Tree-based models for selecting treatments and determining expected treatment outcomes

JP2023552391A5Active Publication Date: 2024-12-04FOUNDATION MEDICINE INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2023533926
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2020-12-03
Filing Date
2021-12-02
Publication Date
2024-12-04
Estimated Expiration
2041-12-02

AI Technical Summary

Technical Problem

Existing methods struggle to predict individual patient responses to treatments due to discrepancies between clinical trial populations and the broader patient population, making it difficult for healthcare providers to determine the most effective treatment option for a specific patient.

Method used

Utilizing tree-based regression models, such as Bayesian additive regression trees, trained on patient characteristics and treatment outcomes from both clinical trials and real-world data to predict expected treatment outcomes for individual patients.

Benefits of technology

Enables personalized treatment predictions by accurately determining the most effective treatment option for a patient based on their specific characteristics, improving treatment efficacy and patient outcomes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
  • Figure 00000000_0001_ABST
    Figure 00000000_0001_ABST
Patent Text Reader

Abstract

Described herein are methods and systems for determining an expected disease treatment outcome when treating a subject, methods and devices for selecting a treatment option for a subject, and methods for treating a disease in a subject. The methods can include receiving a plurality of subject characteristics for the subject, accessing a tree-based model corresponding to treatment options for the disease, where the tree-based model is generated based on a plurality of prior patient characteristics and associated treatment outcomes for the corresponding treatment options, and determining from the plurality of subject characteristics and the tree-based model an expected treatment outcome if the subject is treated with the corresponding treatment option.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] Cross - reference to Related Applications This application claims the benefit of U.S. Provisional Patent Application No. 63 / 121,195, filed on December 3, 2020, which is hereby incorporated by reference in its entirety.

[0002] This specification describes methods and devices for determining disease treatment outcomes expected when treating a subject, methods and devices for selecting treatment options for a subject, and methods for treating a subject's disease.

Background Art

[0003] Often, the highest - quality evidence regarding the effectiveness of a drug is obtained from clinical trials, which provide summary statistics of the average effect in the test - subject population. The patient population enrolled in a clinical trial is usually smaller than the intended - use population, which may differ in disease prognosis, treatment location, or other factors regarding test - selection criteria. Thus, it can be difficult to predict how patients outside the clinical - trial study group will respond to a given treatment, especially in clinical trials when patients are not stratified or when patients with a total profile of a given intended treatment are not included. When a physician tries to make a treatment decision about an individual patient, the effect on that particular patient is more important than the average effect on the clinical - trial population. That is, a healthcare provider during treatment desires to know how a particular patient will respond to treatment, regardless of the clinical - trial results for the selected patient population.

[0004] Physicians often have two or more treatment options for treating a patient, but must determine which of these treatment options is most effective in treating a particular patient, even if the treatment options are less effective when used to treat the clinical - trial patient population.

Summary of the Invention

[0005] This specification describes methods and devices for selecting treatments for subjects with a disease, methods for determining expected treatment outcomes when administering treatment options to patients, and devices for implementing these methods. It also describes methods and systems for treating patients using the treatments selected by the above methods.

[0006] The method may involve using one or more regression models (tree-based models such as Bayesian additive regression tree models or other tree-based regression models), each of which can determine the expected treatment outcome for a patient if the patient were treated with the treatment option corresponding to the model. Each decision tree model may be trained on a number of previous patient characteristics and the associated treatment outcomes (e.g., patient treatment outcomes) associated with such characteristics. Previous patients may be, for example, patients who participated in a clinical trial or real-world patients (i.e., patients whose treatment outcomes were determined outside of a clinical trial).

[0007] In some examples, the method may include: providing a plurality of nucleic acid molecules obtained from a sample from a subject having a disease; ligating one or more adapters onto one or more nucleic acid molecules from the plurality of nucleic acid molecules; amplifying one or more ligated nucleic acid molecules from the plurality of nucleic acid molecules; capturing amplified nucleic acid molecules from the amplified nucleic acid molecules; sequencing the captured nucleic acid molecules with a sequencer to obtain a plurality of sequence readings representing the captured nucleic acid molecules; generating a genome profile including sequence reading analysis data based on the sequence readings with one or more processors; identifying subject characteristics including biomarker values ​​for one or more loci based on the sequence reading analysis data using one or more processors; receiving a plurality of additional subject characteristics for the subject in one or more processors; accessing a tree-based model corresponding to treatment options for the disease using one or more processors, wherein the tree-based model is generated based on a plurality of previous patient characteristics and relevant treatment outcomes for the corresponding treatment options; and determining an expected treatment outcome for the subject from the plurality of subject characteristics and the tree-based model, provided that the subject is treated with the corresponding treatment option, using one or more processors.

[0008] In some implementations, the method includes: receiving multiple sample-related sequence readings from a subject in one or more processors; analyzing the multiple sequence readings using one or more processors to generate a genomic profile including sequence reading analysis data; identifying subject characteristics including biomarker values ​​for one or more loci based on the sequence reading analysis data using one or more processors; receiving multiple additional subject characteristics about the subject in one or more processors; accessing a tree-based model corresponding to treatment options for a disease using one or more processors, the tree-based model being generated based on multiple previous patient characteristics and relevant treatment outcomes for the corresponding treatment options; and determining an expected treatment outcome for the subject from the multiple subject characteristics and the tree-based model, provided that the subject is treated with the corresponding treatment option.

[0009] In some cases, determining the expected treatment outcome involves weighting multiple nodes in each tree-based model based on commonalities between multiple target characteristics and characteristics of multiple previous patients.

[0010] In some examples, each tree-based model is a tree-based regression model. In some examples, each tree-based model is an ensemble tree model. In some examples, each tree-based model is a Bayesian additive regression tree model.

[0011] In some cases, the treatment outcomes for at least the first portion of multiple previous patients are the treatment outcomes determined during one or more clinical trials.

[0012] In some cases, treatment outcomes and expected treatment outcomes include overall survival, progression-free survival, tumor response, changes in ctDNA levels in the subject, disease remission, or resistance to treatment.

[0013] In some examples, multiple target characteristics include one or more of age, sex, race, year of cancer diagnosis, time since cancer diagnosis, time since metastatic disease diagnosis, biomarker values, comorbidities, list of prior therapies, baseline ctDNA values, clinic environment, or general condition, and multiple prior patient characteristics include one or more of age, sex, race, year of cancer diagnosis, time since cancer diagnosis, time since metastatic disease diagnosis, biomarker values, comorbidities, list of prior therapies, baseline ctDNA values, clinic environment, or general condition, and at least a portion of the multiple target characteristics are of the same type as at least a portion of the multiple prior patient characteristics for at least a portion of the prior patients.

[0014] In some examples, biomarker values ​​for multiple target and patient characteristics include EGFR variant status, NTRK variant status, RET variant status, ALK rearrangement variant status, tumor mutational burden (TMB), HLA heterozygosity loss (LOH) status, germline status of variants, somatic status of variants, microsatellite instability (MSI) status, homologous recombination repair deficiency (HRD) status, whole-genome heterozygosity loss (gLOH) status, copy number variation (CNA) status, PD-L1 expression level status, hormone receptor status, PSA expression level status, or PSA expression level increase rate.

[0015] In some cases, the disease is cancer. For example, cancers include B-cell carcinoma (multiple myeloma), melanoma, breast cancer, lung cancer, bronchial cancer, colorectal cancer, prostate cancer, pancreatic cancer, stomach cancer, ovarian cancer, bladder cancer, brain cancer, central nervous system cancer, peripheral nervous system cancer, esophageal cancer, cervical cancer, uterine cancer, endometrial cancer, oral cancer, pharyngeal cancer, liver cancer, kidney cancer, testicular cancer, biliary tract cancer, small intestine cancer, appendiceal cancer, salivary gland cancer, thyroid cancer, adrenal cancer, osteosarcoma, chondrosarcoma, and hematological cancer. Cancer, adenocarcinoma, inflammatory myofibroblastoma, gastrointestinal stromal tumor (GIST), colon cancer, multiple myeloma (MM), myelodysplastic syndrome (MDS), myeloproliferative disorder (MPD), acute lymphoblastic leukemia (ALL), acute myeloid leukemia (AML), chronic myeloid leukemia (CML), chronic lymphocytic leukemia (CLL), polycythemia vera, Hodgkin lymphoma, non-Hodgkin lymphoma (NHL), soft tissue sarcoma, fibrosarcoma, myxosarcoma, fatty tissue Tumors, osteosarcoma, chordoma, angiosarcoma, endosarcoma, lymphangiosarcoma, lymphangioendosarcoma, synoviomas, mesothelioma, Ewing's tumor, leiomyosarcoma, rhabdomyosarcoma, squamous cell carcinoma, basal cell carcinoma, adenocarcinoma, sweat gland carcinoma, sebaceous gland carcinoma, papillary carcinoma, papillary adenocarcinoma, medullary carcinoma, bronchogenic carcinoma, renal cell carcinoma, liver cancer, bile duct cancer, choriocarcinoma, seminocarcinoma, embryonal carcinoma, Wilms' tumor, bladder cancer, epithelial carcinoma, glioma, astrocytoma, medulloblastoma, craniopharyngioma It may be an ependymoma, pineal cell tumor, glioblastoma, acoustic neuroblastoma, oligodendroglioma, meningioma, neuroblastoma, retinoblastoma, follicular lymphoma, diffuse large B-cell lymphoma, mantle cell lymphoma, hepatocellular carcinoma, thyroid cancer, gastric cancer, head and neck cancer, small cell carcinoma, essential thrombocythemia, aplastic myelogenesis, eosinophilic syndrome, systemic mastocytosis, familial eosinophilia, chronic eosinophilic leukemia, neuroendocrine carcinoma, or carcinoid tumor.

[0016] In some examples, the method further includes obtaining a sample from the subject. In some examples, the sample includes a tissue biopsy sample, a liquid biopsy sample, or a normal control. In some examples, the sample is a liquid biopsy sample and includes blood, plasma, cerebrospinal fluid, sputum, stool, urine, or saliva. In some examples, the sample is a liquid biopsy sample and includes circulating tumor cells (CTCs). In some examples, the sample is a liquid biopsy sample and includes cell-free DNA (cfDNA), circulating tumor DNA (ctDNA), or any combination thereof. In some examples, the multiple nucleic acid molecules may include a mixture of tumor nucleic acid molecules and non-tumor nucleic acid molecules. In some examples, the tumor nucleic acid molecules are derived from the tumor portion of the heterogeneous tissue biopsy sample, and the non-tumor nucleic acid molecules are derived from the normal portion of the heterogeneous tissue biopsy sample.

[0017] In some cases, the sample includes a liquid biopsy sample, tumor nucleic acid molecules originate from the circulating tumor DNA (ctDNA) fraction of the liquid biopsy sample, and non-tumor nucleic acid molecules originate from the non-tumor cell-free DNA (cfDNA) fraction of the liquid biopsy sample.

[0018] In some examples, one or more adapters include amplification primers, flow cell adapter sequences, substrate adapter sequences, or sample index sequences.

[0019] In some cases, the captured nucleic acid molecule is captured from a nucleic acid molecule amplified by hybridization to one or more bait molecules. In some cases, one or more bait molecules contain one or more nucleic acid molecules, each nucleic acid molecule containing a region complementary to the region of the captured nucleic acid molecule.

[0020] In some cases, amplifying nucleic acid molecules may involve performing polymerase chain reaction (PCR) amplification techniques, non-PCR amplification techniques, or isothermal amplification techniques.

[0021] In some examples, sequencing involves the use of massively parallel sequencing (MPS) techniques, whole-genome sequencing (WGS), whole-exome sequencing, targeted sequencing, direct sequencing, or Sanger sequencing techniques. In some examples, sequencing includes massively parallel sequencing, and massively parallel sequencing techniques include next-generation sequencing (NGS).

[0022] In some examples, the sequencer includes next-generation sequencers.

[0023] In some cases, one or more of the sequence readings overlap with one or more gene loci within one or more subgenome segments in the sample.

[0024] In some examples, the method further includes generating a report containing a list of expected treatment outcomes for subjects if they were treated with the corresponding treatment options. In some examples, the method includes, for example, sending the report to a healthcare provider. In some examples, the report is sent via a computer network or peer-to-peer connection.

[0025] A method (which may be a computer-implemented method) for determining the expected treatment outcome for a subject having a disease (such as cancer) includes, in one or more processors, receiving a plurality of subject characteristics for the subject; accessing, using one or more processors, a tree-based model (such as a tree-based regression model like an ensemble regression tree model), where the tree-based model may be stored on a memory configured to be accessed by one or more processors, corresponding to treatment options for the disease, and the tree-based model is generated based on characteristics of a plurality of previous patients and the associated treatment outcomes corresponding to the treatment options; and determining, using one or more processors, from the plurality of subject characteristics and the tree-based model, the expected treatment outcome for the subject, conditional on the subject being treated by the treatment option to which the subject corresponds. The characteristics of the plurality of previous patients and the associated treatment outcomes may be determined during a clinical trial, or outside of a clinical trial (i.e., real-world data), or both. For example, the tree-based model may be fitted using real-world data, and the nodes of the tree-based model may be weighted or determined using clinical trial data. In some embodiments, the tree-based model is a Bayesian additive regression tree model.

[0026] A method (which may be a computer implementation) for determining the expected treatment outcome (e.g., one, two, or more treatment options) for a subject with a disease (such as cancer) may include: receiving multiple subject characteristics for the subject in one or more processors; accessing two or more tree-based models (e.g., tree-based regression models such as ensemble regression tree models) using one or more processors, wherein the tree-based models may be stored in memory configured to be accessed by one or more processors, each tree-based model corresponding to a treatment option for the disease, and each tree-based model is generated based on multiple previous patient characteristics and associated treatment outcomes for the corresponding treatment options; and determining the expected treatment outcome for the subject, on the condition that the subject is treated with the corresponding treatment option, using one or more processors, based on the multiple subject characteristics and the two or more tree-based models. Multiple previous patient characteristics and associated treatment outcomes may be determined during a clinical trial, outside of a clinical trial (i.e., real-world data), or both. For example, the tree-based models may be fitted using real-world data, and the nodes of the tree-based models may be weighted or determined using clinical trial data. In some embodiments, the tree-based model is a Bayesian additive regression tree model.

[0027] A method of selecting a treatment for a subject having a disease (such as cancer), for example, a treatment selected from two treatment options or three or more treatment options, includes, in one or more processors, receiving a plurality of subject characteristics for the subject, and using the one or more processors to access two or more tree-based models (such as a tree-based regression model such as an ensemble regression tree model), where the tree-based models can be stored on a memory configured to be accessed by the one or more processors, each tree-based model corresponding to a treatment option for the disease, and each tree-based model being generated based on characteristics of a plurality of previous patients and associated treatment outcomes corresponding to the treatment options, accessing, using the one or more processors, from the plurality of subject characteristics and the tree-based models, determining an expected treatment outcome for the subject on condition that the subject is treated by the treatment option the subject corresponds to, and selecting, based on the expected treatment outcome of each treatment option, the treatment option that is most likely to be effective in treating the disease in the subject. The characteristics of the plurality of previous patients and the associated treatment outcomes can be determined during a clinical trial, or outside of a clinical trial (i.e., real-world data), or both. For example, a tree-based model can be fitted using real-world data, and the nodes of the tree-based model can be weighted or determined using clinical trial data. In some embodiments, the tree-based model is a Bayesian additive regression tree model.

[0028] In some embodiments, determining the expected treatment outcome includes weighting a plurality of nodes in each tree-based model based on the commonality between the plurality of subject characteristics and the characteristics of the plurality of previous patients.

[0029] Uncertainty in the tree-based model is considered using a Markov chain Monte Carlo method, a bagging process, or a non-Bayesian model. In some embodiments, uncertainty in the tree-based model is considered using a Markov chain Monte Carlo method.

[0030] Treatment outcomes and / or expected treatment outcomes may include, for example, overall survival, progression-free survival, tumor response, changes in ctDNA levels in the subject, disease remission, or resistance to treatment.

[0031] For example, the expected treatment outcome could be a distribution of expected treatment outcome likelihoods, such as a distribution of expected treatment outcome likelihoods over multiple periods.

[0032] In some embodiments, the disease is cancer, muscle sclerosis, or end-stage renal disease. In some embodiments, the disease is cancer. In some embodiments, the treatment options include cancer immunotherapy options, chemotherapy options, or both.

[0033] Multiple subject characteristics may include one or more of the following: age, sex, race, year of cancer diagnosis, time since cancer diagnosis, time since metastatic disease diagnosis, biomarker values, comorbidities, list of previous therapies, baseline ctDNA value, clinic environment, or general health. Multiple prior patient characteristics may include one or more of the following: age, sex, race, year of cancer diagnosis, time since cancer diagnosis, time since metastatic disease diagnosis, biomarker values, comorbidities (e.g., diabetes or heart disease), list of previous therapies (optionally, a list of the time spent on each previous therapy), baseline ctDNA value, clinic environment (e.g., classified as academic or social environment), or general health. At least a portion of the multiple subject characteristics may be of the same type as at least a portion of the multiple prior patient characteristics for at least a portion of the prior patients. Exemplary biomarker values ​​may include EGFR variant status, NTRK variant status, RET variant status, ALK rearrangement variant status, tumor mutational burden (TMB), HLA heterozygosity loss (LOH) status, germline status of variants, somatic status of variants, microsatellite instability (MSI) status, homologous recombination repair deficiency (HRD) status, whole-genome heterozygosity loss (gLOH) status, copy number variation (CAN) status, PD-L1 expression level status, hormone receptor status, PSA expression level status, or PSA expression level increase rate.

[0034] Any of the above methods may further include displaying on an electronic display the treatment option that is most likely to be effective in treating the disease in the subject, or the expected treatment outcome for one or more treatment options.

[0035] Any of the above methods may further include providing the subject or the subject's healthcare provider with a report (which may be an electronic or non-electronic report) indicating the treatment option most likely to be effective in treating the disease in the subject, or the expected treatment outcome for one or more treatment options.

[0036] This specification also describes a system (e.g., one or more electronic devices) including one or more processors and a memory for storing one or more programs configured to be executed by the one or more processors, wherein one or more programs include instructions for performing any of the methods described above.

[0037] For example, the system may include one or more processors and memory storing one or more programs configured to run by one or more processors, the one or more programs including instructions for determining expected treatment outcomes for subjects having a disease (such as cancer), the program receiving multiple subject characteristics for a subject in one or more processors, accessing a tree-based model (e.g., a tree-based regression model such as an ensemble regression tree model) using one or more processors, the tree-based model may be stored in memory, corresponding to treatment options for the disease, and the tree-based model is generated based on the characteristics of multiple previous patients and associated treatment outcomes for the corresponding treatment options, and the system may include determining, using one or more processors, the expected treatment outcomes for a subject if the subject were treated with the corresponding treatment options, from the multiple subject characteristics and the tree-based model. The characteristics of multiple previous patients and associated treatment outcomes may be determined during a clinical trial, outside of a clinical trial (i.e., real-world data), or both. For example, the tree-based model may be fitted using real-world data, and the nodes of the tree-based model may be weighted or determined using clinical trial data. In some embodiments, the tree-based model is a Bayesian additive regression tree model.

[0038] The system may include one or more processors and memory storing one or more programs configured to run by one or more processors, the one or more programs including instructions for determining expected treatment outcomes (e.g., one or two treatment options, or three or more treatment options) for subjects having a disease (e.g., cancer), the program may include receiving multiple subject characteristics for a subject in one or more processors, accessing two or more tree-based models (e.g., tree-based regression models such as ensemble regression tree models) using one or more processors, the tree-based models may be stored in memory, each tree-based model corresponding to a treatment option for a disease, each tree-based model being generated based on multiple previous patient characteristics and associated treatment outcomes for the corresponding treatment options, and using one or more processors to determine the expected treatment outcome for a subject if the subject were treated with the corresponding treatment option from the multiple subject characteristics and tree-based models. Multiple previous patient characteristics and associated treatment outcomes may be determined during a clinical trial, or outside of a clinical trial (i.e., real-world data), or both. For example, tree-based models can be fitted using real-world data, and the nodes of a tree-based model can be weighted or determined using clinical trial data. In some embodiments, the tree-based model is a Bayesian additive regression tree model.

[0039] The system may include one or more processors and memory storing one or more programs configured to run by one or more processors, wherein one or more programs include instructions for selecting a treatment for a subject having a disease (such as cancer) (e.g., a treatment selected from two or more treatment options), and the program may include receiving multiple subject characteristics about the subject in one or more processors, accessing two or more tree-based models (e.g., tree-based regression models such as ensemble regression tree models) using one or more processors, wherein the tree-based models may be stored in memory, each tree-based model corresponding to a treatment option for the disease, and each tree-based model is generated based on multiple previous patient characteristics and relevant treatment outcomes for the corresponding treatment options, and using one or more processors to determine the expected treatment outcome for the subject if the subject were treated with the corresponding treatment option from the multiple subject characteristics and tree-based models, and selecting the treatment option that is most likely to be effective in treating the disease in the subject based on the expected treatment outcomes of each treatment option. The characteristics and associated treatment outcomes of multiple previous patients can be determined during the clinical trial, outside of the clinical trial (i.e., using real-world data), or both. For example, a tree-based model may be fitted using real-world data, and the nodes of the tree-based model may be weighted or determined using clinical trial data. In some embodiments, the tree-based model is a Bayesian additive regression tree model.

[0040] This specification also describes a method for treating a subject with a disease, which may include selecting a treatment for the subject according to one of the methods described above, and administering to the subject the treatment option that is most likely to be effective in treating the disease. For example, a method of treating a subject may include selecting a treatment for a subject with a disease (such as cancer) (e.g., a treatment selected from two or more treatment options), and may include receiving multiple subject characteristics about the subject in one or more processors; accessing two or more tree-based models (e.g., tree-based regression models such as ensemble regression tree models) using one or more processors, wherein the tree-based models may be stored in memory, each tree-based model corresponding to a treatment option for the disease, and each tree-based model is generated based on the characteristics of multiple previous patients and the associated treatment outcomes for the corresponding treatment options; determining the expected treatment outcome for the subject if the subject were treated with the corresponding treatment option from the multiple subject characteristics and tree-based models using one or more processors; selecting the treatment option that is most likely to be effective in treating the disease in the subject based on the expected treatment outcomes of each treatment option; and administering the treatment option that is most likely to be effective in treating the disease to the subject. The characteristics of multiple previous patients and the associated treatment outcomes may be determined during a clinical trial, outside of a clinical trial (i.e., real-world data), or both. For example, tree-based models can be fitted using real-world data, and the nodes of a tree-based model can be weighted or determined using clinical trial data. In some embodiments, the tree-based model is a Bayesian additive regression tree model. [Brief explanation of the drawing]

[0041] [Figure 1] This document describes exemplary methods for fitting ensemble tree regression models according to several embodiments. [Figure 2A]This document illustrates the exemplary use of real-world data-generated ensemble tree regression models, used to analyze clinical trial data and populate tree nodes, in several embodiments. [Figure 2B] This document illustrates how clinical trial data can be used to weight nodes, based on several embodiments. [Figure 3] We demonstrate several embodiments of a learning loop that generates updated tree-based models based on real-world data. [Figure 4] This document presents an exemplary process for determining the expected treatment outcome for a patient with a disease. [Figure 5A] Examples of computing devices according to several embodiments are shown. [Figure 5B] An example of a computing system according to one embodiment is shown. [Modes for carrying out the invention]

[0042] The methods described herein may be useful in determining expected treatment outcomes for subjects with a disease or in selecting treatment options for a subject. For example, a subject or healthcare provider for a subject may consider whether a treatment option is worth the risk of adverse events or treatment costs, and informing a subject that a treatment is likely to be successful can be a significant advantage in providing effective treatment. In some situations, a subject or healthcare provider for a subject may determine which of two or more different treatment options will treat the disease most effectively. For example, the mere fact that a clinical trial demonstrated that a first treatment option is more effective than a second treatment option in treating the disease across the entire clinical trial population does not necessarily mean that the first treatment option is more likely to be effective than the second treatment option for a particular subject undergoing treatment. The methods described herein can provide a personal prediction of one or more treatment options, allowing subjects or healthcare providers to select the most suitable treatment option for a subject based on the characteristics of the subject.

[0043] The methods described herein are robust and can be used for a wide range of diseases. For example, these methods can be used when the disease is cancer, muscle sclerosis, or end-stage renal disease. In some embodiments, the disease is cancer.

[0044] An exemplary computer implementation for determining expected treatment outcomes for a subject with a disease may include: receiving multiple subject characteristics about the subject in one or more processors; using one or more processors to access a tree-based model corresponding to treatment options for the disease, where the tree-based model is generated based on the characteristics of multiple previous patients and the relevant treatment outcomes for the corresponding treatment options; and using one or more processors to determine the expected treatment outcome for the subject if the subject is treated with the corresponding treatment option, based on the multiple subject characteristics and the tree-based model. This prediction method can be extended to determine expected treatment outcomes for multiple different treatment options. For example, a computer implementation for determining expected treatment outcomes for a subject with a disease may include: receiving multiple subject characteristics about the subject in one or more processors; using one or more processors to access two or more tree-based models, where each tree-based model corresponds to a treatment option for the disease, where each tree-based model is generated based on the characteristics of multiple previous patients and the relevant treatment outcomes for the corresponding treatment options; and using one or more processors to determine the expected treatment outcome for the subject if the subject is treated with the corresponding treatment option, based on the multiple subject characteristics and the tree-based model.

[0045] Knowing the expected treatment outcome or two or more different treatment options is useful in selecting the treatment outcome that is most likely to be effective in treating the disease in a subject. For example, a method for selecting treatment for a subject with a disease may include determining the expected treatment outcome for the subject for two or more treatment options, and, based on the expected treatment outcome for each treatment option, selecting the treatment option that is most likely to be effective in treating the disease in the patient. The expected treatment outcomes for a subject for two or more treatment options can be determined by using a computer implementation method that includes, for example, receiving multiple subject characteristics for the subject in one or more processors; accessing two or more tree-based models using one or more processors, each tree-based model corresponding to a treatment option for the disease, and each tree-based model being generated based on multiple previous patient characteristics and relevant treatment outcomes for the corresponding treatment options; and determining the expected treatment outcome for the subject if the subject were treated with the corresponding treatment option, from the multiple subject characteristics and tree-based models using one or more processors.

[0046] Once a treatment option is selected for a patient, the patient can be treated for their illness by administering that treatment option.

[0047] definition Unless otherwise defined, all technical terms used herein have the same meaning as those generally understood by those skilled in the art of the field to which this disclosure pertains.

[0048] Where used herein, the singular forms “a,” “an,” and “the” include plural references unless otherwise explicitly indicated by the context. Any reference to “or” herein includes “and / or” unless otherwise specified.

[0049] In this specification, references to values ​​or parameters "about" include (and describe) variations relating to the value or parameter itself. For example, the statement "about X" includes the statement "X".

[0050] The terms "individual," "patient," and "subject" are used synonymously and refer to mammals.

[0051] The aspects and variations of the Invention described herein are understood to include "consisting of" and / or "essentially consisting of." In this specification, the terms "comprising" (and any form or variation of "comprising," such as "comprise" and "comprises"), "having" (and any form or variation of "having," such as "have" and "has"), "including" (and any form or variation of "including," such as "includes" and "include"), or "containing" (and any form or variation of "containing," such as "contains" and "contain") are inclusive or open-ended and do not omit any additives, components, integers, elements, or method steps not further described.

[0052] When used herein, the term “subgenome interval” (or “subgenome sequence interval”) refers to a portion of a genome sequence.

[0053] When used herein, the term "target interval" refers to a subgenome interval or an expressed subgenome interval (e.g., a transcribed sequence of a subgenome interval).

[0054] The terms “variant sequence” or “variant,” as used herein, are interchangeable and refer to a modified nucleic acid sequence relative to the corresponding “normal” or “wild-type” sequence. In some examples, a variant sequence may be a “short variant sequence” (or “short variant”), i.e., a variant sequence less than approximately 50 base pairs in length.

[0055] Where a range of values ​​is provided, it should be understood that each intermediate value between the upper and lower limits of that range, and any other stated or intermediate values ​​within that range, are included within the scope of this disclosure. If the stated range includes an upper or lower limit, the range excluding either of those limits is also included within this disclosure.

[0056] The section headings used herein are for organizational purposes only and should not be construed as limiting the subject matter described herein. Descriptions are presented to enable those skilled in the art to manufacture and use the invention and are provided in the context of a patent application and its requirements. Various modifications to the embodiments described herein will be readily apparent to those skilled in the art, and the general principles herein may be applied to other embodiments. Therefore, the invention is not intended to be limited to the embodiments shown, but rather should be given the broadest scope consistent with the principles and features described herein.

[0057] The diagram illustrates the process according to various embodiments. In the exemplary process, some blocks are combined at will, the order of some blocks is changed at will, and some blocks are omitted at will. In some examples, additional steps can be performed in combination with the exemplary process. Therefore, the actions illustrated (and described in more detail below) are illustrative in nature and should not be considered limiting.

[0058] All publications, patents, and patent application disclosures referenced herein are incorporated herein by reference in their entirety. In the event of any conflict between the references incorporated herein and this disclosure, this disclosure shall prevail.

[0059] Treatment outcome prediction model The methods described herein can use tree-based models and multiple subject characteristics for a subject to predict the treatment outcome when a given patient is treated with a corresponding treatment option (i.e., to determine the expected treatment outcome for a corresponding treatment option). Each of the tree-based models can correspond to a given treatment option, and multiple tree-based models can be used to compare expected treatment outcomes to determine, for example, the treatment option that is most likely to be effective in treating the disease in the subject.

[0060] Tree-based models can be generated based on the characteristics of multiple previous patients and the relevant treatment outcomes for corresponding treatment options. The characteristics of the multiple previous patients do not need to strictly match the characteristics of the multiple target patients, but it is preferable that the sets of characteristics overlap to some extent.

[0061] A tree-based model(s) can be a tree-based regression model. In some embodiments, the tree-based model is an ensemble tree model (e.g., an ensemble tree-based regression model). For example, a tree-based model can be a Bayesian additive regression tree (BART) model. An exemplary Bayesian additive regression model is described in Chipman et al., BART: Bayesian additive regression trees, Ann. Appl. Stat., vol. 4, no. 1, pp. 266-298 (2010). See also Athey et al., Generalized random forests, Annals of Statistics, vol. 47, no. 2, pp. 1148-1178 (2019).

[0062] Data about previous patients (e.g., characteristics of multiple previous patients or previous patients, along with treatment outcomes for treatment options) is used to generate a tree-based model. In some embodiments, at least a portion of the data is determined during one or more (e.g., multiple) clinical trials (i.e., “clinical trial data”). When multiple clinical trials are used for the data, the clinical trials may collect the same or different characteristics of previous patients. In some embodiments, at least a portion of the data is determined outside of clinical trials (i.e., “real-world data”). Data determined outside of clinical trials may be stored in a database that associates the characteristics of previous patients with treatment options. For example, a clinical genome database (CGDB), such as the Foundation-Flatiron Clinical Genome Database, may be used. The set of characteristics of previous patients determined during clinical trials may be the same as, different from, and at least partially overlap with, the set of characteristics of previous patients determined outside of clinical trials.

[0063] Clinical trial data and real-world data can be used differently to generate tree-based models. For example, a tree-based model may be fitted using real-world data, and the nodes of the tree-based model may be determined using clinical trial data. That is, real-world data can be used to generate a tree structure, and the tree structure is used to analyze clinical trial data and populate the tree nodes. As more real-world data is collected, the tree-based model can be re-determined using additional real-world data. Refer to Figure 1, which illustrates an exemplary method for fitting an ensemble tree regression model. Uncertainty in the tree-based model can be considered using, for example, Markov chain Monte Carlo, bagging processes, or non-Bayesian models. In some embodiments, uncertainty in the tree-based model can be considered using Markov chain Monte Carlo. Nodes in the tree-based model can be weighted based on commonalities between multiple previous patient characteristics from real-world data and multiple previous patient characteristics from clinical trial data. For example, nodes related to previous patient characteristics from real-world data and similar previous patient characteristics from clinical trial data are weighted more heavily to estimate expected treatment outcomes, while more significant discrepancies are weighted less heavily. Referring to Figures 2A and 2B, Figures 2A and 2B illustrate the use of an ensemble tree regression model generated by real-world data, which is used to analyze clinical trial data and populate the tree nodes.

[0064] Next, the generated tree-based model can be used to predict how a subject with a set of subject characteristics will respond to treatment options for the tree-based model. That is, using a tree-based model corresponding to treatment options for a disease, the expected treatment outcome for the subject if the subject were treated with the corresponding treatment option can be determined. In some embodiments, the expected treatment outcome for the subject is a weighted average of treatment outcomes for previous patients (e.g., previous patients from a clinical trial). For example, the personalized expected treatment outcome Y for a subject for treatment option z can be determined using the following: In some embodiments, the weights are normalized.

number

[0065] In some embodiments, tree-based models may incorporate targeted smoothing based on continuous measures, such as time or propensity scores. An exemplary smoothing technique is described in Starling, Jennifer E., et al. “BART with targeted smoothing: An analysis of patient-specific stillbirth risk.” The Annals of Applied Statistics 14.1(2020):28-50.). In another example, Bayesian causal forest models may incorporate targeted smoothing techniques. An exemplary smoothing technique used with Bayesian causal forests is described in Starling, Jennifer E., et al. “Targeted smooth Bayesian causal forests: An analysis of heterogeneous treatment effects for simultaneous versus interval medical abortion regimens over gestation.” arXiv preprint arXiv:1905.09405(2019).

[0066] Previous patient characteristics or subject characteristics may be any of the patient's features. Exemplary characteristics include, but are not limited to, one or more of the following: age, sex, race, year of cancer diagnosis, time since cancer diagnosis, time since metastatic disease diagnosis, biomarker values, comorbidities (e.g., diabetes or heart disease), list of previous therapies (which may optionally further include a list of the time spent on each previous therapy), baseline ctDNA values ​​(e.g., tumor fraction), clinic environment (e.g., social or academic environment for treatment), or general health status. Exemplary biomarker values ​​may include, but are not limited to, EGFR variant status, NTRK variant status, RET variant status, ALK rearrangement variant status, tumor mutational burden (TMB), loss of HLA heterozygosity (LOH) status, germline status of variants, somatic status of variants, microsatellite instability (MSI) status, homologous recombination repair deficiency (HRD) status, loss of whole-genome heterozygosity (gLOH) status, copy number variation (CAN) status, PD-L1 expression level status, hormone receptor status, PSA expression level status, or PSA expression level increase rate.

[0067] In some embodiments, previous patient data may be insufficient to determine one or more treatment options using a tree-based model, for example, when a new target prediction falls into a covariate space segment determined by a tree where the previous patient data is insufficient. In this example, the data used in the tree-based model may be selected by (1) feeding in draws from previous predictive distributions, (2) using data from previous models where those patients had similar characteristics to the new target (e.g., a pre-specified previous distribution or a data-informed previous distribution), (3) pruning the trees in the tree-based model to create sparser segments of the covariate space, or (4) removing iterations from the computation (e.g., omitting a particular "b" draw from the composite Markov chain Monte Carlo, or, in the case of a tree-based ensemble, selectively removing a violating tree from the ensemble), (5) returning negative or "no result" insights for that particular combination of covariate space segments (e.g., choosing not to return results for patient profiles that result in empty nodes), or (6) by one or more of these techniques.

[0068] The tree-based model can be updated when additional real-world data becomes available, and this can include updates using data from treated subjects. This allows for the generation of a learning loop, as shown in Figure 3.

[0069] Treatment options (i.e., two or more) can be compared using a tree-based model corresponding to each treatment option. In some embodiments, two treatment options are compared; that is, the expected treatment outcome for a subject if treated with the corresponding treatment options can be determined for two treatment options. In some embodiments, three or more treatment options are compared; that is, the expected treatment outcome for a subject if treated with the corresponding treatment options can be determined for three or more treatment options. As an example, treatment options z and z' can be compared according to the following:

number

[0070] The treatment options evaluated using the methods described herein may be any therapeutic agent, surgical procedure, or any other medical or homeopathic intervention. The treatment options may be disease type-dependent. For example, the methods may be used to select from, for example, cancer immunotherapy options, chemotherapy options, or combination therapies (e.g., a combination of cancer immunotherapy and chemotherapy) to treat a subject with cancer.

[0071] The determined expected treatment outcomes for one or more treatment options, or the treatment option most likely to be effective in treating the disease in a subject, may be displayed (e.g., using an electronic display as part of an electronic system) or reported (e.g., reported to the subject or the subject's healthcare provider). This report may be an electronic report (e.g., an electronic medical record) or a non-electronic report (e.g., a report that may be stored as a paper record or delivered to the subject or the subject's healthcare provider). In some cases, all or part of the report may be displayed in a graphical user interface of an online or web-based healthcare portal. In some cases, the report may be transmitted over a computer network or peer-to-peer connection.

[0072] Figure 4 provides a non-restrictive example of a method for determining expected treatment outcomes for a subject with a disease. Process 400 can be performed, for example, using one or more electronic devices implementing a software platform. In some examples, Process 400 is performed using a client-server system, and blocks of Process 400 are divided in any way between the server and client devices. In other embodiments, blocks of Process 400 are divided between the server and multiple client devices. Thus, while parts of Process 400 are described herein as being performed by specific devices in a client-server system, it should be understood that Process 400 is not so limited. In other embodiments, Process 400 is performed using only one client device, or only multiple client devices. In Process 400, some blocks are optionally combined, the order of some blocks is optionally changed, and some blocks are optionally omitted. In some examples, additional steps can be performed in combination with Process 400. Thus, the operations illustrated (and described in more detail below) are essentially illustrative and should not be considered limiting.

[0073] In step 402, process 400 includes receiving several subject characteristics about the subject (e.g., in one or more processors). Exemplary subject characteristics may include, but are not limited to, age, sex, race, year of cancer diagnosis, time since cancer diagnosis, time since metastatic disease diagnosis, biomarker values, comorbidities, list of prior therapies, baseline ctDNA values, clinic environment, or general health status. In 404, access is made to a tree-based model corresponding to treatment options for the disease (e.g., using one or more processors). The tree-based model is generated based on several prior patient characteristics and relevant treatment outcomes for the corresponding treatment options. Treatment outcomes may be determined, for example, during one or more clinical trials or outside of clinical trials. Exemplary treatment outcomes may include, for example, overall survival, progression-free survival, tumor response, change in ctDNA levels in the subject, disease remission, or resistance to treatment. Exemplary previous patient characteristics may include, but are not limited to, age, sex, race, year of cancer diagnosis, time since cancer diagnosis, time since metastatic disease diagnosis, biomarker values, comorbidities, list of previous therapies, baseline ctDNA values, clinic environment, or general health status. The tree-based model may be a tree-based regression model and / or an ensemble tree model, and the tree-based model may be, for example, a Bayesian additive regression model. Uncertainty in the tree-based model may be considered using Markov chain Monte Carlo, bagging processes, non-Bayesian models, or any other appropriate process. The tree-based model may be stored, for example, in a non-temporary computer-readable medium.

[0074] The expected treatment outcome for the subject is determined in 406. The prediction assumes that the patient will be treated with the corresponding treatment option. The determination is made from multiple characteristics of the subject and the tree-based model. The prediction can be made by weighting multiple nodes in each tree-based model based on commonalities between multiple subject characteristics and multiple previous patient characteristics.

[0075] A subject with a disease can be treated using the methods described herein. For example, a method for treating a subject with a disease may include determining the expected treatment outcome for the subject for two or more treatment options according to the methods described herein. From the two or more treatment options, the treatment option that is most likely to be effective in treating the disease in the subject may be selected, and the selected treatment option is administered to the subject.

[0076] Diseases and Treatment Options The methods described herein may be used to predict the treatment outcomes of diseases. Illustrative diseases may include, but are not limited to, hyperproliferative diseases (e.g., cancer), aneuploidy disorders (Down syndrome, Edwards syndrome, Patau syndrome), fragile X syndrome, or neurological disorders (e.g., amyotrophic lateral sclerosis, Alzheimer's disease, Parkinson's disease, stroke, or Guillain-Barré syndrome). In some examples, the disease is cancer. In some examples, the cancer is a solid tumor or a metastatic form thereof. In some examples, the cancer is a blood cancer, e.g., leukemia or lymphoma.

[0077] In some cases, the subject has cancer or is at risk of developing cancer. For example, in some cases, the subject has a genetic predisposition to cancer (e.g., having a mutation that increases the baseline risk of developing cancer). In some cases, the subject is exposed to environments that increase the risk of developing cancer (e.g., radiation or chemicals). In some cases, the subject needs to be monitored for the development of cancer. In some cases, the subject needs to be monitored for cancer progression or regression, for example, after treatment with cancer therapy. In some cases, the subject needs to be monitored for cancer recurrence. In some cases, the subject needs to be monitored for minimal residual disease (MRD). In some cases, the subject has been treated for cancer or is being treated. In some cases, the subject has not been treated with anti-cancer therapy (or anti-cancer treatment).

[0078] In some cases, the subject (e.g., patient) is being treated with or has been treated with one or more targeted therapies. In some cases, post-targeted therapy samples (e.g., specimens) are obtained (e.g., collected) from patients who have previously been treated with targeted therapy. In some cases, post-targeted therapy samples are samples obtained after the completion of targeted therapy.

[0079] In some cases, the patients had not been previously treated with targeted therapy. In some cases, for patients who had not been previously treated with targeted therapy, the specimens included excisions, e.g., initial excision or excision after recurrence (e.g., excision after disease recurrence following treatment).

[0080] Exemplary cancers include, but are not limited to, B-cell cancer (e.g., multiple myeloma), melanoma, breast cancer, lung cancer (non-small cell lung cancer or NSCLC, etc.), bronchial cancer, colorectal cancer, prostate cancer, pancreatic cancer, stomach cancer, ovarian cancer, bladder cancer, brain or central nervous system cancer, peripheral nervous system cancer, esophageal cancer, cervical cancer, uterine or endometrial cancer, oral or pharyngeal cancer, liver cancer, kidney cancer, testicular cancer, biliary tract cancer, small intestine or This includes adnexal cancer, salivary gland cancer, thyroid cancer, adrenal gland cancer, osteosarcoma, chondrosarcoma, hematological cancers, adenocarcinoma, inflammatory myofibroblastic tumor, gastrointestinal stromal tumor (GIST), colon cancer, multiple myeloma (MM), myelodysplastic syndrome (MDS), myeloproliferative disorder (MPD), acute lymphoblastic leukemia (ALL), acute myeloid leukemia (AML), chronic myeloid leukemia (CML), chronic lymphocytic leukemia (CLL), polycystic leukemia vera, and Hodgkin lymphoma. Non-Hodgkin lymphoma (NHL), soft tissue sarcoma, fibrosarcoma, myxosarcoma, liposarcoma, osteogenic sarcoma, chordoma, angiosarcoma, endosarcoma synoviomas, mesothelioma, Ewing's tumor, leiomyosarcoma, rhabdomyosarcoma, squamous cell carcinoma, basal cell carcinoma, adenocarcinoma, sweat gland carcinoma, sebaceous gland carcinoma, papillary carcinoma, papillary adenocarcinoma, medullary carcinoma, bronchogenic carcinoma, renal cell carcinoma, hepatocellular carcinoma, cholangiocarcinoma, choriocarcinoma, seminoma, embryonic carcinoma, Wilms' tumor, bladder cancer, epithelial carcinoma, glioma, This includes astrocytoma, medulloblastoma, craniopharyngioma, ependymoma, pineal glandoma, hemangioblastoma, acoustic neuroma, oligodendroglioma, meningioma, neuroblastoma, retinoblastoma, cellular lymphoma, mantle cell lymphoma, hepatocellular carcinoma A, thyroid cancer, gastric cancer, head and neck cancer, small cell carcinoma, essential thrombocythemia, agnogenous myeloid metaplasia, hypereosinophilic syndrome, systemic mastocytosis, the familiar hypereosinophilia, chronic eosinophilic leukemia, neuroendocrine carcinoma, and cancerous tumors.

[0081] In some cases, cancer is a hematological malignancy (or pre-malignancy). As used herein, hematological malignancy refers to tumors of hematopoietic or lymphoid tissue, such as tumors affecting the blood, bone marrow, or lymph nodes. Exemplary hematological malignancies include leukemia (e.g., acute lymphoblastic leukemia (ALL), acute myeloid leukemia (AML), chronic lymphocytic leukemia (CLL), chronic myeloid leukemia (CML), hairy cell leukemia, acute monocytic leukemia (AMoL), chronic myelomonocytic leukemia (CMML), juvenile myelomonocytic leukemia (JMML), or macrogranular lymphocytic leukemia), lymphoma (e.g., AIDS-associated lymphoma, cutaneous T-cell lymphoma), and Hodgkin lymphoma (e.g., classical Hodgkin lymphoma or nodular lymphocyte-dominant lymphoma). This includes, but is not limited to, primary central nervous system lymphomas (e.g., B-cell non-Hodgkin lymphomas (e.g., Burkitt lymphoma, small lymphocytic lymphoma / small cell lymphoma (CLL / SLL), diffuse large B-cell lymphoma, follicular lymphoma, immunoblastic large cell lymphoma, progenitor B-lymphoblastic lymphoma, or mantle cell lymphoma) or T-cell non-Hodgkin lymphomas (e.g., mycosis fungoides, anaplastic large cell lymphoma, or progenitor T-lymphoblastic lymphoma)), and primary central nervous system lymphomas.

[0082] Treatment options may depend on the disease being treated. Exemplary treatment options for cancer, but not limited to, include, in some cases, the use of poly(ADP-ribose) polymerase inhibitors (PARPi), kinase inhibitors (e.g., multi-kinase inhibitors or kinase-specific inhibitors), alkylating agents, antimetabolites, hormones, platinum compounds, chemotherapy, radiotherapy, targeted therapy (e.g., immunotherapy), surgery, or any combination thereof. In some implementations, treatment options may be the administration of therapeutic agents. For example, this method may be used to determine the treatment outcomes for two or more different doses of the same therapeutic agent, and the dose with the better outcome may be selected for and / or administered to the subject. In some cases, this method may be used to adjust the dose of therapeutic agent administered to the subject.

[0083] Characteristics of the target patient and characteristics of previous patients Previous patient characteristics or subject characteristics may be any of the patient's features. Exemplary characteristics include, but are not limited to, one or more of the following: age, sex, race, year of cancer diagnosis, time since cancer diagnosis, time since metastatic disease diagnosis, biomarker values, comorbidities (e.g., diabetes or heart disease), list of previous therapies (which may optionally further include a list of the time spent on each previous therapy), baseline ctDNA values, clinic environment (e.g., social or academic environment for treatment), or general health status. Exemplary biomarker values ​​may include, but are not limited to, EGFR variant status (e.g., presence or absence of tumor-associated fusions), NTRK variant status (e.g., presence or absence of tumor-associated fusions), RET variant status (e.g., presence or absence of tumor-associated fusions), ALK rearrangement variant status (e.g., presence or absence of tumor-associated fusions), tumor mutational burden (TMB), HLA heterozygosity loss (LOH) status, germline status of variants, somatic status of variants, microsatellite instability (MSI) status, homologous recombination repair deficiency (HRD) status, whole-genome heterozygosity loss (gLOH) status, copy number variation (CAN) status, PD-L1 expression level status, hormone receptor status, PSA expression level status, or PSA expression level increase rate. In some examples, biomarker values ​​may be the presence or absence of tumor-associated variants, such as fusion variants, in one or more genes.

[0084] Some or all of the characteristics of a previous patient and / or subject can be determined by sequencing nucleic acid molecules in biological samples obtained from the previous patient or subject, respectively. For example, biomarker values ​​and / or baseline ctDNA values ​​can be determined using sequencing of nucleic acid molecules in biological samples. Sequence readings can be used to generate sequence reading analysis data, which may be part of a genomic profile of the subject. "Sequence reading analysis data" is, but is not limited to, data determined from sequencing readings, including variant calls (e.g., fusions, rearrangements, nucleotide variants, or copy change variants), tumor mutational loadings, heterozygosity loss, microsatellite instability, or gene expression data.

[0085] In some examples, the disclosed method includes the steps of (i) obtaining a sample from a subject (e.g., a subject suspected of having cancer or determined to have cancer), (ii) extracting nucleic acid molecules (e.g., a mixture of tumor nucleic acid molecules and non-tumor nucleic acid molecules) from the sample, (iii) ligating one or more adapters (e.g., one or more amplification primers, flow cell adapter sequences, substrate adapter sequences, or sample index sequences) to the nucleic acid molecules extracted from the sample, (iv) amplifying the nucleic acid molecules (e.g., using polymerase chain reaction (PCR) amplification techniques, non-PCR amplification techniques, or isothermal amplification techniques), and (v) merging one or more bait molecules (e.g., each containing one or more nucleic acid molecules, each containing a region complementary to the region of the captured nucleic acid molecule). (vi) capturing nucleic acid molecules from nucleic acid molecules amplified by (e.g., by hybridization), (vi) sequencing nucleic acid molecules (or library proxies derived therefrom) extracted from a sample using, for example, a next-generation (e.g., massively parallel) sequencer, using, for example, next-generation (massively parallel) sequencing technology, whole-genome sequencing (WGS) technology, whole-exome sequencing technology, targeted sequencing technology, direct sequencing technology, or Sanger sequencing technology, and (vii) generating, displaying, transmitting, and / or delivering a report (e.g., an electronic report, a web-based report, or a paper report) to a subject (or patient), caregiver, healthcare provider, physician, oncologist, electronic medical record system, hospital, clinic, practice, third-party payer, insurance company, or government agency. In some examples, the report includes output from the method described herein. In some examples, all or part of the report may be displayed in a graphical user interface of an online or web-based healthcare portal. In some examples, the report is transmitted over a computer network or peer-to-peer connection.

[0086] The disclosed method may be used with any of a variety of samples. Samples may be obtained from the subject or a previous patient to determine the characteristics of the subject and / or the characteristics of a previous patient. For example, in some cases, the sample may include a tissue biopsy sample, a liquid biopsy sample, or a normal control. In some cases, the sample may be a liquid biopsy sample and may include blood, plasma, cerebrospinal fluid, sputum, stool, urine, or saliva. In some cases, the sample may be a liquid biopsy sample and may include circulating tumor cells (CTCs). In some cases, the sample may be a liquid biopsy sample and may include cell-free DNA (cfDNA), circulating tumor DNA (ctDNA), or a combination thereof. In certain cases, the sample may be a formalin-fixed paraffin-embedded (FFPE) sample.

[0087] In some cases, specimens may be collected by tissue excision (e.g., surgical excision), needle biopsy, bone marrow biopsy, bone marrow aspiration, skin biopsy, endoscopic biopsy, fine-needle aspiration, oral swab, nasal swab, vaginal swab, or cytological smear, scraping, washing, or lavage (ductal or bronchoalveolar lavage).

[0088] In other examples, the sample may contain one or more pre-malignant or malignant cells. As used herein, pre-malignant tumor refers to cells or tissue that are not yet malignant but are ready to become malignant. In certain examples, the sample may be obtained from a solid tumor, a soft tissue tumor, or a metastatic lesion. In certain examples, the sample may be obtained from a hematological malignancy or a pre-malignant tumor. In other examples, the sample may contain tissue or cells from a surgical margin. In certain examples, the sample may contain tumor-infiltrating lymphocytes. In some examples, the sample may contain one or more non-malignant cells. In some examples, the sample may be a primary tumor or a metastasis (e.g., a metastatic biopsy sample), or a portion thereof. In some examples, the sample may be obtained from the site with the highest percentage of tumor cells (e.g., tumor site) compared to adjacent sites (e.g., sites adjacent to the tumor). In some cases, the sample may be taken from the site with the largest tumor lesion (e.g., the site with the largest number of tumor cells visualized under a microscope) compared to adjacent sites (e.g., sites adjacent to the tumor).

[0089] In some cases, nucleic acids extracted from a sample may include deoxyribonucleic acid (DNA) molecules. Examples of DNA suitable for analysis by the disclosed methods include, but are not limited to, genomic DNA or fragments thereof, mitochondrial DNA or fragments thereof, cell-free DNA (cfDNA), and circulating tumor DNA (ctDNA). Cell-free DNA (cfDNA) consists of DNA fragments released from normal and / or cancer cells during apoptosis and necrosis, circulating in the bloodstream and / or accumulating in other bodily fluids. Circulating tumor DNA (ctDNA) consists of DNA fragments released from cancer cells and tumors that circulate in the bloodstream and / or accumulate in other bodily fluids. In some cases, DNA is extracted from nucleated cells derived from the sample. In some cases, the sample may have low nucleation, for example, when the sample consists mainly of red blood cells, diseased cells containing excess cytoplasm, or tissue with fibrosis. In some cases, low nucleation may require more, for example, a larger tissue volume for DNA extraction.

[0090] In some cases, nucleic acids extracted from a sample may include ribonucleic acid (RNA) molecules. Examples of RNA suitable for analysis by the disclosed methods include, but are not limited to, total cellular RNA, total cellular RNA after depletion of a specific large amount of RNA sequence (e.g., ribosomal RNA), cell-free RNA (cfRNA), messenger RNA (mRNA) or its fragments, poly(A) tail mRNA fraction of total RNA, ribosomal RNA (rRNA) or its fragments, transfer RNA (tRNA) or its fragments, and mitochondrial RNA or its fragments. In some cases, RNA may be extracted from a sample and converted to complementary DNA (cDNA), for example, using a reverse transcription reaction. In some cases, cDNA is prepared by a random-primed cDNA synthesis method. In other cases, cDNA synthesis is initiated at the poly(A) tail of mature mRNA by priming with an oligo(dT)-containing oligonucleotide. Methods for depletion, poly(A) enrichment, and cDNA synthesis are well known to those skilled in the art.

[0091] In some cases, a sample may contain tumor cells or tumor cell nuclei. In some cases, a sample may contain tumor cell nuclei with tumor cell content of at least 5-50%, 10-40%, 15-25%, or 20-30%. In some cases, a sample may contain tumor cell nuclei with tumor cell content of at least 5%, at least 10%, at least 20%, at least 30%, at least 40%, or at least 50%. In some cases, the percentage of tumor cell nuclei is determined (e.g., calculated) by dividing the number of tumor cells in the sample by the total number of all cells in the sample that have nuclei. In some cases, for example, when the sample is a liver sample containing hepatocytes, a different tumor cell content calculation may be required due to the presence of other hepatocytes, such as non-hepatocytes, that have nuclei with twice or more the DNA content of somatic cell nuclei. In some cases, the detection sensitivity of genetic alterations, such as variant sequences, or the determination of microsatellite instability, may depend on the tumor cell content of the sample. For example, a sample with a lower tumor cell content may have lower detection sensitivity for a given sample size.

[0092] In some examples, as described above, the sample may include nucleic acids (e.g., DNA, RNA (or RNA-derived cDNA), or both) derived from tumor or normal tissue. In certain examples, the sample may further include non-nucleic acid components derived from tumor or normal tissue, such as cells, proteins, carbohydrates, or lipids.

[0093] DNA or RNA can be extracted from tissue samples, biopsy samples, blood samples, or other bodily fluid samples using any of the various techniques known to those skilled in the art (see, for example, Example 1 of International Patent Application Publication No. 2012 / 092426 (Tan, et al. (2009)), “DNA, RNA, and Protein Extraction: The Past and The Present”, J. Biomed. Biotech. 2009:574398, the technical literature for the Maxwell® 16 LEV Blood DNA Kit (Promega Corporation, Madison, WI), and the Maxwell 16 Buccal Swab LEV DNA Purification Kit Technical Manual (Promega Literature #TM333, January 1, 2011, Promega Corporation, Madison, WI)). A protocol for RNA isolation is disclosed, for example, in Maxwell® 16 Total RNA Purification Kit Technical Bulletin (Promega Literature #TB351, August 2009, Promega Corporation, Madison, WI).

[0094] A typical DNA extraction procedure includes, for example, (i) collecting a fluid, cell, or tissue sample from which DNA is to be extracted; (ii) disrupting the cell membrane (i.e., lysing the cell) as necessary to release DNA and other cytoplasmic components; (iii) treating the fluid or lysed sample with concentrated saline solution to precipitate proteins, lipids, and RNA, and then separating the precipitated proteins, lipids, and RNA by centrifugation; and (iv) purifying the DNA from the supernatant to remove detergents, proteins, salts, or other reagents used during the cell membrane lysis step.

[0095] Cell membrane disruption can be performed using various mechanical shearing techniques (e.g., by French pressing or piercing with a fine needle) or sonication techniques. The cell lysis step often involves solubilizing lipids, cell membranes, and nuclear membranes using detergents and surfactants. In some examples, the lysis step may further include degrading proteins using proteases and / or digesting RNA in the sample using RNases.

[0096] Examples of suitable techniques for DNA purification include, but are not limited to, (i) precipitation in frozen ethanol or isopropanol followed by centrifugation (DNA precipitation can be enhanced by increasing the ionic strength, for example by adding sodium acetate); (ii) phenol-chloroform extraction followed by centrifugation to separate the nucleic acid-containing aqueous phase from the protein-containing organic phase; and (iii) individual-phase chromatography in which nucleic acids are adsorbed onto individual phases (e.g., silica) depending on the pH and salt concentration of the buffer.

[0097] In some cases, DNA-bound cell and histone proteins can be removed by adding a protease, precipitating proteins with sodium acetate or ammonium acetate, or by extraction with a phenol-chloroform mixture prior to the DNA precipitation step.

[0098] In some cases, DNA can be extracted using one of a variety of suitable commercially available DNA extraction and purification kits. Examples include, but are not limited to, the QIAamp kit (for isolating genomic DNA from human samples) and the DNAeasy kit (for isolating genomic DNA from animal or plant samples) from Qiagen (Germantown, MD), or the Maxwell® and ReliaPrep® series kits from Promega (Madison, WI).

[0099] As described above, in some examples, the sample may include formalin-fixed (also known as formaldehyde-fixed or paraformaldehyde-fixed) and paraffin-embedded (FFPE) tissue specimens. For example, an FFPE specimen may be a tissue specimen embedded in a matrix, such as an FFPE block. Methods for separating nucleic acids (e.g., DNA) from formaldehyde-fixed or paraformaldehyde-fixed, paraffin-embedded (FFPE) tissues include, for example, Cronin, et al., (2004) Am J Pathol. 164(1):35-42; Masuda, et al., (1999) Nucleic Acids Res. 27(22):4436-4443; Specht, et al., (2001) Am J Pathol. 158(2):419-429; the Ambion RecoverAll® Total Nucleic Acid Isolation Protocol (Ambion, Cat. No. AM1975, September 2008); the Maxwell® 16 FFPE Plus LEV DNA Purification Kit Technical Manual (Promega Literature #TM349, February 2011); and the EZNA® FFPE DNA Kit. This is disclosed in the Handbook (OMEGA bio-tek, Norcross, GA, product numbers D3399-00, D3399-01, and D3399-02, June 2009) and the QIAamp® DNA FFPE Tissue Handbook (Qiagen, Cat. No. 37625, October 2007). For example, the RecoverAll® Total Nucleic Acid Isolation Kit uses xylene at high temperature to solubilize paraffin-embedded samples and capture nucleic acids by filtering through a glass fiber filter. The Maxwell® 16 FFPE Plus LEV DNA Purification Kit is used with the Maxwell® 16 Instrument to purify genomic DNA from 1-10 μm sections of FFPE tissue.DNA is purified using silica-clad paramagnetic particles (PMPs) and eluted at low elution volumes. The EZNA® FFPE DNA Kit uses a spin column and buffer system for genomic DNA isolation. The QIAamp® DNA FFPE Tissue Kit uses QIAamp® DNA Micro technology for the purification of genomic and mitochondrial DNA.

[0100] In some examples, the disclosed method may further include determining or obtaining a yield of nucleic acids extracted from a sample and comparing the determined value to a reference value. For example, if the determined or obtained value is less than the reference value, the nucleic acids may be amplified before proceeding to library construction. In some examples, the disclosed method may further include determining or obtaining a size (or average size) value of the nucleic acid purification in the sample and comparing the determined or obtained value to a reference value, for example, the size (or average size) of at least 100, 200, 300, 400, 500, 600, 700, 800, 900, or 1000 base pairs (bps). In some examples, one or more parameters described herein may be adjusted or selected in response to this determination.

[0101] After isolation, nucleic acids are typically dissolved in a slightly alkaline buffer, such as Tris-EDTA(TE) buffer or ultrapure water. In some cases, isolated nucleic acids (e.g., genomic DNA) can be purified or sheared using any of the various techniques well known to those skilled in the art. For example, genomic DNA can be fragmented by physical shearing, enzymatic cleavage, chemical cleavage, and other methods well known to those skilled in the art. Example 4 of International Patent Application Publication No. 2012 / 092426 describes a method for DNA shearing. In some cases, alternative methods to DNA shearing can be used to avoid the ligation step during library preparation.

[0102] In some cases, nucleic acids isolated from a sample can be used to construct a library (e.g., a nucleic acid library as described herein). In some cases, nucleic acids are fragmented using one of the methods described above, optionally repaired for strand-end damage, optionally ligated to synthetic adapters, primers, and / or barcodes (e.g., amplification primers, sequencing adapters, flow cell adapters, substrate adapters, sample barcodes or indices, and / or unique molecular identifier sequences), sizing selected (e.g., by preparation gel electrophoresis), and / or amplified (e.g., using PCR, non-PCR amplification techniques, or isothermal amplification techniques). In other cases, the fragmented and adapter-ligated nucleic acid group is used without explicit sizing or amplification prior to hybridization-based selection of target sequences. In some cases, nucleic acids are amplified by one of a variety of specific or non-specific nucleic acid amplification methods well known to those skilled in the art. In some cases, nucleic acids are amplified by whole-genome amplification methods, such as random prime strand substitution amplification. Examples of nucleic acid library preparation techniques for next-generation sequencing are described, for example, in van Dijk, et al. (2014), Exp. Cell Research 322:12-20, and in Illumina's genomic DNA sample preparation kits.

[0103] In some examples, a nucleic acid library may contain all or substantially all of the complexity of the genome. The term “substantially all” in this context actually refers to the possibility that there may be some undesirable loss of genomic complexity during the initial steps of the procedure. The methods described herein are also useful when the nucleic acid library contains a portion of the genome, for example, when the complexity of the genome is reduced by design. In some examples, any selected portion of the genome may be used in conjunction with the methods described herein. For example, in certain embodiments, the entire exome or a subset thereof is isolated. In some examples, the library may contain at least 95%, 90%, 80%, 70%, 60%, 50%, 40%, 30%, 20%, 10%, or 5% of the genomic DNA. In some examples, the library may consist of cDNA copies of the genomic DNA containing copies of at least 95%, 90%, 80%, 70%, 60%, 50%, 40%, 30%, 20%, 10%, or 5% of the genomic DNA. In certain cases, the amount of nucleic acid used to generate a nucleic acid library may be less than 5 micrograms, less than 1 microgram, or less than 500 ng, less than 200 ng, less than 100 ng, less than 50 ng, less than 10 ng, less than 5 ng, or less than 1 ng.

[0104] In some examples, a library (e.g., a nucleic acid library) includes a collection of nucleic acid molecules. As described herein, the nucleic acid molecules in a library may include target nucleic acid molecules (e.g., tumor nucleic acid molecules, reference nucleic acid molecules, and / or regulatory nucleic acid molecules; also referred herein to as the first, second, and / or third nucleic acid molecules, respectively). The nucleic acid molecules in a library may originate from a single subject or individual. In some examples, a library may include nucleic acid molecules from two or more subjects (e.g., two, three, four, five, six, seven, eight, nine, ten, twenty, thirty, or more subjects). For example, two or more libraries from different subjects can be combined to form a library having nucleic acid molecules from two or more subjects (each subject-derived nucleic acid molecule is optionally ligated to a unique sample barcode corresponding to a specific subject). In some examples, subjects are humans who have or are at risk of having cancer or tumors.

[0105] In some cases, a library (or a portion thereof) may contain one or more subgenome segments. In some cases, a subgenome segment may be a single nucleotide location, for example, where a variant of that location is associated (positively or negatively) with a tumor phenotype. In some cases, a subgenome segment contains two or more nucleotide locations. Such examples include sequences of length of at least 2, 5, 10, 50, 100, 150, 250, or more than 250 nucleotide locations. A subgenome segment may include, for example, one or more whole genes (or parts thereof), one or more exons or coding sequences (or parts thereof), one or more introns (or parts thereof), one or more microsatellite regions (or parts thereof), or any combination thereof. A subgenome segment may include all or part of a spontaneously occurring nucleic acid molecule, for example, a fragment of a genomic DNA molecule. For example, a subgenome segment may correspond to a fragment of genomic DNA subjected to a sequencing reaction. In some cases, a subgenome segment is a continuous sequence from a genomic source. In some cases, subgenome segments contain non-contiguous sequences within the genome; for example, subgenome segments in cDNA may contain exon-exon junctions formed as a result of splicing. In some cases, subgenome segments contain tumor nucleic acid molecules. In some cases, subgenome segments contain non-tumor nucleic acid molecules.

[0106] The methods described herein can be used, for example, in combination with, or as part of, a method for evaluating multiple target intervals (e.g., target sequences) or sets of target intervals from a set of genomic loci (e.g., loci or fragments thereof) described herein.

[0107] In some examples, the set of genomic loci evaluated by the disclosed method includes, in mutational form, a plurality of genes related to effects on cell division, proliferation, or survival, or related to cancer, for example, the cancers described herein.

[0108] In some examples, the set of loci evaluated by the disclosed method includes at least one, at least two, at least three, at least four, at least five, at least six, at least seven, at least eight, at least nine, at least ten, at least twenty, at least thirty, at least forty, at least fifty, at least sixty, at least seventy, at least eighty, at least ninety, at least one hundred, or more than one hundred loci.

[0109] In some examples, the selected locus (also referred to herein as the target locus or target sequence), or a fragment thereof, may include a target interval comprising a non-coding sequence, coding sequence, intragene region, or intergene region of the target genome. For example, the target interval may include a non-coding sequence or fragment thereof (e.g., a promoter sequence, an enhancer sequence, a 5' untranslated region (5'UTR), a 3' untranslated region (3'UTR), or a fragment thereof), a coding sequence or a fragment thereof, an exon sequence or a fragment thereof, or an intron sequence or a fragment thereof.

[0110] Methods described herein may involve contacting a nucleic acid library with multiple target capture reagents to select and capture multiple specific target sequences (e.g., gene sequences or fragments thereof) for analysis. In some examples, a target capture reagent (i.e., a molecule that binds to a target molecule and thereby enables the capture of the target molecule) is used to select the target segment to be analyzed. For example, the target capture reagent may be a bait molecule, e.g., a nucleic acid molecule (e.g., a DNA molecule or RNA molecule), that hybridizes to (i.e., complements) the target molecule and thereby enables the capture of the target nucleic acid. In some examples, the target capture reagent, e.g., the bait molecule (or bait sequence), is a capture oligonucleotide (or capture probe). In some examples, the target nucleic acid may be a genomic DNA molecule, an RNA molecule, an RNA-derived cDNA molecule, a microsatellite DNA sequence, etc. In some examples, the target capture reagent is suitable for solution-phase hybridization to the target. In some examples, the target capture reagent is suitable for individual-phase hybridization to the target. In some examples, the target capture reagent is suitable for both solution-phase hybridization and individual-phase hybridization to the target. The design and construction of the target capture reagent are described in detail, for example, in International Patent Application Publication No. 2020 / 236941, the entire contents of which are incorporated herein by reference.

[0111] The methods described herein enable optimized sequencing of numerous genomic loci (e.g., genes or gene products (e.g., mRNA), microsatellite loci, etc.) from samples from one or more subjects (e.g., cancer tissue samples, liquid biopsy samples, etc.) by appropriately selecting target capture reagents to select target nucleic acid molecules to be sequenced. In some examples, the target capture reagent may hybridize to a specific target locus, e.g., a specific target gene locus or fragment thereof. In some examples, the target capture reagent may hybridize to a specific group of target loci, e.g., a specific group of gene loci or fragment thereof. In some examples, multiple target capture reagents may be used, including a mixture of target-specific and / or group-specific target capture reagents.

[0112] In some examples, the number of target capture reagents (e.g., bait molecules) in a set of target capture reagents (e.g., bait sets) that contact a nucleic acid library to capture multiple target sequences for nucleic acid sequencing may be greater than 10, 50, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 1250, 1500, 1750, 2000, 3000, 4000, 5000, 10000, 25000, or 50000.

[0113] In some examples, the total length of the target capture reagent sequence may be approximately 70 to 1000 nucleotides. In one example, the length of the target capture reagent is approximately 100 to 300 nucleotides, 110 to 200 nucleotides, or 120 to 170 nucleotides. In addition to the above, intermediate oligonucleotide lengths of approximately 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 300, 400, 500, 600, 700, 800, and 900 nucleotides can be used in the method described herein. In some embodiments, oligonucleotides with approximately 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, or 230 bases can be used.

[0114] In some examples, each target capture reagent sequence may include (i) a target-specific capture sequence (e.g., a locus- or microsatellite-specific complement sequence), (ii) an adapter, primer, barcode, and / or unique molecular identifier sequence, and (iii) a universal tail on one or both ends. When the term “target capture reagent” is used herein, it may refer to a target-specific target capture sequence or the entire target capture reagent oligonucleotide containing the target-specific target capture sequence.

[0115] In some examples, the target-specific capture sequence in the target capture reagent is approximately 40 to 1000 nucleotides long. In some examples, the target-specific capture sequence is approximately 70 to 300 nucleotides long. In some examples, the target-specific sequence is approximately 100 to 200 nucleotides long. In yet other examples, the target-specific sequence is approximately 120 to 170 nucleotides long, typically 120 nucleotides long. In addition to the above, target-specific sequences of intermediate lengths, for example, approximately 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 300, 400, 500, 600, 700, 800, and 900 nucleotides long, as well as target-specific sequences of lengths between the above lengths, may also be used in the methods described herein.

[0116] In some examples, target capture reagents may be designed to select a target interval containing one or more reconstructions, such as an intron containing a genomic reconstruction. In such examples, the target capture reagent is designed to mask repetitive sequences to enhance selection efficiency. In embodiments where the reconstruction has a known ligation sequence, a complementary target capture reagent may be designed to recognize the ligation sequence to enhance selection efficiency.

[0117] In some examples, the disclosed methods may involve the use of target capture reagents designed to capture two or more different target categories, each category having a different target capture reagent design strategy. In some examples, the hybridization-based capture methods and target capture reagent compositions disclosed herein can provide capture and homogeneous coverage of a set of target sequences, while minimizing coverage of genomic sequences other than the target set of sequences. In some examples, the target sequences may include an entire exome from genomic DNA or a selected subset thereof. In some examples, the target sequences may include a large chromosomal region (e.g., an entire chromosomal arm). The methods and compositions disclosed herein provide different target capture reagents for achieving different sequencing depth and coverage patterns for a composite set of target nucleic acid sequences.

[0118] Typically, DNA molecules are used as target capture reagent sequences, but RNA molecules can also be used. In some examples, the DNA molecular target capture reagent may be single-stranded DNA (ssDNA) or double-stranded DNA (dsDNA). In some examples, RNA-DNA double helix is ​​more stable than DNA-DNA double helix and therefore potentially provides better nucleic acid capture.

[0119] In some examples, the disclosed methods include providing a selected set of nucleic acid molecules captured from one or more nucleic acid libraries (e.g., a library catch). For example, the method may include providing one or more nucleic acid libraries, each containing multiple nucleic acid molecules extracted from one or more samples derived from one or more subjects (e.g., multiple target nucleic acid molecules and / or reference nucleic acid molecules), and contacting one or more libraries (e.g., in a solution-based hybridization reaction) with one, two, three, four, five, or more than five target capture reagents (e.g., oligonucleotide target capture reagents) to form a hybridization mixture containing multiple target capture reagent / nucleic acid molecule hybrids, and separating the multiple target capture reagent / nucleic acid molecule hybrids from the hybridization mixture by contacting the hybridization mixture with a binding entity that enables the separation of the multiple target capture reagent / nucleic acid molecule hybrids from the hybridization mixture, thereby providing a library catch (e.g., a subgroup of nucleic acid molecules also selected or enriched from one or more libraries).

[0120] In some examples, the disclosed method may further include amplifying the library catch (e.g., by performing PCR). In other examples, the library catch is not amplified.

[0121] In some cases, the target capture reagent may be part of a kit that optionally includes instructions, standards, buffers, enzymes, or other reagents.

[0122] As described above, the methods disclosed herein may include the step of contacting a library (e.g., a nucleic acid library) with a plurality of target capture reagents to provide selected library target nucleic acid sequences (i.e., library catches). The contact step can be carried out, for example, by solution-based hybridization. In some examples, the method includes repeating the hybridization step for one or more further solution-based hybridizations. In some examples, the method further includes subjecting the library catches to one or more further solution hybridizations with the same or different sets of target capture reagents.

[0123] In some examples, the contact step is carried out in a solid support, for example, an array. Suitable solid supports for hybridization are described, for example, in Albert, T.Jet al. (2007) Nat. Methods 4(11):903-5, Hodges, E. et al. (2007) Nat. Genet. 39(12):1522-7, and Okou, D. et al. (2007) Nat. Methods 4(11):907-9, the contents of which are incorporated herein by reference in their entirety.

[0124] Hybridization methods that can be adapted for use in the methods described herein are described in the art, for example, in International Patent Application Publication No. 2012 / 092426. A method for hybridizing a target capture reagent to multiple target nucleic acids is described in detail in International Patent Application Publication No. 2020 / 236941, the entirety of which is incorporated herein by reference.

[0125] The methods and systems disclosed herein can be used in combination with, or as part of, methods or systems (e.g., next-generation sequencing systems) for sequencing nucleic acids to generate multiple sequence readings that overlap with one or more loci within a subgenome section in a sample, thereby determining allelic sequences at multiple loci. As used herein, “next-generation sequencing,” or “NGS,” is also referred to as “high-volume parallel sequencing,” and can be used to sequence the nucleotide sequences of individual nucleic acid molecules (e.g., as in single-molecule sequencing) or clonely expanded proxies of individual nucleic acid molecules (e.g., 10 3 , 10 4 , 10 5 This refers to any sequencing method that uses a high-throughput approach to determine molecules exceeding a certain threshold (or molecules exceeding that threshold that are sequenced simultaneously).

[0126] Next-generation sequencing methods are known in the art and are described, for example, in Metzker, M. (2010) Nature Biotechnology Reviews 11:31-46, which are incorporated herein by reference. Other examples of sequencing methods suitable for use when carrying out the methods and systems disclosed herein are described, for example, in International Patent Application Publication No. 2012 / 092426. In some examples, sequencing may include, for example, whole-genome sequencing (WGS), whole-exome sequencing, targeted sequencing, or direct sequencing. In some examples, sequencing may be performed using, for example, Sanger sequencing. In some examples, sequencing may include paired-end sequencing techniques that enable sequencing of both ends of a fragment and generate high-quality, alignable sequence data for detecting, for example, genome rearrangements, repetitive sequence elements, gene fusions, and novel transcripts.

[0127] The disclosed methods and systems may be implemented using sequencing platforms such as Roche 454, Illumina Solexa, ABI-SOLiD, ION Torrent, Complete Genomics, Pacific Bioscience, Helicos, and / or Polonator platforms. In some examples, sequencing may include Illumina MiSeq sequencing. In some examples, sequencing may include Illumina HiSeq sequencing. In some examples, sequencing may include Illumina NovaSeq sequencing. Optimized methods for sequencing multiple target genomic loci in nucleic acids extracted from a sample are described in detail, for example, in International Patent Application Publication No. 2020 / 236941, the entire contents of which are incorporated herein by reference.

[0128] In some examples, the disclosed method involves (a) obtaining a library containing multiple normal and / or tumor nucleic acid molecules derived from a sample; (b) contacting the library with one, two, three, four, five, or more than five target capture reagents under conditions that allow hybridization of the target capture reagents to target nucleic acid molecules, thereby providing a selected set of captured normal and / or tumor nucleic acid molecules (i.e., a library catch); and (c) contacting the hybridization mixture with a binding entity that allows separation of the target capture reagent / nucleic acid molecule hybrid from the hybridization mixture, for example, to obtain a selected subset of nucleic acid molecules from the hybridization mixture (e.g., (d) separating a library catch; (e) sequencing the library catch to obtain from the library catch, which may include variant sequences containing mutations (or changes), such as somatic mutations or germ cell mutations, a plurality of readings (e.g., sequence readings) that overlap with one or more target intervals (e.g., one or more target sequences); and (e) aligning the sequence readings using an alignment method described in other parts of this Spec, and / or assigning nucleotide values ​​to nucleotide positions within the target interval from one or more sequence readings (e.g., invoking mutations using a Bayesian method or other method described herein).

[0129] In some examples, obtaining sequence reads for one or more target intervals may involve sequencing at least one, at least five, at least ten, at least 20, at least 30, at least 40, at least 150, at least 200, at least 250, at least 300, at least 350, at least 400, at least 450, at least 500, at least 550, at least 600, at least 650, at least 700, at least 750, at least 800, at least 850, at least 900, at least 950, at least 1000, at least 1250, at least 1500, at least 1750, at least 2000, at least 2250, at least 2500, at least 2750, at least 3000, at least 3500, at least 4000, at least 4500, and at least 5000 loci. In some examples, obtaining sequence readings for one or more target intervals may involve sequencing the target interval for any number of loci within the range described in this paragraph, for example, at least 2850 loci.

[0130] In some examples, obtaining sequence readings for one or more target intervals involves sequencing the target interval using a sequencing method that provides sequence reading lengths (or average sequence reading lengths) of at least 20, 30, 40, 50, 60, 70, 80, 90, 100, 120, 140, 160, 180, 200, 220, 240, 260, 280, 300, 320, 340, 360, 380, or 400 bases. In some examples, obtaining sequence readings for one or more target intervals may involve sequencing the target intervals using a sequencing method that provides sequence reading lengths (or average sequence reading lengths) for any number of bases within the range described in this paragraph, for example, a sequence reading length (or average sequence reading length) for 56 bases.

[0131] In some examples, obtaining sequence readings for one or more target intervals may involve sequencing with an average coverage (or depth) of at least 100x. In some examples, obtaining sequence readings for one or more target intervals may involve sequencing with an average coverage (or depth) of at least 100x, at least 150x, at least 200x, at least 250x, at least 500x, at least 750x, at least 1000x, at least 1500x, at least 2000x, at least 2500x, at least 3000x, at least 3500x, at least 4000x, at least 4500x, at least 5000x, at least 5500x, or at least 6000x, or greater. In some examples, obtaining sequence readings for one or more target intervals may involve sequencing with an average coverage (or depth) of at least 160x, which is any value within the range described in this paragraph.

[0132] In some examples, obtaining sequence readings for one or more target intervals involves sequencing using an average sequencing depth of at least 6000x, with a mean value ranging from at least 100x to at least 6000x, representing approximately 90%, 92%, 94%, 95%, 96%, 97%, 98%, or 99% of the sequenced loci. For example, in some examples, obtaining readings for a target interval involves sequencing at least 125x for at least 99% of the sequenced loci. In another example, in some examples, obtaining readings for a target interval involves sequencing at least 4100x for at least 95% of the sequenced loci.

[0133] In some cases, the relative abundance of nucleic acid species in a library can be estimated by counting the relative number of occurrences of those congeneral sequences in the data generated by sequencing experiments (e.g., the number of sequence readings for a given congeneral sequence).

[0134] In some examples, the disclosed methods and systems provide nucleotide sequences for a set of target intervals (e.g., loci) as described herein. In certain examples, the sequences are provided without using methods that include matching normal controls (e.g., wild-type controls) and / or matching tumor controls (e.g., primary vs. metastatic).

[0135] In some embodiments, the sequencing depth level (e.g., X times the sequencing depth level), when used herein, refers to the number of duplicate readings, e.g., the number of readings after detection and removal of PCR duplicate readings (e.g., unique readings). In other examples, duplicate readings are evaluated, for example, to assist in the detection of copy number variations (CNAs).

[0136] Alignment is the process of matching a reading to a specific location, such as a genomic location or locus. In some cases, NGS readings can be aligned to a reference sequence (e.g., a wild-type sequence). In some cases, NGS readings can be newly assembled. Methods for NGS sequence alignment are described, for example, in Trapnell C. and Salzberg SL Nature Biotech., 2009, 27:455-457. Examples of de novo sequence assembly are described, for example, in Warren R. et al., Bioinformatics, 2007, 23:500-501; Butler J. et al., Genome Res., 2008, 18:810-820; and Zerbino DR and Birney E., Genome Res., 2008, 18:821-829. The optimization of sequence alignment is described in the Art, for example, in International Patent Application Publication No. 2012 / 092426. Further descriptions of sequence alignment methods are provided, for example, in International Patent Application Publication No. 2020 / 236941, the entire contents of which are incorporated herein by reference.

[0137] Misalignments (e.g., the placement of base pairs from short reads at inaccurate locations within the genome), such as those resulting from the sequence context of reads around actual cancer mutations (e.g., the presence of repetitive sequences), can lead to reduced sensitivity in mutation detection because the reads of alternative alleles may be shifted from the histogram peaks of alternative allele reads. Other examples of sequence contexts that can cause misalignments include short tandem repeats, scattered repeats, low-complexity regions, insertions-deletions (indels), and paralogs. When problematic sequence situations arise in the absence of actual mutations, misalignments can introduce artifact reads of "mutant" alleles by placing the actual reads of reference genome bases in the wrong location. Since mutation calling algorithms for multiplex analysis must be highly sensitive even to low-abundance mutations, these misalignments can increase the false-positive detection rate / decrease specificity.

[0138] In some examples, the methods and systems disclosed herein may integrate multiple individually tailored alignment methods or algorithms to optimize base calling performance in sequencing methods, particularly those that rely on large-scale parallel sequencing of numerous diverse genetic events at numerous diverse genomic loci. In some examples, the disclosed methods and systems may include the use of one or more global alignment algorithms. In some examples, the disclosed methods and systems may include the use of one or more local alignment algorithms.Examples of alignment algorithms that may be used include, but are not limited to, the Burrows-Wheeler Alignment (BWA) software bundle (see, e.g., Li, et al. (2009), “Fast and Accurate Short Read Alignment with Burrows-Wheeler Transform”, Bioinformatics 25:1754-60, Li, et al. (2010), “Fast and Accurate Long-Read Alignment with Burrows-Wheeler Transform”, Bioinformatics epub.PMID:20080505), the Smith-Waterman algorithm (see, e.g., Smith, et al. (1981), “Identification of Common Molecular Subsequences”, J. Molecular Biology 147(1):195-197), and the Striped Smith-Waterman algorithm (see, e.g., Farrar (2007), “Striped Smith-Waterman Speeds Database Searches Six Times Over This includes other SIMD implementations (see "Other SIMD Implementations", Bioinformatics 23(2):156-161), the Needleman-Wunsch algorithm (Needleman, et al. (1970) "A General Method Applicable to the Search for Similarities in the Amino Acid Sequence of Two Proteins", J. Molecular Biology 48(3):443-53), or any combination thereof.

[0139] In some examples, the methods and systems disclosed herein may involve the use of a sequence assembly algorithm, such as the Arachne sequence assembly algorithm (see, for example, Batzoglou, et al. (2002), “ARACHNE: A Whole-Genome Shotgun Assembler”, Genome Res. 12: 177-189).

[0140] In some cases, the alignment method used to analyze sequence readings is neither individually customized nor adjusted to detect different variants (e.g., point mutations, insertions, deletions, etc.) at different genomic loci. In some cases, different alignment methods are used to analyze readings that are individually customized or adjusted to detect at least a subset of different variants detected at different genomic loci. In some cases, different alignment methods are used to analyze readings that are individually customized or adjusted to detect each different variant at different genomic loci. In some cases, the adjustment can be a function of (i) the sequencing of a locus (e.g., a locus, microsatellite locus, or other target interval), (ii) the tumor type associated with the sample, (iii) the sequencing of variants, or (iv) the characteristics of the sample or target. The selection or use of alignment conditions individually adjusted to several specific target intervals to be sequenced allows for optimization of speed, sensitivity, and specificity. This method is particularly effective when the alignment of readings is optimized for a relatively large number of diverse target intervals. In some examples, the method involves using an alignment method optimized for reconstruction in combination with other alignment methods optimized for target intervals not related to reconstruction.

[0141] In one embodiment, the method further comprises selecting or using an alignment method for analyzing, e.g., aligning sequence readings, wherein the alignment method is a function of, or selected accordingly, or optimized for, (i) tumor type, e.g., tumor type in a sample; (ii) location of a sequenced target interval (e.g., locus); (iii) type of variant within the sequenced target interval (e.g., point mutation, insertion, deletion, substitution, copy number variation (CNV), rearrangement, or fusion); (iv) site being analyzed (e.g., nucleotide position); (v) type of sample (e.g., a sample as described herein); and / or (vi) adjacent sequences within or near the target interval being evaluated (e.g., depending on the expected tendency for misalignment of the target interval due to the presence of repetitive sequences within or near the target interval).

[0142] In some cases, the methods disclosed herein enable the rapid and efficient alignment of cumbersome readings, such as readings with rearrangements. Thus, in some cases where the readings for a target interval (e.g., a target interval or a represented target interval) include nucleotide positions with rearrangements, such as translocations, the method may include using an alignment method that is appropriately tuned and includes (i) selecting a rearrangement reference sequence for alignment with the readings, wherein the rearrangement reference sequence aligns with a rearrangement (in some embodiments, the reference sequence is not identical to the genome rearrangement), and (ii) comparing the readings with the rearrangement reference sequence, for example, aligning them.

[0143] In some cases, alternative methods may be used to align cumbersome readings. These methods are particularly effective when read alignment is optimized for a relatively large number of diverse target intervals. For example, a method for analyzing a sample may involve (i) performing a comparison of readings (e.g., alignment comparison) using a first set of parameters (e.g., by using a first mapping algorithm or by comparing with a first reference sequence) to determine whether the readings satisfy a first alignment criterion (e.g., the readings can be aligned with the first reference sequence, e.g., the number of mismatches is less than a certain number), and (ii) if the readings do not satisfy the first alignment criterion, (e.g., by using a first mapping algorithm or by comparing with a second reference sequence) (iii) performing a second alignment comparison using a second set of parameters (by comparison), and optionally determining whether the above readings satisfy the second criterion (e.g., the readings can be aligned with the above second reference sequence, e.g., the number of mismatches is less than a certain number), wherein the above second set of parameters includes, for example, the use of the above second reference sequence, and the above second reference sequence is more likely to align with the readings for variants (e.g., reconfiguration, insertion, deletion, or translocation) compared with the first set of parameters.

[0144] In some cases, the alignment of sequence readings in the disclosed method may be combined with mutation calling methods described in other parts of this Spec. As considered herein, the reduced sensitivity for detecting actual mutations can be addressed by evaluating (in a manual or automated manner) the quality of the alignment around the expected mutation site of the gene being analyzed. In some cases, the sites to be evaluated can be obtained from the human genome (e.g., the HG19 human reference genome) or a database of cancer mutations (e.g., COSMIC). Regions identified as problematic can be repaired by alignment optimization (or realignment) using a slower but more accurate alignment algorithm, such as the Smith-Waterman alignment, using an algorithm selected to give better performance in the relevant sequence context. If general alignment algorithms fail to resolve the problem, customized alignment methods can be created, for example, by adjusting the maximum difference mismatch penalty parameter for genes likely to contain substitutions, adjusting specific mismatch penalty parameters based on specific mutation types common to a particular tumor type (e.g., C→T in melanoma), or adjusting specific mismatch penalty parameters based on specific mutation types common to a particular sample type (e.g., substitutions common in FFPE).

[0145] The decrease in specificity of the evaluated target interval (increase in false positive rate) due to misalignment can be assessed by manually or automatically checking all mutation calls in the sequencing data. Regions found to be prone to false mutation calls due to misalignment can undergo the alignment rescue discussed above. If algorithmic improvements are not possible, "mutations" from the problem region can be classified or screened from a panel of target loci.

[0146] A base call refers to the raw output of a sequencing device, for example, the determined sequence of nucleotides in an oligonucleotide molecule. A mutation call refers to the process of selecting a nucleotide value, for example, A, G, T, or C, for a given nucleotide position being sequenced. Typically, a sequence reading (or base call) for a position provides two or more values, for example, some readings indicate T and some readings indicate G. A mutation call is the process of assigning the correct nucleotide value, for example, one of those values, to a sequence. Although referred to as a "mutation" call, it can be applied to assign a nucleotide value to any nucleotide position, for example, a position corresponding to a mutant allele, a wild-type allele, an allele not characterized as mutant or wild-type, or a position not characterized as variability.

[0147] In some examples, the disclosed methods may include optimizing performance by using a customized or adjusted mutation calling algorithm or its parameters when applied to sequencing data, particularly in methods that rely on large-scale parallel sequencing of numerous diverse genetic events at numerous diverse genomic loci (e.g., loci, microsatellite regions, etc.) in a sample, e.g., a sample derived from a subject with cancer. Optimization of mutation calling is described in the art, for example, in International Patent Application Publication No. 2012 / 092426.

[0148] Methods for mutation calling may include one or more of the following: making independent calls based on information at each location in the reference sequence (e.g., examining sequence reads; examining base calls and quality scores; calculating the probabilities of observed bases and quality scores given potential genotypes; and assigning genotypes (e.g., using Bayes' rules)); removing false positives (e.g., using depth thresholds to reject SNPs with read depths much lower or higher than expected; local readjustment to remove false positives due to small indels); and performing linkage disequilibrium (LD) / attribution-based analysis to refine the calls.

[0149] Formulas for calculating genotype likelihood associated with specific genotypes and locations are described, for example, in Li H. and Durbin R., Bioinformatics, 2010;26(5):589-95. Prior predictions for specific mutations in a particular cancer type can be used when evaluating samples from that cancer type. Such possibilities can be obtained from publicly available cancer mutation databases, such as the Catalogue of Somatic Mutation in Cancer (COSMIC), HGMD (Human Gene Mutation Database), The SNP Consortium, Breast Cancer Mutation Data Base (BIC), and Breast Cancer Gene Database (BCGD).

[0150] For an example of LD / imputation-based analysis, see, for example, Browning BL and Yu Z. Am. Hum. Genet. 2009, 85(6):847-61. For an example of a low-coverage SNP calling method, see, for example, Li, Y., et al., Annu. Rev. Genomics Hum. Genet. 2009, 10:387-406.

[0151] After alignment, substitution detection can be performed using a mutation calling method, such as Bayesian mutation calling. This method is applied to each base in the target interval, for example, the exon of the gene being evaluated, and the presence of alternative alleles is observed. This method compares the probability of observing read data in the presence of a mutation with the probability of observing read data in the presence of base call errors only. If this comparison strongly supports the presence of a mutation, the mutation can be called.

[0152] The advantage of Bayesian mutation detection methods is that the comparison between the probability of mutation presence and the probability of base call errors alone can be weighted by prior predictions of the presence of mutations at that site. If several reads of alternative alleles are observed at sites that frequently mutate for a given cancer type, the presence of a mutation can be reliably called even if the amount of evidence for mutation does not meet the usual threshold. This flexibility can then be used to increase the sensitivity of detecting rarer mutations / lower purity samples or to make the test more robust against reduced read coverage. The probability that a random base pair in the genome is mutated in cancer is approximately 1e-6. For example, the probability of specific mutations occurring at many sites in a typical polygenic cancer genome panel can be orders of magnitude higher. These likelihoods can be obtained from publicly available databases of cancer mutations (e.g., COSMIC).

[0153] Indel calling is the process of finding bases in sequence identification data that differ from a reference sequence due to insertions or deletions, typically including associated confidence scores or statistical evidence measures. Indel calling methods may include steps of identifying candidate indels, calculating genotype likelihood by local realignment, and performing LD-based genotype inference and calling. Typically, Bayesian methods are used to obtain potential indel candidates, which are then tested against reference sequences within a Bayesian framework.

[0154] Algorithms for generating candidate indels are described, for example, in McKenna, A., et al., Genome Res. 2010; 20(9): 1297-303, Ye, K., et al., Bioinformatics, 2009; 25(21): 2865-71, Lunter, G., and Goodson, M., Genome Res. 2011; 21(6): 936-9, and Li, H., et al. (2009), Bioinformatics 25(16): 2078-9.

[0155] Methods for generating indel calls and individual-level genotype likelihoods include, for example, the Dindel algorithm (Albers CA et al., Genome Res. 2011;21(6):961-73). For example, a Bayesian EM algorithm can be used to analyze readings, perform initial indel calls, generate genotype likelihoods for each candidate indel, and then, for example, QCALL (Le SQ and Durbin R. Genome Res. 2011;21(6):952-60) can be used to complete the genotype. Parameters such as pre-observation predictions of indels can be adjusted based on the size or location of the indels (e.g., increased or decreased).

[0156] Methods have been developed to address limited deviations from 50% or 100% allele frequencies for the analysis of cancer DNA. (See, e.g., SNVMix-Bioinformatics.2010 March 15;26(6):730-736) However, the methods disclosed herein can account for the possibility of mutant alleles being present at frequencies (or allele fractions) in the range of 1% to 100% (i.e., allele fractions in the range of 0.01 to 1.0), particularly at levels lower than 50%. This approach is especially important for detecting mutations in natural (multiclonal) tumor DNA, for example, in low-purity FFPE samples.

[0157] In some cases, the mutation calling methods used to analyze sequence readings are not individually customized or fine-tuned for detecting different mutations at different genomic loci. In some cases, different mutation calling methods are individually customized or fine-tuned for at least a subset of different mutations detected at different genomic loci. In some cases, different mutation calling methods are used that are individually customized or fine-tuned for each different mutant detected at different genomic loci. The customization or adjustment may be based on one or more of the factors described herein, such as the type of cancer in the sample, the gene or locus where the sequenced target interval is located, or the sequenced variant. By selecting or using mutation calling methods that are individually customized or fine-tuned for several target intervals to be sequenced in this manner, it becomes possible to optimize the rate, sensitivity, and specificity of mutation calling.

[0158] In some examples, a unique mutation calling method is used to assign nucleotide values ​​to nucleotide positions in each of X unique target intervals, where X is at least 2, at least 3, at least 4, at least 5, at least 10, at least 15, at least 20, at least 30, at least 40, at least 50, at least 60, at least 70, at least 80, at least 90, at least 100, at least 200, at least 300, at least 400, at least 500, at least 1000, at least 1500, at least 2000, at least 2500, at least 3000, at least 3500, at least 4000, at least 4500, at least 5000, or more. The calling method is different and can thus be unique, for example, by depending on different Bayesian prior values.

[0159] In some cases, assigning nucleotide values ​​is a function of previously observed (e.g., literature) predicted values ​​or values ​​representing variants at nucleotide positions in a type of tumor, e.g., mutations.

[0160] In some examples, the method involves assigning nucleotide values ​​(e.g., mutation calls) for at least 10, 20, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900, or 1,000 nucleotide positions, each assignment being a function of a unique value (in contrast to the values ​​of other assignments) that represents a variant at a nucleotide position in a type of tumor, e.g., a previously observed (e.g., literature) predicted value or a reading indicating a mutation.

[0161] In some examples, assigning a nucleotide value is a function of a set of values ​​that represent the probability of observing a reading indicating a variant at a nucleotide position when a variant is present in the sample at a certain frequency (e.g., 1%, 5%, 10%) and / or when the variant is not present (e.g., observed in the reading due solely to a base calling error).

[0162] In some examples, the mutation calling method described herein may include the following steps: (a) for each of the X target intervals, (i) a first value which is or represents a predicted value (e.g., from the literature) prior to observing a reading indicating a variant, e.g., a mutation, at the nucleotide position in the nucleotide position in a tumor of type X, and (ii) a second set of values ​​which represent the probability of observing a reading indicating the variant at the nucleotide position if the variant is present in the sample at a certain frequency (e.g., 1%, 5%, 10%) and / or the variant is not present (e.g., observed in the reading due to a nucleotide calling error only); and (b) in response to the above values, assigning a nucleotide value (e.g., a mutation calling) from the above reading for each of the nucleotide positions by weighting the comparison between the values ​​in the second set, e.g., by the Bayesian method described herein, using the first value (e.g., calculating the probability of the presence of a mutation), and thereby analyzing the sample.

[0163] A further description of the mutation induction method is provided, for example, in International Patent Application Publication No. 2020 / 236941, the entirety of which is incorporated herein by reference.

[0164] In some cases, a genomic profile of a subject may be generated using one or more sequence readings, and the genomic profile may further include additional data (i.e., in addition to the data generated from the sequence readings). In some cases, the genomic profile may include information about the presence of genes (or variant sequences), copy number variations, epigenetic traits, proteins (or their modifications), and / or other biomarkers in the individual's genome and / or proteome, as well as the individual's corresponding phenotypic traits, and information about the interactions between genetic or genomic traits, phenotypic traits, and environmental factors. The genomic profile of a subject may include the results of an analysis of one or more sequence readings using a comprehensive genomic profiling (CGP) test, gene expression profiling test, cancer hotspot panel test, DNA methylation test, DNA fragmentation test, RNA fragmentation test, or any combination thereof (e.g., sequence reading analysis data). Other data that illustrates the genomic profile of the subject may be included in the genomic profile. Based on the sequence reading analysis data, the genomic profile may further include information that classifies the subject and / or identifies one or more treatments used for the subject. The genomic profile may be updated to include classification and / or therapeutic information after further processing of the data has been performed, for example, based on data modeling, as described herein.

[0165] Computer systems and methods The methods described herein can be implemented using one or more computer systems. Such computer systems may include one or more programs configured to run one or more processors in order for the computer systems to perform such methods. One or more steps of the computer implementation method may be performed automatically.

[0166] In some examples, the disclosed systems may further include sequencers, such as next-generation sequencers (also known as high-speed parallel sequencers). Examples of next-generation (or high-speed parallel) sequencing platforms include, but are not limited to, the Roche 454, Illumina Solexa, ABI-SOLiD, ION Torrent, or the Pacific Bioscience sequencing platform.

[0167] Figure 5A shows an example of a computing device according to one embodiment. Device 500 can be a host computer connected to a network. Device 500 can be a client computer or a server. As shown in Figure 5, device 500 can be any suitable type of microprocessor-based device, such as a personal computer, workstation, server, or handheld computing device (portable electronic device, e.g., telephone or tablet). The device may include, for example, one or more of the following: a processor 510, an input device 520, an output device 530, storage 540, and a communication device 560. The input device 520 and the output device 530 can generally correspond to those described above and can be connected to or integrated with the computer.

[0168] The input device 520 can be any suitable device that provides input, such as a touchscreen, keyboard or keypad, mouse, or voice recognition device. The output device 530 can be any suitable device that provides output, such as a touchscreen, haptic device, or speaker. In some embodiments, the input and output devices 520 / 530 can be the same or different devices.

[0169] Storage 540 can be any suitable device that provides storage, such as electrical, magnetic, or optical memory, including RAM (volatile and non-volatile), cache, hard drive, or removable storage disk. Communication device 560 can include any suitable device that can send and receive signals over a network, such as a network interface chip or device. The computer components can be connected in any suitable way, such as via the physical bus 580 or wirelessly (e.g., Bluetooth®, Wi-Fi®, or any other wireless technology).

[0170] The software 550, which is stored in the storage 540 and can be executed by the processor 510, may include, for example, programming to embody the functions of this disclosure (for example, as embodied in the device described above).

[0171] The software 550 may also be stored and / or transferred in any non-temporary computer-readable storage medium for use by or in connection with an instruction execution system, apparatus, or device such as those described above, and may fetch and execute instructions associated with the software from the instruction execution system, apparatus, or device. In the context of this disclosure, the computer-readable storage medium may be any medium such as storage 540, and may contain or store programming for use by or in connection with an instruction execution system, apparatus, or device.

[0172] The software 550 can also be propagated by an instruction execution system, apparatus, or any other device, or in any transmission medium for use in connection with them, to fetch and execute instructions associated with the software from the instruction execution system, apparatus, or device. In the context of this disclosure, the transmission medium can be any medium that can communicate, propagate, or transmit transmission programming by or for use in connection with the instruction execution system, apparatus, or device. The transmission-readable medium may include, but is not limited to, wired or wireless transmission media of electronic, magnetic, optical, electromagnetic, or infrared.

[0173] Device 500 can be connected to a network which can be any preferred type of interconnected communication system. The network can implement any preferred communication protocol and can be protected by any preferred security protocol. The network may include network links of any preferred configuration that can implement the transmission and reception of network signals, such as wireless network connections (T1 or T3 lines), cable networks, DSL, or telephone lines.

[0174] Device 500 can implement any operating system suitable for running over a network. Software 550 can be written in any suitable programming language such as C, C++, Java®, or Python. In various embodiments, application software embodying the functions of this disclosure can be deployed in different configurations (e.g., in a client / server configuration, or via a web browser as a web-based application or web service). In some embodiments, the operating system is run by one or more processors, for example, processor 510.

[0175] The device 500 may further include a sequencer 570, which can be any suitable nucleic acid sequencing instrument.

[0176] Figure 5B shows an example of a computing system according to one embodiment. In the computing system 590, device 500 (e.g., as described above and shown in Figure 5A) is connected to network 592, and network 592 is also connected to device 594. In some embodiments, device 594 is a sequencer (e.g., a next-generation sequencer). Exemplary sequencers include, but are not limited to, Roche / 454's Genome Sequencer (GS) FLX System, Illumina / Solexa's Genome Analyzer (GA), Illumina's HiSeq 2500, HiSeq 3000, HiSeq 4000, and NovaSeq 6000 sequencing systems, Life / APG's Support Oligonucleotide Ligation Detection (SOLiD) system, Polonator's G.007 system, Helicos BioSciences' HeliScope Gene sequencing system, or Pacific BioSciences' PacBio RS system.

[0177] Devices 500 and 594 can communicate using an appropriate communication interface over a network 592, such as a local area network (LAN), a virtual private network (VPN), or the internet. In some embodiments, network 592 can be, for example, the internet, an intranet, a virtual private network, a cloud network, a wired network, or a wireless network. Devices 500 and 594 can communicate partially or entirely over wireless or wired communication such as Ethernet® or IEEE 802.11b wireless. Furthermore, devices 500 and 594 can communicate over a second network, such as a mobile / cellular network, using a suitable communication interface. Communication between devices 500 and 594 may further include, or communicate with, various servers such as mail servers, mobile servers, media servers, and telephone servers. In some embodiments, devices 500 and 594 can communicate directly (instead of, or in addition to, communication over network 592) over wireless or wired communication such as Ethernet® or IEEE 802.11b wireless. In some embodiments, devices 500 and 594 communicate via communication 596, which can be a direct connection or occur via a network (e.g., network 592).

[0178] One or all of devices 500 and 594 are generally programmed to include logic (e.g., HTTP web server logic) accessed from local or remote databases or other sources of data and content, or to format data, in order to provide and / or receive information over the network 592 in accordance with the various examples described herein.

[0179] For example, a system (e.g., an electronic device) includes one or more processors and non-temporary computer-readable memory that stores one or more programs, the one or more programs being configured to be executed by one or more processors, and the one or more programs containing instructions for performing a method to determine the expected treatment outcome for a subject having cancer. When executed by one or more processors, the one or more programs cause one or more processors to: receive a plurality of subject characteristics about the subject; access a tree-based model (which may be stored in non-temporary computer-readable memory) corresponding to treatment options for the disease, the tree-based model being generated based on a plurality of previous patient characteristics and relevant treatment outcomes for the corresponding treatment options; and determine the expected treatment outcome for the subject if the subject were treated with the corresponding treatment options from the plurality of subject characteristics and the tree-based model. In some examples, when executed by one or more programs, one or more programs cause one or more processors to generate further tree-based models.

[0180] A non-temporary computer-readable storage medium can store one or more programs, which, when executed by one or more processors of the system, cause the system to perform a method for determining expected treatment outcomes for a subject with a disease, as described herein. When executed by one or more processors, one or more programs cause one or more processors to: receive a plurality of subject characteristics about a subject; access a tree-based model (which may be stored in non-temporary computer-readable memory) corresponding to treatment options for the disease, the tree-based model being generated based on a plurality of previous patient characteristics and relevant treatment outcomes for the corresponding treatment options; and determine, from the plurality of subject characteristics and the tree-based model, the expected treatment outcome for the subject if the subject were treated with the corresponding treatment options. In some examples, when executed by one or more programs, one or more programs cause one or more processors to generate further tree-based models.

[0181] While the present disclosure and examples are fully described with reference to the accompanying drawings, it should be noted that various changes and modifications will be apparent to those skilled in the art. Such changes and modifications should be understood to be within the scope of the present disclosure and examples as defined by the claims.

[0182] The foregoing description is provided for illustrative purposes and refers to specific embodiments. However, the above exemplary considerations are not intended to be exhaustive or to limit the invention to the exact form disclosed. Many modifications and variations are possible in light of the above teachings. The embodiments have been selected and described to best illustrate the principles of the art and their practical applications. Thereafter, those skilled in the art will be able to best utilize the art and various embodiments with various modifications suitable for the specific use intended.

Claims

1. 1. A method comprising: Providing a plurality of nucleic acid molecules obtained from a sample from a subject having a disease; ligating one or more adaptors onto one or more nucleic acid molecules from said plurality of nucleic acid molecules; amplifying the one or more ligated nucleic acid molecules from the plurality of nucleic acid molecules; capturing amplified nucleic acid molecules from the amplified nucleic acid molecules; sequencing the captured nucleic acid molecule with a sequencer to obtain a plurality of sequence reads representative of the captured nucleic acid molecule; generating, by one or more processors, a genomic profile comprising sequence read analysis data based on said sequence reads; using the one or more processors to identify a characteristic of interest comprising biomarker values ​​for one or more loci based on the sequence read analysis data; receiving, at one or more processors, a plurality of additional object characteristics for the object; accessing, using the one or more processors, a tree-based model corresponding to treatment options for the disease, the tree-based model being generated based on a plurality of prior patient characteristics and associated treatment outcomes for the corresponding treatment options; and determining, using the one or more processors, from the plurality of subject characteristics and the tree-based model, a predicted treatment outcome for the subject, provided that the subject is treated with the corresponding treatment option.

2. 1. A system for determining a probable treatment outcome for a subject having a disease, the system comprising one or more processors and one or more programs, the one or more programs comprising: receiving, at the one or more processors, a plurality of object characteristics for the object; accessing, using the one or more processors, a tree-based model corresponding to treatment options for the disease, the tree-based model being generated based on a plurality of prior patient characteristics and associated treatment outcomes for the corresponding treatment options; determining, using the one or more processors, from the plurality of subject characteristics and the tree-based model, a predicted treatment outcome for the subject, provided that the subject is treated with the corresponding treatment option; The system further comprises instructions for causing the one or more processors to execute:

3. 1. A system for determining a probable treatment outcome for a subject having a disease, the system comprising one or more processors and one or more programs, the one or more programs comprising: receiving, at the one or more processors, a plurality of object characteristics for the object; accessing, using the one or more processors, two or more tree-based models, each tree-based model corresponding to a treatment option for the disease, each tree-based model generated based on a plurality of prior patient characteristics and associated treatment outcomes for the corresponding treatment option; determining, using the one or more processors, from the plurality of subject characteristics and the two or more tree-based models, a predicted treatment outcome for the subject, provided that the subject is treated with the corresponding treatment option; The system further comprises instructions for causing the one or more processors to execute:

4. 1. A system for selecting a treatment for a subject having a disease, the system comprising one or more processors and one or more programs, the one or more programs comprising: Determining a predicted treatment outcome for the subject for two or more treatment options by the system of claim 3; selecting the treatment option most likely to be effective in treating the disease in the subject based on the predicted treatment outcome for each treatment option; The system further comprises instructions for causing the one or more processors to execute:

5. 1. A method for providing a predicted treatment outcome as an indication for determining treatment options for treating a subject having a disease, comprising: Determining the predicted treatment outcomes for the subject for two or more treatment options with the system of claim 3; selecting the treatment option most likely to be effective in treating the disease in the subject based on the predicted treatment outcome for each treatment option; A method comprising:

6. 6. The method of claim 5, wherein determining the predicted treatment outcome comprises weighting a plurality of nodes in each tree-based model based on commonalities between the plurality of subject characteristics and characteristics of the plurality of prior patients.

7. The method of claim 5 , wherein each tree-based model is a tree-based regression model, an ensemble tree model, or a Bayesian additive regression tree model.

8. The method of claim 5 , wherein uncertainty in the tree-based model is accounted for using a Markov Chain Monte Carlo method, a bagging process, or a non-Bayesian model.

9. The method of claim 5 , wherein the tree-based model is populated based at least in part on draws from a prior predictive distribution.

10. The method of claim 5 , wherein the tree-based model is populated at least in part using a predetermined prior distribution or a prior distribution informed by data.

11. The method of claim 5 , wherein the tree-based model is pruned to produce a sparser partition of covariate space.

12. 6. The method of claim 5, wherein determining the expected treatment outcome comprises omitting one or more trees from the tree-based model if the one or more trees fail to provide a treatment outcome prediction.

13. 6. The method of claim 5, wherein the treatment options include cancer immunotherapy or chemotherapy treatment options, or both.

14. 6. The method of claim 5, further comprising comparing predicted treatment outcomes for two or more different treatment options.

15. 6. The method of claim 5, wherein the treatment outcome for at least a first portion of the plurality of prior patients is a treatment outcome determined during one or more clinical trials.

16. 6. The method of claim 5, wherein the treatment outcome for at least a second portion of the plurality of prior patients is a treatment outcome determined outside of a clinical trial.

17. 6. The method of claim 5, wherein the treatment outcome and the predicted treatment outcome are treatment outcomes for overall survival, progression-free survival, tumor response, changes in ctDNA levels in the subject, disease remission, or resistance to treatment.

18. 6. The method of claim 5, wherein the predicted treatment outcomes are a distribution of predicted treatment outcome likelihoods.

19. 20. The method of claim 18, wherein the predicted treatment outcome is a distribution of predicted treatment outcome likelihoods for a plurality of time periods.

20. the plurality of subject characteristics comprises one or more of age, sex, race, year of cancer diagnosis, time since cancer diagnosis, time since metastatic disease diagnosis, biomarker value, comorbidities, list of prior therapies, baseline ctDNA value, clinic environment, or performance status; the plurality of prior patient characteristics comprises one or more of age, sex, race, year of cancer diagnosis, time since cancer diagnosis, time since metastatic disease diagnosis, biomarker value, comorbidities, list of prior therapies, baseline ctDNA value, clinic setting, or performance status; The method of claim 5 , wherein at least a portion of the plurality of target characteristics is a same type of characteristic as at least a portion of the plurality of prior patient characteristics for at least a portion of the prior patients.