Method for detecting homologous recombination defects in cancer patients

Abstract

The present disclosure provides methods and compositions, e.g., kits, for detecting homologous recombination deficiencies in cancer patients. In certain embodiments, the disclosed methods are based on genome copy number summarization and deep learning models to detect homologous recombination defects from very low coverage whole genome sequencing data.

Description

[0001] Related Applications

[0002] This application claims priority to U.S. Provisional Application No. 63 / 312,823, filed on February 22, 2022, the entire disclosure of which is incorporated herein by reference. Technical Field

[0003] The present invention generally relates to the diagnosis, prognosis and treatment of cancer. In particular, the present invention relates to a method for detecting defects in the DNA homologous recombination pathway in cancer patients. Background Art

[0004] Homologous recombination is a highly accurate DNA repair mechanism in eukaryotic cells. When genes involved in the homologous recombination pathway, such as BRCA1 and BRCA2, are inactivated, defects in the homologous recombination pathway, i.e., homologous recombination defect (HRD), occur. HRD can lead to disorders in the repair of double-strand breaks in DNA, which in turn lead to tumorigenesis (see Lord, CJ, & Ashworth, A. (2012) The DNA damage response and cancer therapy. Nature, 481 (7381), 287-294; Venkitaraman, AR (2014) Cancer suppression by the chromosome custodians, BRCA1 and BRCA2. Science, 343 (6178), 1470-1475). HRD is often observed in patients with breast and ovarian cancer and has been shown to be associated with the patient's response to poly (ADP-ribose) polymerase (PARP) inhibitors such as Veliparib (Nik-Zainal, S. et al. (2016) Landscape of somatic mutations in 560 breast cancer whole-genome sequences. Nature, 534 (7605), 47-54; Coleman, RL et al. (2019) Veliparib with first-line chemotherapy and as maintenance therapy in ovarian cancer. New England Journal of Medicine, 381 (25), 2403-2415). Therefore, the identification of HRD is a good marker for selecting patients who are more likely to benefit from PARP inhibitor treatment.

[0005] Currently, there are two FDA-approved companion diagnostic tests for HRD on the market, myChoice CDx from Myriad Genetics and FoundationOne CDx from Foundation Medicine, as well as several other gene panel tests covering BRCA1 and BRCA2. Both myChoice CDx and FoundationOne CDx use deep sequencing approaches to identify mutation status of BRCA1 and BRCA2 or high genomic instability scores such as loss of heterozygosity (LOH) as positive HRD markers. These methods generally require high sequencing depth (FoundationOne CDx has a median coverage of 500), high tumor proportions in FFPE samples and matched normal tissue samples (FoundationOne CDx recommends above 25%) to have good sensitivity. New methods still need to be developed for detecting HDR in cancer patients. Summary of the invention

[0006] The present disclosure provides a method for detecting a defect in the DNA homologous recombination pathway in a cancer patient in one aspect. In some embodiments, the method includes: obtaining WGS data (whole genome sequencing data) of a tumor sample from a patient; generating a genome copy number spectrum from the WGS data; using a deep learning model to generate a chromosome instability (CI) score for each chromosome from the genome copy number spectrum, thereby generating a CI score group; generating a risk score for a defect in the DNA homologous recombination pathway from the CI score group; determining that the risk score is greater than a threshold number; and determining that the patient has a risk of a defect in the DNA homologous recombination pathway. In some embodiments, the method does not require WGS data from a patient's normal tissue / sample.

[0007] In some embodiments, the threshold value of the risk score is 0.4. In some embodiments, when the risk score is greater than 0.6, the risk of DNA homologous recombination pathway deficiency is very high. In some embodiments, when the risk score is greater than 0.4 and less than or equal to 0.6, the risk of DNA homologous recombination pathway deficiency is moderate.

[0008] In some embodiments, the deep learning model includes a long short-term memory layer and a fully connected network.

[0009] In some embodiments, the risk score is the average of a group of CI scores weighted by the base length of each chromosome, the position of each chromosome, or the position of each CNA break.

[0010] In some embodiments, a method for detecting a defect in a DNA homologous recombination pathway in a cancer patient comprises: obtaining whole genome sequencing (WGS) data of a tumor sample from the patient, wherein the WGS data has a sequencing depth of 0.05-1; generating a genomic variation spectrum from the WGS data, the genomic variation spectrum containing copy number changes, gene fusions, inversion spectrum, or translocation; generating a plurality of large-scale genomic breakpoints (LGBs) for each genome based on the genomic variation spectrum, wherein an LGB is a breakpoint between two adjacent genomic fragments with different genomic copy numbers, and each such genomic fragment has a length of at least a specific threshold (e.g., 10 megabases); determining that the LGB number is greater than the LGB threshold number; and identifying the patient as being at risk for a defect in a DNA homologous recombination pathway.

[0011] In some embodiments, the threshold number of LGBs is 25. In some embodiments, when the number of LGBs is greater than 35, the patient has a high risk of having a DNA homologous recombination pathway defect. In some embodiments, when the number of LGBs is greater than 25 and less than or equal to 35, the patient has a moderate risk of having a DNA homologous recombination pathway defect.

[0012] In some embodiments, the tumor sample is a FFPE sample or a CTC sample.

[0013] In some embodiments, the cancer is selected from breast cancer, ovarian cancer, pancreatic cancer, head and neck cancer, and melanoma. In some embodiments, the cancer is breast cancer.

[0014] In some embodiments, the genomic copy number profile comprises any one or a combination of parameters selected from the following: genomic fragments along the genome, the number of copies of genomic fragments, copy number variation (CNA) breakpoints, or the number of CNA breakpoints. In some embodiments, the genomic copy number profile also includes the number of fusions, inversions, and translocation events involved.

[0015] In some embodiments, a genomic copy number profile is generated by the following steps: aligning WGS data with a reference genome, thereby generating a set of aligned WGS reads along the genome, counting the aligned WGS reads on the genome, and generating genomic fragments along the genome, wherein two adjacent genomic fragments have significantly different read counts. Fusion or translocation events are determined by identifying split reads or read pairs that are mapped to different locations within or across chromosomes, and inversions are determined by read pairs with the wrong direction on the same chromosome.

[0016] In some embodiments, the methods disclosed herein further comprise administering to the patient a therapeutically effective amount of a PARP inhibitor and / or an alkylating agent. In some embodiments, the PARP inhibitor and / or alkylating agent is selected from iniparib, olaparib, rucaparib, CEP9722, MK4827, BMN673, 3-aminobenzamide, platinum complexes, chlormethine, chlorambucil, melphalan, cyclophosphamide, ifosfamide, estramustine, carmustine, lomustine, fotemustine, streptozocin, busulfan, pipobroman, procarbazine, dacarbazine, thiotepa, and temozolomide.

[0017] In another aspect, the present disclosure provides a method for treating cancer in a patient. In some embodiments, the method comprises administering a therapeutically effective amount of a PARP inhibitor and / or an alkylating agent to a patient, wherein the patient has been identified as being at risk for a defect in a DNA homologous recombination pathway by the methods disclosed herein. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The following drawings form part of this specification and are included to further demonstrate certain aspects of the present disclosure. The present disclosure may be better understood by reference to one or more of these drawings in combination with the detailed description of specific embodiments presented herein.

[0019] Figure 1 A schematic diagram of the deep learning model used to generate CI scores is shown.

[0020] Figure 2 LGB scores of HRD and non-HRD patients are shown.

[0021] Figure 3 The correlation between LGB and HRD risk scores for 0.5 coverage and other coverages is shown.

[0022] Figure 4 The correlation between LGB and HRD risk scores of the original and diluted samples is shown.

[0023] Figure 5Results of LGB testing in exemplary samples with sequencing depths of 0.01-0.5 are shown.

[0024] Figure 6 The correlation between LGB scores at 0.5 coverage and LGB scores at 0.2 or 0.1 coverage is shown.

[0025] Figure 7 Results of HRD risk testing in exemplary samples with sequencing depths of 0.01-0.5 are shown.

[0026] Figure 8 The correlation between HRD risk scores at 0.5 coverage and HRD risk scores at 0.2, 0.1, or 0.05 coverage is shown. DETAILED DESCRIPTION

[0027] Before describing the present disclosure in more detail, it should be understood that the present disclosure is not limited to the specific embodiments described, and therefore can certainly be varied. It should also be understood that the terms used herein are only used for the purpose of describing specific embodiment examples, and are not intended to be limiting, because the scope of the present disclosure will only be limited by the appended claims.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present disclosure belongs. Although any methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present disclosure, preferred methods and materials are now described.

[0029] All publications and patents cited in this specification are incorporated herein by reference, just as if each individual publication or patent was specifically and individually indicated to be incorporated by reference and is incorporated herein by reference to disclose and describe the methods and / or materials in connection with which the publications are cited. The citation of any publication is for disclosure prior to the filing date and should not be construed as an admission that the present disclosure is not entitled to antedate such publication by virtue of prior disclosure. In addition, the publication dates provided may be different from the actual publication dates which may need to be independently confirmed.

[0030] It will be apparent to those skilled in the art after reading this disclosure that, without departing from the scope or spirit of the present disclosure, the discrete components and features of each of the embodiments described and illustrated herein can be easily separated or combined with the features of any other embodiments. Any enumerated method can be implemented in the order of the enumerated events or in any other logically possible order.

[0031] definition

[0032] The following definitions are provided to assist the reader. Unless otherwise defined, all technical terms, symbols and other scientific or medical terms or terminology used herein are intended to have meanings commonly understood by those skilled in the art of chemistry and medicine. In some cases, for clarity and / or ease of reference, terms with commonly understood meanings are defined herein, and the inclusion of such definitions herein should not necessarily be interpreted as representing substantial differences compared to the definitions of terms commonly understood in the art.

[0033] As used herein, the singular forms "a", "an" and "the" include plural forms unless the context clearly dictates otherwise.

[0034] As used herein, the term "administering" refers to providing a pharmaceutical formulation or composition to a subject, and includes, but is not limited to, administration by a medical professional and self-administration.

[0035] As used herein, the term "cancer" refers to any disease involving abnormal cell growth, and includes all stages and all forms of the disease affecting any tissue, organ or cell in the body. The term includes all known cancers and neoplastic conditions, whether characterized by malignant, benign, soft tissue or solid tumors, and all stages and grades of cancer, including pre- and post-metastatic cancers. In general, cancer can be classified based on the tissue or organ in which the cancer is located or originates and the morphology of the cancerous tissue and cells. As used herein, cancer types include, but are not limited to, acute lymphoblastic leukemia (ALL), acute myeloid leukemia, adrenocortical carcinoma, anal cancer, astrocytoma, basal cell carcinoma of the cerebellum or brain in children, bile duct cancer, bladder cancer, bone tumors, brain cancer, cerebellar astrocytoma, brain astrocytoma / glioblastoma, ependymoma, medulloblastoma, supratentorial primitive neuroectodermal tumor (supratentorial primitive neuroectodermal tumor). tumors), visual pathway and hypothalamic gliomas, breast cancer, Burkitt's lymphoma, cervical cancer, chronic lymphocytic leukemia, chronic myeloid leukemia, colon cancer, emphysema, endometrial cancer, ependymoma, esophageal cancer, Ewing's sarcoma, retinoblastoma, stomach cancer, glioma, head and neck cancer, heart cancer, Hodgkin's lymphoma, islet cell carcinoma (endocrine pancreas), Kaposi's sarcoma, kidney cancer (renal cell carcinoma), laryngeal cancer, leukemia, liver cancer, lung cancer, neuroblastoma, non-Hodgkin's lymphoma, ovarian cancer, pancreatic cancer, pharyngeal cancer, prostate cancer, rectal cancer, renal cell carcinoma (kidney cancer), retinoblastoma, Ewing family of tumors, skin cancer, stomach cancer, testicular cancer, laryngeal cancer, thyroid cancer, vaginal cancer.

[0036] It is noted that in this disclosure, terms such as "comprising," "including," "consisting of," "containing," "including," and the like have the meanings ascribed to them in U.S. patent law; they are inclusive or open-ended, and do not exclude additional, unrecited elements or method steps. Terms such as "consisting essentially of" and "consisting essentially of" have the meanings ascribed to them in U.S. patent law; they permit the inclusion of additional ingredients or steps that do not materially affect the basic and novel characteristics of the claimed invention. The terms "consisting of" and "composed of" have the meanings ascribed to them in U.S. patent law; i.e., these terms are closed.

[0037] The terms "determine", "evaluate", "analytical test", "measure", and "detect" are used interchangeably to refer to quantitative and semi-quantitative determinations. In the case of quantitative and semi-quantitative determinations, the phrases "determine the level of a polynucleotide or polypeptide of interest" or "detect" a polynucleotide or polypeptide of interest can all be used.

[0038] The terms "nucleic acid" and "polynucleotide" are used interchangeably and refer to a polymeric form of nucleotides of any length, either deoxyribonucleotides or ribonucleotides, or their analogs. A polynucleotide may have any three-dimensional structure and may perform any known or unknown function. Non-limiting examples of polynucleotides include genes, gene fragments, exons, introns, messenger RNA (mRNA), transfer RNA, ribosomal RNA, ribozymes, cDNA, shRNA, single-stranded short or long RNA, recombinant polynucleotides, branched polynucleotides, plasmids, vectors, isolated DNA of any sequence, control regions, isolated RNA of any sequence, nucleic acid probes, and primers. Nucleic acid molecules may be linear or circular.

[0039] As used herein, the term "subject" refers to a human or any non-human animal (e.g., mouse, rat, rabbit, dog, cat, cow, pig, sheep, horse, or primate). Humans include both prenatal and postnatal forms. In many embodiments, the subject is a human. The subject can be a patient, which refers to a person seeking disease diagnosis or treatment from a medical provider. The term "subject" can be used interchangeably with "individual" or "patient" herein. A subject may suffer from or be susceptible to a disease or disorder, but may or may not show symptoms of the disease or disorder.

[0040] As used herein, the term "therapeutically effective amount" refers to an amount sufficient to prevent, treat, reduce and / or alleviate the symptoms and / or underlying causes of any condition or disease, or an amount sufficient to produce a desired effect on cells. In one embodiment, a "therapeutically effective amount" is an amount sufficient to reduce or eliminate the symptoms of a disease. In another embodiment, a therapeutically effective amount is an amount sufficient to overcome the disease itself.

[0041] The terms "therapeutic treatment," "treatment," or "therapy" refer to methods of reducing the effects of cancer (e.g., breast cancer, lung cancer, ovarian cancer, etc.) or cancer symptoms. Thus, in the disclosed methods, treatment can refer to a 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 100% reduction in the severity of a cancer or cancer symptom. For example, a method of treating a disease is identified as treatment if one or more disease symptoms of the subject are reduced by 10% compared to a control. Thus, the reduction can be 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, 100%, or any percentage reduction between 10% and 100% compared to the original or control level. It should be understood that treatment does not necessarily refer to a cure or complete eradication of a disease, disorder, or symptom of a disease or disorder.

[0042] Cancer and Homologous Recombination Deficiency

[0043] Cancer development is a multifactorial process and includes key events such as the ability to replicate without limit, induction of angiogenesis, and activation of invasion and metastasis. A major potential mechanism for cancer development is the emergence of genomic instability, which is caused by exogenously or endogenously induced DNA damage or by genetic mutations resulting from failure of DNA damage repair. Organisms have evolved different strategies to cope with various forms of DNA damage. Incorrect bases introduced during replication are removed by proteins of the mismatch repair (MMR) pathway. The nucleotide excision repair (NER) machinery recognizes structural aberrations within the DNA duplex and removes the affected bases. The base excision repair (BER) pathway is activated by damaged DNA bases. There are two major repair pathways for double-strand breaks: precise homologous recombination (HR) and error-prone non-homologous end joining (NHEJ).

[0044] DNA double-strand breaks (DSBs) can be caused by exogenous sources such as ionizing radiation or endogenous sources such as cellular metabolic byproducts. DSBs caused by internal factors such as replication blockage can usually be directly repaired by using the sister chromatid located nearby as a template for homologous recombination. On the other hand, DSBs caused by ionizing radiation mainly occur on compact chromatin structures. Therefore, interactions using intact homologous sequences as templates for HR are impossible. In order to deal with such DSBs in condensed chromatin structures, vertebrates often use NHEJ to simply reconnect the DSB end chains. Another possible reason why NHEJ is the main DSB repair mechanism in humans may be because HR tends to occur in the S / G2 phase, while NHEJ dominates in the G1 phase, which is longer than the S / G2 phase.

[0045] Homologous recombination is a precise repair mechanism for DSBs because it uses a complete DNA copy of the sister chromatid or homologous chromosome as a template to repair the break. In the first step, the damage is recognized by the MRE11-RAD50-NBS1 (MRN) complex, which activates the ATM (Ataxia telangiectasiamutated) kinase. Once the DNA 5′-end is resected, the replication protein (RPA) covers the single-stranded DNA region and activates the ATR (Ataxia telangiectasia and Rad3-related) kinase. RPA is replaced by RAD51, a process that requires the help of further repair-related proteins such as CHEK2, BRCA1, BRCA2, and PALB2, which are loaded with RAD51. Subsequently, the defective DNA strand is attached to its sister chromatid, which is used as a template for DNA resynthesis.

[0046] The HR pathway includes proteins that are important for double-strand DNA break repair, such as BRCA1, BRCA2, PALB2 / FANCN, BRIP1 / FANCJ, BARD1, RAD51, and RAD51 paralogs (RAD51B, RAD51C, RAD51D, XRCC2, XRCC3). When the genes for these proteins mutate or are underexpressed, errors in DNA repair may occur, ultimately leading to cancer. Although recurrent mutations have not yet been found in human tumors, other factors of the HR pathway may potentially be dysregulated in cancer, such as FANCA, FANCB, FANCC, FANCD2, FANCE, FANCG, FANCI, FANCL, FANCM, FAN1, SLX4 / FANCP, or ERCC1. As used herein, defects in the HR pathway and HR defects can be used interchangeably and refer to a situation in which one or more proteins of the HR pathway involved in repairing DNA are deficient or inactivated.

[0047] The core component of the non-homologous end joining (NHEJ) pathway is the Ku70–Ku80 complex, which can bind to the ends of double-stranded broken DNA and recruit DNA PKcs to initiate NHEJ. Further, the remaining factors involved in this process process the broken DNA to build DNA ends suitable for ligation. Subsequently, DNA ligase IV (LIG4) is recruited simultaneously with XRCC4 and XRCC4-like factors (XLF) to connect the processed DNA and reshape genome integrity. During NHEJ, changes in the DNA sequence or random blunt-end ligation may occur. Therefore, the integrity of chromosomes can be easily lost, resulting in chromosome rearrangements.

[0048] Cancer patients with HRD have been treated with poly (ADP-ribose) polymerase (PARP) inhibitors, which disable single-strand break repair. When homologous recombination deficiency (HRD) occurs in cells, cells are forced to switch to the error-prone NHEJ pathway to repair DSBs. This leads to the accumulation of genomic damage and promotes genomic instability. When HRD and PARP inhibition occur simultaneously in cells, DNA damage repair mechanisms are greatly impaired, leading to synthetic lethality and forced cell death.

[0049] However, defining those HRD patients who may benefit from PARPi therapy remains challenging. Alterations in genes encoding proteins in the HR pathway, such as BRCA1 and BRCA2, are commonly used to identify HRD patients. However, in a recent clinical trial of ovarian cancer, it has been confirmed that nearly 20% of the study population was HRD positive without BRCA mutations (Miller, RE et al. (2020) ESMO recommendations on predictive biomarker testing for homologous recombination deficiency and PARP inhibitor benefit in ovarian cancer. Ann. Oncol. Off. J. Eur. Soc. Med. Oncol. 31, 1606–1622). Therefore, it is still necessary to identify patients with HR pathway defects.

[0050] Methods for detecting homologous recombination deficiency

[0051] The present disclosure provides, in one aspect, a method for detecting a defect in a DNA homologous recombination pathway in a cancer patient. In some embodiments, the method involves detecting chromosomal instability in a sample using whole genome sequencing (WGS) data generated from a patient tumor sample.

[0052] The tumor sample used in the methods disclosed herein refers to a biological sample or a sample from a biological source containing one or more tumor cells. Biological samples include samples from body fluids, such as blood, plasma, serum or urine, or samples obtained from cells, tissues or organs by biopsy, preferably tumor tissue suspected of including or substantially consisting of cancer cells. The tumor sample can be a fresh tumor sample or a frozen tumor sample. The tumor sample can be obtained by biopsy, for example, by surgery or needle biopsy.

[0053] In some embodiments, the tumor sample used in the method disclosed herein has a low tumor ratio or a low tumor DNA ratio. Such tumor samples include formalin-fixed paraffin-embedded (FFPE) tissue samples, single cell samples, and peripheral circulating tumor cells. In this case, the WGS data generated from the tumor sample and used for the method disclosed herein have a very low sequencing depth. As used herein, sequencing depth refers to the ratio of the total number of bases obtained by sequencing to the size of the genome, or refers to the average number of times each base is sequenced in the genome. In some embodiments, the WGS data used in the method disclosed herein have a depth of 0.05-0.5, for example, 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45 and 0.5.

[0054] In some embodiments, WGS is performed using DNA isolated from a tumor sample. DNA can be isolated from a tumor sample using a variety of methods. Standard methods for extracting DNA from tissues or cells are described, for example, in Ausubel et al., Current Protocols of Molecular Biology (1997) John Wiley & Sons and Sambrook and Russell, Molecular Cloning: A Laboratory Manual 3rd ed. (2001). Commercially available kits, for example, Fecal DNA genomic extraction kit ( DNA Stool Mini Kit, Qiagen) can also be used to isolate DNA from tumor samples.

[0055] Whole-genome sequencing methods

[0056] In the present disclosure, whole genome sequencing (WGS) data can be generated by any method known in the art. In some embodiments, WGS is performed by high-throughput sequencing.

[0057] High-throughput sequencing, or next-generation sequencing, is highly scalable and capable of sequencing an entire genome or transcriptome at once by using a different approach than traditional methods such as Sanger sequencing. Based on its mechanism, high-throughput sequencing includes sequencing by synthesis, pyrosequencing, sequencing by ligation, and nanopore sequencing.

[0058] Sequencing by synthesis involves the synthesis of a complementary strand of a target nucleic acid by the introduction of labeled nucleotides or nucleotide analogs during polymerase amplification. Immediately after or upon successful introduction of the labeled nucleotide, the labeled signal is measured and the characteristics of the nucleotide are recorded. The detectable label on the introduced nucleotide is removed before repeating the introduction, detection, and identification steps. Examples of sequencing by synthesis methods are known in the art and are described in, for example, U.S. Pat. No. 7,056,676, U.S. Pat. No. 8,802,368, and U.S. Pat. No. 7,169,560, the contents of which are incorporated herein by reference. Sequencing by synthesis can be performed on a solid surface (or microarray or chip) using fold-back PCR and anchor primers. Target nucleic acid fragments can be attached to a solid surface by hybridization with an anchor primer and bridge amplification. For example, this technology is applicable to, for example, Sequencing platform.

[0059] Pyrophosphate sequencing involves hybridizing a target nucleic acid region with a primer and extending a new chain by sequentially introducing deoxynucleotide triphosphates corresponding to the bases A, C, G, and T (U) in the presence of a polymerase. Each base introduction is accompanied by the release of pyrophosphate, which is converted to ATP by sulfate pyrophosphatase, thereby driving the synthesis of oxyluciferin and the release of visible light. Since the release of pyrophosphate is equimolar to the number of bases introduced, the light emitted is proportional to the number of nucleotides added in any one step. The process is repeated until the entire sequence is determined.

[0060] Sequencing by ligation is a DNA sequencing method that uses DNA ligase to identify the nucleotide at a given position in a DNA sequence. Unlike most popular DNA sequencing methods, sequencing by ligation does not use DNA polymerase to generate the second strand. Instead, sequencing by ligation uses the mismatch sensitivity of DNA ligase to determine the basic sequence of the target DNA molecule. Sequencing by ligation relies on the sensitivity of DNA ligase to base pairing mismatches. The target molecule to be sequenced is a single-stranded unknown DNA sequence with a known sequence attached to at least one end. A short "anchor" strand is then introduced to bind the known sequence. A mixture of probe oligonucleotides (eight or nine bases long) is then introduced, labeled according to the position to be sequenced (usually with a fluorescent dye). These probes hybridize to the target DNA sequence, adjacent to the anchor sequence, and DNA ligase preferentially ligates this probe to the above anchor when its base matches the unknown DNA sequence. Based on the fluorescence generated by the probe, the identity of the nucleotide at that position in the unknown sequence can be inferred. Oligonucleotide probes can also construct cleavable junctions, and the probe is cleaved when the label is recognized. This will both remove the label and regenerate the 5' phosphate at the end of the ligated probe, preparing the system for the next round of ligation. This cycle can be repeated several times to read longer sequences.

[0061] Nanopore sequencing is a third generation sequencing method that works by monitoring changes in electrical current as nucleic acids pass through protein nanopores in solid membranes. The biological or solid membrane in which the nanopore is located is surrounded by an electrolyte solution. This membrane divides the solution into two chambers. Applying a bias voltage across the membrane induces an electric field that drives the movement of charged particles (in this case, ions). This effect is called electrophoresis. For sufficiently high concentrations, the electrolyte solution is well distributed and all the voltage drop is concentrated near and inside the nanopore. This means that charged particles in the solution can only feel the force from the electric field when they are close to the pore area. This area is often called the capture zone. In the capture zone, the ions have a directed motion that can be recorded as a steady ionic current by placing electrodes near the membrane. Now imagine a nanoscale polymer, such as a nucleic acid, placed in one of the chambers. The nucleic acid also has a net charge that feels the force from the electric field in the capture zone. The nucleic acid approaches this capture zone due to Brownian motion and any possible attraction to the membrane surface. Once inside the nanopore, the nucleic acid passes through the nanopore one nucleotide after another via a combination of electrophoretic, electroosmotic, and sometimes thermophoretic forces. Inside the pore, the nucleotides of the nucleic acid occupy volume and restrict the flow of ions, which manifests as a decrease in the ionic current. Based on various factors of the nucleotide, such as geometry, size and chemical composition, the magnitude of the change in ionic current and the duration of translocation will vary. Based on this modulation of the ionic current, different nucleotides in the nucleic acid can then be sensed and potentially identified.

[0062] Large-scale genomic breakpoints

[0063] In some embodiments, chromosomal instability in the genome of a tumor sample is detected by measuring the number of large-scale rearrangement or large-scale genomic alteration breakpoints (LGB) per genome in the tumor sample. In some embodiments, LGB refers to a breakpoint between two adjacent genomic fragments with different copy numbers, each such genomic fragment being at least 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 megabases long.

[0064] A normal diploid genome has exactly two copies of each chromosome. However, copy number variation (CNA) occurs when the genome of a cell becomes unstable, and the copy number changes throughout the genome. In order to determine CNAs in a genome, the genome is divided into continuous regions of maximum length, so that the copy number of each site within the region is the same. Each such region is called a fragment, and the number of copies shared by all positions within the fragment is called the absolute copy number. In the case of a diploid genome without CNAs, there is only one fragment, the entire genome, and its absolute copy number is 2. On the other hand, cells with HRD will have CNAs and multiple fragments, especially large-sized fragments. Therefore, the LGB number can be used as a biomarker for HRD in cells.

[0065] In some embodiments, LGB can be detected by a genomic copy number profile generated from WGS data. In some embodiments, the genomic copy number profile is a copy number profile within the entire genome and includes parameters of genomic fragments in the genome, the copy number of genomic fragments, copy number variation (CNA) breakpoints, the number of CNA breakpoints, and combinations thereof.

[0066] In some embodiments, the genome copy number profile can be generated according to the following steps. In the first step, the WGS data, i.e., nucleic acid sequence reads, are mapped to a reference genome. As used herein, the term "mapping" or "mapping to a reference genome" refers to comparing nucleic acid sequence reads to a reference genome whose sequence is known. Various programs and algorithms have been developed to map nucleic acid sequence reads to a reference (see Flicek P, Birney E. (2009) Sense from sequence reads: methods for alignment and assembly, Nat Methods 6(11 Suppl): S6-S12; Neilsen R, Paul JS et al. (2011) Genotype and SNP calling from next-generation sequencing data. Nat Rev Genet 12: 443-52; Ruffalo M et al. (2011) Comparative analysis of algorithms for next-generation sequencing read alignment. Bioinformatics 27: 2790-96; Patnaik S et al. (2012) Customisation of the exome data analysis pipeline using a combinatorial approach. PLoS ONE 7: e30080). Among various programs and algorithms, the Burrows-Wheeler aligner (BWA) based on the Burrows-Wheeler transform (Li H, Durbin R (2009) Fast and accurate short rad alignment with Burrows-Wheeler transform. Bioinformatics 25: 1754-60) has been shown to have a good balance between running time, memory usage, and accuracy, and is generally applicable to different computing workflows.

[0067] In the next step, the number of aligned WGS reads is counted in each site or region along the genome. In some embodiments, in order to count the number of aligned WGS reads, the genome is divided into fixed or variable size fragments or windows, which are called bins, and the number of reads in the bins is summarized, and the resolution at which the genome copy number is called is defined by the bins rather than by individual sites. This is to reduce the impact of variable amplification and sequence sampling. In some embodiments, a fixed size, for example, 30, 40, 50, 60, 70, 80, 90 or 100 kilobases, preferably a bin of 50 kilobases (all bins have the same size), is used to count the number of aligned WGS reads. In some embodiments, the number of aligned WGS reads is counted using a bin of variable size to avoid false positive deletion calls (i.e., false calls of breakpoints) within the repeat region, thereby removing low mapping quality reads.

[0068] In some embodiments, read counts are adjusted by GC correction and mappability correction. GC correction adjusts read counts according to the GC content in the corresponding genomic region to eliminate the effect of GC content on read counts. Mappability correction adjusts read counts according to the degree of mappability of the genomic region. The higher the repetitiveness in a region, the lower its mappability and the lower the number of uniquely aligned reads.

[0069] The next step in genomic copy number profile generation is segmentation, which identifies the boundaries (physical locations) between genomic regions with different absolute copy numbers. There are three common segmentation methods: sliding window methods, objective function-based methods, and HMM-based methods.

[0070] Sliding window methods segment the genome using statistical tests, looking for regions where read counts differ greatly from other regions. Methods using sliding window methods do not calculate absolute copy numbers at the same time as segmentation. A post-processing method is required, which usually involves testing different ploidy candidates and selecting the candidate whose copy number profile is as close to integer as possible.

[0071] Objective function-based methods combine the approximation of the data and the extreme values ​​of the breakpoints into a single formula. This method models the (normalized) read counts via a piecewise constant function that (i) is as faithful to the data as possible and (ii) has as little variation as possible. Like sliding window methods, objective function-based methods do not necessarily also assign an absolute copy number to each fragment.

[0072] In HMM-based methods, states correspond to different possible absolute copy numbers, and transitions between states capture segmentation (i.e., a transition from a state at bin i indicates that bins i-1 and i belong to two different fragments). The objective function approach is more suitable for simultaneous breakpoint identification within sampled cells, as it can model hundreds of cells within one objective function.

[0073] After segmentation, the number of LGBs per genome in the tumor sample, i.e., the number of breakpoints between two adjacent genomic segments with different copy numbers, can be counted. In some embodiments, only genomic segments of at least 5 megabases (preferably 6, 7, 8, 9, or 10 megabases) are counted to ensure that the copy number changes are caused by chromosomal rearrangements. In some embodiments, all genomic segments are counted or a relaxed standard (1, 2, 3, 4 megabases) is used.

[0074] In some embodiments, the number of LGBs per genome in a tumor sample is compared to a threshold number to determine whether the patient has HRD. In some embodiments, the threshold number of LGBs is 25 / genome, that is, when the number of LGBs / genome in a tumor sample is greater than 25 / genome, it can be determined that the patient has a risk of HRD. In some embodiments, when the number of LGBs is greater than 35, it can be determined that the patient has a high risk of HRD. In some embodiments, when the number of LGBs is greater than 25 and less than or equal to 35, it can be determined that the patient has a moderate risk of HRD.

[0075] Deep Learning Models

[0076] In some embodiments, the methods disclosed herein involve using a deep learning model to determine a patient's risk of developing HRD.

[0077] Deep learning is a machine learning method based on artificial neural networks that use representation learning. Artificial neural networks (ANNs) are computing systems inspired by the biological neural networks that make up animal brains. ANNs are based on a collection of connected units called artificial neurons (similar to biological neurons in biological brains). Each connection (synapse) between neurons is capable of passing a signal to another neuron. The receiving (post-synaptic) neuron is capable of processing the signal and then sending a signal to the downstream neuron connected to it. Neurons may have states, which are usually represented by real numbers, usually between -1 and 1. Neurons and synapses may also have weights that change as learning proceeds, which may increase or decrease the strength of the signal they send downstream. Generally speaking, neurons are organized in layers. Different layers can perform different types of transformations on their inputs. Signals are transmitted from the first layer (input) to the last layer (output), possibly after traversing these layers many times. Deep learning refers to the use of multiple layers in the network to extract progressively higher-level features from the raw input.

[0078] In some embodiments, the method disclosed herein involves using a deep learning model to predict HRD status using CNV fragment information. In some embodiments, the deep learning model includes one or more layers of long short-term memory (LSTM) or convolutional neural network (CNN), followed by a fully connected network. LSTM is an artificial recurrent neural network architecture that has feedback connections, unlike standard feedforward neural networks. CNN is an artificial neural network architecture that uses shared weight convolution kernels to extract features by sliding along the input. In some embodiments, the deep learning model generates an instability (CI) score for each of the 22 chromosomes from WGS data.

[0079] In the next step, the HRD risk score is calculated by calculating the average of the CI scores. In some embodiments, the HRD risk score is a weighted average of the CI scores based on chromosome base length, each chromosome position, or each CNA breakpoint. The structure of the exemplary deep learning model used in this article is as follows Figure 1 as shown in .

[0080] In some embodiments, to train a deep learning model, a chromosome instability (CI) score is assigned to each chromosome in the sample based on the HRD label of the sample. All 22 chromosomes of HRD-positive samples have a CI score of 1, while the CI score of negative samples is 0. In some embodiments, a deep learning model is trained for chromosome-level data to predict CI scores.

[0081] In some embodiments, the trained model is applied to the sample CNV segmentation results generated from the WGS data to predict the HRD risk score. In some embodiments, the HRD risk score is a value between 0-1. In some embodiments, when the risk score is greater than 0.6, the patient has a high risk of HRD. In some embodiments, when the risk score is greater than 0.4 and less than or equal to 0.6, the patient has a moderate risk of HRD. In some embodiments, when the risk score is not greater than 0.4, the patient has a low risk of HRD.

[0082] Computer-implemented methods, systems and devices

[0083] Any method described herein can be performed in whole or in part with a computer system including one or more processors, and one or more processors can be configured to perform each step. Therefore, each embodiment relates to a computer system configured to perform each step of any method herein, which may have different components for performing corresponding steps or corresponding step groups. Although presented as numbered steps, the steps of each method herein can be performed at the same time or in different orders. In addition, parts of these steps can be used together with parts of other steps from other methods. And all or part of the steps can be optional. Any step of any method can be performed with modules, circuits or other devices for performing these steps.

[0084] Any computer system mentioned herein can utilize any suitable number of subsystems. In some embodiments, the computer system includes a single computer device, wherein the subsystem can be a component of the computer device. In other embodiments, the computer system can include multiple computer devices, each of which is a subsystem with internal components. The subsystems can be interconnected through a system bus. Other subsystems include, for example, printers, keyboards, storage devices, monitors coupled to display adapters, and the like. Peripheral devices and input / output (I / O) devices coupled to an I / O controller can be connected to a computer system by any number of means known in the art, such as serial ports. For example, a serial port or an external interface (e.g., Ethernet, Wi-Fi, etc.) can be used to connect a computer system to a wide area network, such as the Internet, a mouse input device, or a scanner. Interconnection through a system bus allows a central processing unit to communicate with each subsystem and control the execution of instructions from a system memory or storage device (e.g., a fixed disk, such as a hard disk or an optical disk), as well as information exchange between subsystems. System memory and / or storage devices can be implemented as computer-readable media. Any data mentioned herein can be output from one component to another and can be output to a user.

[0085] A computer system can include multiple identical components or subsystems, for example, connected together through external interfaces or through internal interfaces. In some embodiments, computer systems, subsystems or devices can communicate within a network. In this case, one computer can be considered a client, while another computer can be considered a server, in which case each computer can be part of the same computer system. Clients and servers can each include multiple systems, subsystems or components.

[0086] It should be understood that any embodiment of the present disclosure can be implemented in the form of control logic using hardware (e.g., an application specific integrated circuit or a field programmable gate array) and / or computer software with a general programmable processor in a modular or integrated manner. As used herein, a processor includes a multi-core processor on the same integrated chip, or multiple processing units on a single circuit board or network. Based on the disclosure and teachings provided herein, those of ordinary skill in the art will know and understand other ways and / or methods of implementing various embodiments of the present disclosure using hardware and a combination of hardware and software.

[0087] Any software component or function described in this application can be implemented as software code for execution by a processor using any suitable computer language, such as Java, C++ or Perl, such as traditional or object-oriented techniques. The software code can be stored and / or transmitted as a series of instructions or commands stored on a computer-readable medium, suitable media including random access memory (RAM), read-only memory (ROM), magnetic media (such as hard disk drives or floppy disks) or optical media (such as compact disks (CDs) or DVDs (digital versatile disks)), flash memory, etc. The computer-readable medium can be any combination of such storage or transmission devices.

[0088] Such program can also be encoded and transmitted using a carrier signal adapted to be transmitted via wired, optical and / or wireless networks that meet the various protocols including the Internet. Therefore, the computer-readable medium according to an embodiment of the present invention can be created using a data signal that adopts such program coding. The computer-readable medium encoded with program code can be packaged with compatible devices, or provided separately from other devices (e.g., downloaded via the Internet). Any such computer-readable medium can be located at or built into a single computer product (e.g., hard drive, CD or entire computer system), and can be located at or built into different computer products in a system or network. A computer system can include a monitor, a printer or other suitable displays for providing any result described herein to the user.

[0089] Methods of treating cancer

[0090] The method of identifying patients with HRD as described above has therapeutic applications. It is believed that if the tumor is HR pathway defective, treatment with drugs that cause DNA double-strand breaks (such as alkylating agents) or treatments that inhibit alternative DNA repair pathways (such as PARPi) will be more effective. Therefore, in another aspect, the present disclosure provides a method for treating cancer in patients with HR pathway defects. In some embodiments, the method includes administering to the patient a therapeutically effective amount of a drug that inhibits alternative DNA repair pathways, such as a PARP inhibitor, or a drug that causes DNA double-strand breaks, such as an alkylating agent, wherein the patient has been determined to suffer from HRD by the above method.

[0091] As used herein, "PARP inhibitors" refers to compounds that inhibit the activity of the poly ADP ribose polymerase (PARP) enzyme, a protein that is important for repairing single-strand breaks ("nicks" in the DNA). If such nicks are not repaired until the DNA is replicated (which must occur before the cell divides), the replication itself can lead to the formation of double-strand breaks. Drugs that inhibit PARP can cause multiple double-strand breaks to form in this way, and in tumors with HRD, these double-strand breaks cannot be effectively repaired, leading to cell death.

[0092] In some embodiments, the PARP inhibitor according to the present invention can be selected from iniparib, olaparib, rucaparib, CEP 9722, MK 4827, BMN-673 and 3-aminobenzamide.

[0093] As used herein, the term "alkylating agent" or "alkylating antitumor agent" refers to a compound that attaches an alkyl group to DNA. In some embodiments, the alkylating agent according to the present invention can be selected from platinum complexes (such as cisplatin, carboplatin and oxaliplatin), bis (2-chloroethyl) methylamine, chlorambucil, melphalan, cyclophosphamide, ifosfamide, estramustine, carmustine, lomustine, fotemustine, streptozotocin, busulfan, pipobroman, procarbazine, dacarbazine, thiotepa and temozolomide.

[0094] The drugs described herein can be administered in any desired and effective manner: oral ingestion, or topical administration to the eye as an ointment or drops, or parenteral or other administration in any appropriate manner, such as intraperitoneal, subcutaneous, topical, intradermal, inhalation, intrapulmonary, rectal, vaginal, sublingual, intramuscular, intravenous, intraarterial, intrathecal or intralymphatic. In addition, the drug can be administered in combination with other treatments.

[0095] Suitable non-limiting examples of drug dosages are about 1 mg / kg to about 2400 mg / kg per day, such as about 1 mg / kg to about 1200 mg / kg per day, 75 mg / kg to about 300 mg / kg per day, including about 1 mg / kg to about 100 mg / kg per day. Other representative dosages of such drugs include about 1 mg / kg, 5 mg / kg, 10 mg / kg, 15 mg / kg, 20 mg / kg, 25 mg / kg, 30 mg / kg, 35 mg / kg, 40 mg / kg, 45 mg / kg, 50 mg / kg, 60 mg / kg, 70 mg / kg, 75 mg / kg, 80 mg / kg, 90 mg / kg, 100 mg / kg, 125 mg / kg, 150 mg / kg, 175 mg / kg, 200 mg / kg, 250 mg / kg, 300 mg / kg, 350 mg / kg, 400 mg / kg, 450 mg / kg, 500 mg / kg, 600 mg / kg, 700 mg / kg, 750 mg / kg, 800 mg / kg, 900 mg / kg, 1000 mg / kg, 1250 mg / kg, 1500 mg / kg, 1750 mg / kg, 2000 mg / kg, 2500 mg / kg, 3000 mg / kg, 35 ... In some embodiments, the dosage of the drug in humans is about 400 mg / day administered every 12 hours. In some embodiments, the dosage range of the drug in humans is 300-500 mg / day, 100-600 mg / day or 25-1000 mg / day. An effective dose of an agent disclosed herein may be administered as two, three, four, five, six or more sub-doses, separated at appropriate intervals throughout the day.

[0096] The following examples are provided to better illustrate the claimed invention and should not be construed as limiting the scope of the invention. All specific compositions, materials and methods described below fall within the scope of the invention in whole or in part. These specific compositions, materials and methods are not intended to limit the present invention, but are only used to illustrate specific embodiments falling within the scope of the invention. Those skilled in the art can develop equivalent compositions, materials and methods without exercising inventive ability and without departing from the scope of the invention. It should be understood that many changes can be made in the process described herein while still remaining within the scope of the invention. The inventor's intention is to include such changes within the scope of the invention.

[0097] Example 1

[0098] Materials and methods

[0099] Very low coverage whole genome sequencing (WGS) of formalin-fixed paraffin-embedded (FFPE) tumor tissue samples

[0100] Tissue sample processing and database construction process:

[0101] In order to prepare the DNA library for sequencing from fresh surgical tissue or FFPE samples, DNA extraction was performed using GeneRead DNA FFPE kit (Qiagene) for the obtained patient samples. The extracted DNA was tested for concentration and purity. 200 ng DNA was extracted from each sample and used as input material for library preparation. Ultra TM II library preparation kit is used to generate NGS sequencing libraries, and a barcode sequence (index code) is added to each sample during the library preparation process. After purifying the PCR product, the size distribution of the library is analyzed and quantified by real-time PCR. The library is sequenced on the Illumina NovaSeq platform.

[0102] In order to prepare a DNA library for sequencing from CTC blood samples, a commercial platform (including, but not limited to, CTC biological characteristics and separation methods based on their physical properties) was used to enrich and separate CTCs (circulating tumor cells) for the obtained patient samples (fresh blood). A single-cell whole genome amplification kit was used for CTC whole genome amplification. After determining the concentration of the amplified whole genome DNA, 200 ng DNA was used as the input material for library preparation for each sample, and the whole genome was amplified using a single-cell whole genome amplification kit. Ultra TM II library preparation kit to generate NGS sequencing libraries. During the library preparation process, barcode sequences are added to each sample for indexing. After PCR product purification, the size distribution of the library is analyzed and quantified by real-time PCR. Each library is sequenced on the Illumina sequencing platform.

[0103] Genome copy number profiling

[0104] The raw sequencing data were aligned to the reference genome using BWA, QC filtering and duplicate removal were performed. Uniquely aligned reads with good mapping quality were then counted along the genome using a 50,000 bp long sliding window. The read counts were adjusted for genomic GC content and mappability for each window. HMMcopy was called on the window read counts in its default settings to obtain genomic copy number profiles for human chromosomes 1-22 (chromosomes X and Y were ignored in the analysis).

[0105] Large-scale genomic breakpoint (LGB) determination

[0106] Define and calculate large-scale genomic breakpoints (LGBs) by the following steps: (i) Remove genomic fragments shorter than 3 Mb; (ii) For each chromosome, when two adjacent fragments are both longer than 10 Mb, count the copy number variation (CNA) breakpoints; and (iii) Sum the CNA breakpoints of all 22 chromosomes to obtain the final LGB score. The high-risk group of HRD is defined as LGB > 35, the medium-risk group as 25 < LGB ≤ 35, and the low-risk group as LGB ≤ 25.

[0107] Predict HRD risk score through HRD risk test

[0108] The present inventors developed an HRD risk test, which is a deep learning-based model that uses CNV fragment information to predict the HRD status of a sample. The HRD risk test consists of two components: (i) The first component is a 1-LSTM layer followed by a fully connected network, which generates a chromosome instability (CI) score for each of the 22 chromosomes; (ii) The second component calculates the HRD risk score by calculating the weighted average of the CI scores (chromosome base length as the weight). The entire model structure is as Figure 1 shown.

[0109] To train the deep learning model, each chromosome in the sample is assigned a chromosome instability (CI) score based on the HRD label of the sample. The CI scores of all 22 chromosomes in HRD-positive samples are 1, while the CI scores of negative samples are 0. The deep learning model is trained for chromosome-level data to predict CI scores. The model is trained on 80 samples (28 HRD-positive) and 1760 chromosomes and validated on 93 samples (27 HRD-negative) and 2046 chromosomes.

[0110] The trained model is applied to the sample CNV segmentation results generated by HMMcopy to predict the HRD risk score. The HRD risk score is a value between 0 and 1. The high-risk group of HRD is defined as risk score > 0.6, the medium-risk group as 0.4 < risk score ≤ 0.6, and the low-risk group as risk score ≤ 0.4.

[0111] Example 2

[0112] This example illustrates the consistency between the results measured by the shallow sequencing method (LGB and HRD risk score) and the deep sequencing method.

[0113] Deep WGS data of 191 cancer patient samples were downloaded from the European Genome-Phenome Archive (EGA) (Nik-Zainal, S. et al. (2016) Landscape of somatic mutations in 560 breast cancer whole-genome sequences. Nature, 534 (7605), 47-54). The HRD labels of the patient data were obtained from HRDetect (Davies, H. et al. (2017) HRDetect is a predictor of BRCA1 and BRCA2 deficiency based on mutational signatures. Nature medicine, 23 (4), 517-525). The original WGS data was downsampled to 0.5 genome coverage to simulate ultra-low coverage WGS. The HDR risk method was then applied to calculate the LGB score. As described in Example 1, 80 samples were tested for HRD risk and 111 samples were verified. The performance of the HDR risk method was evaluated using the HRDetect predicted label as the basic truth. Figure 2 It was shown that both LGB and HRD risk scores were able to distinguish HRD-negative and HRD-positive groups well, with high AUROCs equal to 0.968 and 0.974, respectively. For the HRD risk test, the AUROC for the training data was 0.990, while the AUROC for the validation data was 0.962. The consistency of the two methods was also checked. A strong linear correlation was found between the LGB score and the HRD risk score through the HRD risk test.

[0114] Example 3

[0115] This example illustrates the robustness of the methods disclosed herein in data of very low sequencing depth.

[0116] The 130 WGS samples were further downsampled to lower coverages of 0.2, 0.1, 0.05, 0.02, and 0.01 to examine the detection limit of the sequencing depth of the methods disclosed herein. A good correlation was observed between the LGB score calculated in the 0.5 coverage sample and the LGB score calculated in the 0.2, 0.1, and 0.05 coverage samples (Table 1, Figure 3 ). Although there is still overall consistency at coverages below 0.05, the AUROC begins to lag behind the optimal range. Overall, the methods disclosed in this article are able to identify HRDs in very low coverage WGS data as low as 0.05 depth.

[0117] Table 1. AUROC of LGB and HRD risk scores with different coverage

[0118]

[0119] Example 4

[0120] This example illustrates the robustness of the methods disclosed herein when applied to low tumor fraction samples.

[0121] A common observation of clinical tumor samples is that the tumor proportion in FFPE samples is variable, and low tumor proportions may potentially impair the sensitivity of the analysis algorithm. In order to test the robustness of the method disclosed herein to low tumor proportions, the inventors generated mixed samples with 80%, 60%, 40% and 20% tumor proportions by randomly sampling the reads from the original tumor sample and the matched normal sample with corresponding probabilities. The same analysis was performed to calculate the LGB and HRD risk scores. The inventors observed a good correlation between the LGB and HRD risk scores of the original and diluted samples ( Figure 4 ). Even when the simulated samples contained 60% less tumor DNA, the sensitivity and specificity of our method were only minimally affected.

[0122] In order to understand the actual detection limit of the tumor score of the method disclosed in this article, the inventors ran Sequenza (Favero, F. et al. (2015) Sequenza: allele-specific copy number and mutation profiles from tumor sequencing data. Annals of Oncology, 26 (1), 64-70) based on tumor sequencing data to estimate the tumor purity in the original sample. The purity of the diluted sample was calculated by multiplying the estimated purity of the original sample by the corresponding dilution factor. The original samples and diluted samples of different degrees were then combined together to form the final analysis pool, and the AUROC of each purity group was calculated. The inventors found that our method can achieve good HRD detection performance in low tumor scores.

[0123] Example 5

[0124] This example further validates the robustness of the LGB score as a HRD biomarker by comparing it with the traditional method of BRCA analysis testing.

[0125] BRCA analysis was performed on 75 breast cancer patients (Harbin, China), of whom 25 were BRCA positive and the remaining 50 were BRCA negative. Figure 4As shown, our method maintains good sensitivity and specificity even when the tumor proportion of the sample is as low as 20%.

[0126] BRCA Testing: BRCA analysis testing is performed using a next-generation sequencing (NGS)-based BRCA mutation detection kit.

[0127] Briefly, total DNA was used for DNA sequencing library preparation, and genomic regions of BRCA1 and BRCA2 were enriched using target probes. Sequencing was performed using the enriched libraries, and the mutation status of BRCA1 / 2 was subsequently analyzed based on the sequencing results.

[0128] Based on such findings, 42 samples (9 BRCA positive and 33 BRCA negative) with tumor proportion greater than 20% were prepared and used to validate the methods disclosed herein. For each sample, tumor proportion was evaluated by pathological examination using FFPE sections.

[0129] like Figure 5 As shown, at different sequencing depths of 0.01-0.5, the LGB test of BRCA-positive samples confirmed that the LGB score of BRCA-positive samples was significantly higher than that of BRCA-negative samples, indicating that the LGB score can distinguish BRCA-positive and BRCA-negative samples.

[0130] As shown in Table 2, the data filled in the "Defective" and "Complete" columns refer to the number of samples in each category, which have the sequencing depth and LGB score labels of the corresponding category; the data filled in the "BRCA Positive Rate" column refers to the proportion of BRCA-positive samples in the total number of 42 samples; the data filled in the "LGB Positive Rate" column refers to the proportion of LGB-positive samples in the total number of 42 samples.

[0131] Table 2. HRD results of LGB test

[0132]

[0133] In addition, under the conditions of 0.5, 0.2, 0.1 and 0.05 sequencing depths, the detection rate of LGB score for HRD-positive samples was much greater than the detection rate measured by the traditional BRCA analysis method (i.e., BRCA-positive rate) (Table 2), indicating that the application scope of LGB score in HRD detection is much wider than that of traditional BRCA analysis method.

[0134] Furthermore, the robustness of the LGB method is illustrated by the good correlation between the LGB scores detected with a sequencing depth of 0.5 and those detected with lower sequencing depths of 0.01 and 0.2 ( Figure 6). Specifically, when the coverage was 0.1 or 0.2, the R value showed a good correlation, indicating that the best detection results of the LGB method can be guaranteed even when the sequencing depth is as low as 0.1.

[0135] Example 6

[0136] This example further validates the robustness of the HRD risk score as an HRD biomarker by comparing it to the traditional method of BRCA analysis test. Similar to Example 5, the HRD risk test was applied to 42 samples. Figure 7 As shown in Table 3, at different sequencing depths of 0.01-0.5, the HRD risk score of BRCA-positive samples is significantly higher than that of BRCA-negative samples, indicating that the HRD risk score can well distinguish BRCA-positive and BRCA-negative samples. As shown in Table 3, the data filled in the "defective" column and the "complete" column refer to the number of samples in each category, which have the sequencing depth and HRD risk score labels of the corresponding category; the data filled in the "BRCA positive rate" column refers to the proportion of BRCA positive samples to the total number of 42 samples; the data filled in the "HRD risk positive rate" column refers to the proportion of HRD risk positive samples to the total number of 42 samples.

[0137] Table 3. HRD results of HRD risk test

[0138]

[0139]

[0140] Furthermore, the robustness of the HRD risk test is exemplified by the good correlation between the HRD risk score detected with a sequencing depth of 0.5 and the HRD risk score detected with a lower sequencing depth of 0.01-0.2 ( Figure 8 ). Specifically, when the coverage was 0.05, 0.1, or 0.2, the R value showed a good correlation, indicating that the best detection results of the LGB method can be guaranteed even if the sequencing depth is as low as 0.05.

[0141] According to the results of the HRD risk test, the optimal cutoff value fluctuates with different sequencing depths. The optimal cutoff values ​​for 0.2 and 0.1 sequencing depths are both 0.3, while the optimal cutoff values ​​for 0.05 or 0.02 sequencing depths are 0.25 or 0.2, respectively. In addition, the analysis results of the HRD risk test at these cutoff values ​​are shown in Tables 4, 5, and 6. (Please note: for a sequencing depth of 0.5, the optimal cutoff value is 0.5).

[0142] Table 4. Results of HRD risk testing using sequencing depths of 0.1 and 0.2

[0143]

[0144] Table 5. Results of HRD risk test using 0.05 sequencing depth

[0145]

[0146] Table 6. Results of HRD risk test using 0.02 sequencing depth

[0147]

[0148]

[0149] Example 7

[0150] This example compares the LGB score analysis test method and the HRD risk score analysis test method with the commercially available Myriad A CDx test (Myriad Genetics, Salt Lake City, UT, USA) was compared.

[0151] Myriad The CDx assay was applied to three samples and compared with the results of the assay described in this article. Tables 7, 8, and 9 summarize the results of the Myriad assay, the LGB assay, and the HRD risk assay and their corresponding scores at various sequencing depths.

[0152] Table 7. HRD results using 0.5 sequencing depth

[0153]

[0154] Note:

[0155] 1. Myriad threshold score ≥42 is positive, otherwise it is negative.

[0156] 2. LGB threshold score ≥30 is positive, otherwise it is negative.

[0157] 3. HRD risk threshold score ≥ 0.5 is positive, otherwise it is negative.

[0158] Table 8. HRD results using 0.2 sequencing depth

[0159]

[0160] Note:

[0161] 1. LGB threshold score ≥30 is positive, otherwise it is negative.

[0162] 2. HRD risk threshold score ≥ 0.3 is positive, otherwise it is negative.

[0163] Table 9. HRD results detected using 0.1 sequencing depth

[0164]

[0165]

[0166] Note:

[0167] 1. LGB threshold score ≥30 is positive, otherwise it is negative.

[0168] 2. HRD risk threshold score ≥ 0.3 is positive, otherwise it is negative.

[0169] Under all conditions at different sequencing depths, the results of the LGB test and HRD risk test of the present disclosure were consistent with those of the Myriad method, which confirms that both methods described herein are capable of detecting HRD in cancer patients.

[0170] While the present disclosure has been particularly shown and described with reference to specific embodiments, some of which are preferred embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present disclosure disclosed herein.

Claims

1. A method for detecting defects in the DNA homologous recombination pathway in cancer patients, characterized in that: The method comprises: obtaining WGS data of a tumor sample from the patient; Generating a genomic variation profile from the WGS data, wherein the genomic variation profile includes a genomic copy number profile, a fusion profile, a translocation profile, or an inversion profile; generating a chromosomal instability (CI) score for each chromosome from the genomic copy number profile using a deep learning model, thereby generating a CI score set; generating a risk score for the DNA homologous recombination pathway deficiency from the CI score group; determining that the risk score is greater than a threshold number; and The patient is determined to be at risk for having a defect in the DNA homologous recombination pathway.

2. The method according to claim 1, characterized in that (i) the threshold number is 0.4; or (ii) when the risk score is greater than 0.6, the risk of the DNA homologous recombination pathway deficiency is high; or (iii) when the risk score is greater than 0.4 and less than or equal to 0.6, the risk of the DNA homologous recombination pathway deficiency is medium.

3. The method according to claim 1, characterized in that The tumor sample is a FFPE sample or a CTC sample.

4. The method according to claim 1, characterized in that The cancer is selected from the group consisting of breast cancer, ovarian cancer, pancreatic cancer, head and neck cancer, and melanoma.

5. The method according to claim 1, characterized in that The genomic copy number profile is selected from any one or a combination of the following parameters: genomic segments along the genome, the copy number of the genomic segments, copy number variation (CNA) breakpoints, or the number of the CNA breakpoints.

6. The method according to claim 1, characterized in that The genomic copy number profile is generated by the following steps: aligning the WGS data to a reference genome to generate a set of aligned WGS reads along the genome, Counting the WGS reads aligned to the genome, and Genomic segments are generated along the genome, wherein two adjacent genomic segments have significantly different numbers of read counts.

7. The method according to claim 1, characterized in that The deep learning model includes one or more long short-term memory layers or convolutional neural network layers, and a fully connected network.

8. The method according to claim 1, characterized in that The risk score is the average of the CI score group weighted by the base length of each chromosome.

9. A method for treating cancer in a patient, characterized in that The method comprises administering to the patient a therapeutically effective amount of a PARP inhibitor and / or an alkylating agent, The patient has been identified as having a risk of having a defect in the DNA homologous recombination pathway by the method of claim 1.

10. The method according to claim 9, characterized in that The PARP inhibitor and / or alkylating agent is selected from iniparib, olaparib, rucaparib, CEP9722, MK4827, BMN673, 3-aminobenzamide, platinum complex, bis(2-chloroethyl)methylamine (chlormethine), chlorambucil, melphalan, cyclophosphamide, ifosfamide, estramustine, carmustine, lomustine, fotemustine, streptozocin, busulfan, pipobroman, procarbazine, dacarbazine, thiotepa and temozolomide.

11. A non-transitory computer readable medium having instructions stored thereon, characterized in that: When executed by a processor, the instructions cause the processor to perform a method comprising the following steps: Obtain WGS data (whole genome sequencing data) of tumor samples from patients; Generate a genomic variation spectrum from the WGS data, wherein the genomic variation spectrum includes a genomic copy number spectrum, a fusion spectrum, a translocation spectrum, or an inversion spectrum; generating a chromosomal instability (CI) score for each chromosome from the genomic copy number profile using a deep learning model, thereby generating a CI score set; and A risk score for DNA homologous recombination pathway deficiency is generated from the CI score set.

12. The non-transitory computer readable medium of claim 11, wherein: The method further comprises: determining that the risk score is greater than a threshold number; and The patient is determined to be at risk for having a defect in the DNA homologous recombination pathway.

13. A method for detecting a defect in the DNA homologous recombination pathway in a cancer patient, characterized in that: The method comprises: Obtaining WGS data (whole genome sequencing data) of a tumor sample from the patient, wherein the WGS data has a sequencing depth of 0.05-0.5; generating a genomic copy number profile from the WGS data; According to the genome copy number profile, a plurality of large-scale genomic breakpoints (LGBs) are generated for each genome, wherein the LGBs are breakpoints between two adjacent genomic fragments of different copy numbers, and each of the genomic fragments is at least 10 megabases in length; determining that the number of LGBs is greater than a threshold number of LGBs; and The patient is determined to be at risk for having a defect in the DNA homologous recombination pathway.

14. The method according to claim 13, characterized in that (i) the LGB threshold number is 25; or (ii) when the number of LGBs is greater than 35, the risk of the DNA homologous recombination pathway defect is high; or (iii) when the number of LGBs is greater than 25 and less than or equal to 35, the risk of the DNA homologous recombination pathway defect is medium.

15. The method according to claim 13, characterized in that The tumor sample is a FFPE sample or a CTC sample.

16. The method according to claim 13, characterized in that The cancer is selected from the group consisting of breast cancer, ovarian cancer, pancreatic cancer, head and neck cancer, and melanoma.

17. The method according to claim 13, characterized in that The genomic copy number profile is generated by the following steps: aligning the WGS data to a reference genome to generate a set of aligned WGS reads along the genome, Counting the WGS reads aligned to the genome, and Genomic fragments are generated along the genome, wherein two adjacent genomic fragments have significantly different read counts.

18. A method for treating cancer in a patient, the method comprising administering to the patient a therapeutically effective amount of a PARP inhibitor and / or an alkylating agent, The patient has been identified as being at risk of having a defect in the DNA homologous recombination pathway by the method of claim 13.

19. The method according to claim 18, wherein the PARP inhibitor and / or alkylating agent is selected from iniparib, olaparib, rucaparib, CEP9722, MK4827, BMN673, 3-aminobenzamide, platinum complex, bis(2-chloroethyl)methylamine (chlormethine), chlorambucil, melphalan, cyclophosphamide, ifosfamide, ifosfamide, estramustine, carmustine, lomustine, fotemustine, streptozocin, busulfan, pipobroman, procarbazine, dacarbazine, thiotepa, and temozolomide.

20. A non-transitory computer readable medium having instructions stored thereon, characterized in that: The instructions, when executed by a processor, cause the processor to perform the method of claim 13 .

21. A PARP inhibitor and / or alkylating agent suitable for use in a method of treating cancer in a patient, characterized in that: The patient has been identified as being at risk of having a defect in the DNA homologous recombination pathway by the method of claim 1 or 12.

22. The PARP inhibitor and / or alkylating agent for use in claim 21, characterized in that: The PARP inhibitor or alkylating agent is selected from enaparib, olaparib, rucaparib, CEP9722, MK4827, BMN673, 3-aminobenzamide, platinum complex, bis(2-chloroethyl)methylamine, chlorambucil, melphalan, cyclophosphamide, ifosfamide, estramustine, carmustine, lomustine, fotemustine, streptozotocin, busulfan, pipobroman, procarbazine, dacarbazine, thiotepa and temozolomide.

Images