivalents and should not be limited to or confined to the foregoing detailed description. Although various embodiments of the present disclosure have been described, it is obvious to those of ordinary skill in the art that more embodiments and embodiments are possible within the scope of the present disclosure. Therefore, except to be consistent with the appended claims and their equivalents, the present disclosure is not bound.

Claims

1. A method for detecting a subtype of a disease state using a system, characterized in that The method includes: receiving, at an input component of the system, a set of sequence reads associated with a nucleic acid sample; generating methylation data by analyzing the set of sequence reads using a processor of the system; and The methylation data is analyzed using the processor to identify a subtype of the disease state.

2. The method according to claim 1, characterized in that The method further comprises: applying, using the processor, information associated with the identified subtypes as training input to a disease state classifier; and The disease state classifier trained based on the information associated with the identified subtype is utilized on one or more subsequent sets of nucleic acid samples.

3. The method according to claim 1, characterized in that Subtypes of the disease state include the embryonic origin of the cancer cells.

4. The method according to claim 1, characterized in that Subtypes of the disease state include histological subtypes.

5. The method according to claim 1, characterized in that Subtypes of the disease state include molecular subtypes.

6. The method according to claim 5, characterized in that The molecular subtypes have been previously defined based on protein expression identified using cancer tissue samples.

7. The method according to claim 5, characterized in that The molecular subtypes have been previously defined based on gene expression identified using cancer tissue samples.

8. The method according to claim 5, characterized in that The molecular subtypes have been previously defined based on genomic alterations identified using cancer tissue samples.

9. The method according to claim 2, characterized in that Define and train molecular subtype information based on outcomes of different treatments.

10. The method according to claim 2, characterized in that Molecular subtype information is defined and trained based on the subject's prognosis for cancer progression.

11. The method according to claim 2, characterized in that Molecular subtype information is defined and trained based on the subject's prognosis for cancer recurrence.

12. A method for training a machine learning model to detect the development of resistance mechanisms in cancer, characterized in that The method comprises: obtaining a set of training data from a source, wherein the training data comprises methylation data derived from a targeted methylation sequencing assay; After said obtaining, annotating said set of training data by assigning a histological label to each piece of training data in said set; applying the set of annotated training data to the machine learning model; and The pattern recognition capabilities of an algorithm associated with the machine learning model are optimized based on the application.

13. The method according to claim 12, characterized in that The source is a plurality of training subjects.

14. The method according to claim 12, characterized in that The histological signature is associated with one of the following: adenocarcinoma and small cell neuroendocrine carcinoma.

15. The method according to claim 12, characterized in that Optimizing the algorithm's pattern recognition capabilities includes: Enable the machine learning model to: A) determining the presence or absence of minimal residual disease in a test set of methylation data; and B) in response to determining that the minimal residual disease is present in the test set, determining whether at least a portion of the cancer cells in the minimal residual disease have converted from a first cancer type to a second cancer type due to development of the resistance mechanism.

16. The method according to claim 15, characterized in that Optimizing the algorithm's pattern recognition capabilities includes: Enable the machine learning model to: C) in response to determining that the minimal residual disease is present and at least a portion of cancer cells in the minimal residual disease have transformed from the first cancer type to the second cancer type, suggesting a treatment recommendation for the second cancer type.

17. A method of using a trained machine learning model associated with a computer system to detect the development of resistance mechanisms in a cancer being treated, characterized in that The method comprises: receiving methylation data derived from a targeted methylation sequencing assay from a biological sample associated with a test subject; After said receiving, applying said methylation data to said trained machine learning model; and After the applying, receiving an output from the trained machine learning model, the output comprising: A) a first indication of the presence or absence of minimal residual disease in said test subject after administration of said treatment for said cancer; and B) a second indication, responsive to the first indication, providing a finding of the minimal residual disease in the test subject, which indicates whether at least a portion of the cancer cells in the minimal residual disease have transformed from the first cancer type to the second cancer type due to the development of the resistance mechanism.

18. The method according to claim 17, characterized in that The resistance mechanism is transdifferentiation.

19. The method according to claim 17, characterized in that The first cancer type is adenocarcinoma, and wherein the second cancer type is small cell neuroendocrine carcinoma.

20. The method of claim 17, wherein: The output further includes: C) a third indication, responsive to the first indication providing a finding that the minimal residual disease is present in the test subject and the second indication providing another finding that at least a portion of the cancer cells in the minimal residual disease have transformed from the first cancer type to the second cancer type, indicating a treatment recommendation for the second cancer type.

21. A method for training a machine learning model to generate a patient prognosis score, characterized in that The method comprises: obtaining a set of training data from a source, wherein the training data comprises methylation data derived from a targeted methylation sequencing assay; After said obtaining, annotating said set of training data by assigning a known patient outcome to each piece of training data in said set; applying the set of annotated training data to the machine learning model; and The patient prognosis prediction ability of an algorithm associated with the machine learning model is optimized based on the application.

22. A method for determining a final prognostic score for a test subject, characterized in that The method includes: receiving methylation data derived from a targeted methylation sequencing assay from a biological sample associated with a test subject; After said receiving, applying said methylation data to a trained machine learning model; receiving, after the applying, an output from the trained machine learning model, the output comprising a first prognostic score for the test subject; Identifying a ctDNA score based on the biological sample; combining the first prognostic score and the ctDNA score; and Based on the combination, a final prognostic score for the test subject is generated.

23. The method of claim 22, wherein: Identifying the ctDNA score includes: determining a tumor score value associated with the biological sample; and The log10 of the tumor score values ​​was calculated.

24. The method of claim 23, wherein: The combining comprises multiplying the log10 of the tumor score value by the first prognostic score.

25. The method of claim 22, wherein: The final prognostic score is a value between 0 and 1.

Images