Cancer diagnosis and classification using non-human metagenomic pathway analysis
Patent Information
- Application Number
- JP2023528760
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2020-11-16
- Filing Date
- 2021-11-16
- Publication Date
- 2024-11-15
AI Technical Summary
Existing cancer diagnosis methods based on taxonomic classifications of microbial DNA do not provide information on the functional capabilities of microbial communities within tumors, limiting the understanding of their contribution to tumor biology and therapeutic responses.
The method involves analyzing the presence and abundance of functional microbial genes and biochemical pathways in biopsy samples to diagnose cancer and predict treatment responses by mapping non-human sequencing reads to protein databases and generating diagnostic models using machine learning algorithms.
This approach allows for accurate cancer diagnosis and prediction of treatment responses by elucidating the role of intratumoral microbiota in metabolite consumption and production, providing insights into tumor-specific biology and personalized therapy options.
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
Technical Field
[0001] Cross-reference This application claims the benefit of U.S. Provisional Patent Application No. 63 / 114,447, filed Nov. 16, 2020, which is hereby incorporated by reference in its entirety.
Background Art
[0002] Recent research on various cancer types has shown that tumors harbor an endogenous microbiome, which can be utilized for improved prognosis, diagnosis, treatment selection, and to enhance our understanding of intratumoral biology. To date, reports have provided evidence of tumor-specific microbiomes in breast, prostate, colon, brain, bone, skin, and pancreatic cancers. There is also an active area of discussion regarding simply how microbes come to colonize tumors, but regardless of etiology, it has been demonstrated that cancer-specific microbial associations can be exploited for diagnostic purposes by detection based on sequencing of microbial nucleic acids. Indeed, Poore et al. showed that detection of microbial DNA (mbDNA) fragments in patient plasma samples can accurately discriminate between various cancer and non-cancer samples (PMID: 32214244 and PCT WO 2020 / 093040).
[0003] In Poore et al., metagenomic shotgun sequencing data derived from total plasma cell-free DNA (which inevitably contains a mixture of human cfDNA and microbial cfDNA) were computer-sorted according to whether the sequenced reads mapped to a human reference genome. All unmapped (i.e., non-human) reads were then classified down to the genus level using a rapid k-mer mapping approach (Kraken, PMID:24580807). The output of the Kraken analysis is a list of taxonomic classifications for the sequenced reads in the sample, and the read counts associated with each taxonomic assignment. In Poore et al., such paired data (genus and read counts) from HIV-negative, healthy donors and cancer cohorts (lung, prostate, and melanoma) were used as input for a machine learning classification algorithm to identify features specific to each cancer type. One disadvantage of using taxonomy-based classification is that while taxonomic assignments are useful for cancer classification, they do not directly inform what cancer-specific biochemical capabilities (if any) may be provided by tumor-associated microbiota. Having methods that enable both cancer classification and diagnosis, while also providing information related to the presence / abundance of biochemical capabilities, can help elucidate how intratumoral microbiota contribute to tumor-specific biology by either providing or consuming metabolites required or produced by the tumor.
[0004] Other prior art relevant to this field is listed below: U.S. Patent Application Publication 2018 / 0223338 describes the use of a solid tissue microbiome or salivary microbiome in the identification and diagnosis of head and neck cancer; U.S. Patent Application Publication 2018 / 0258495Al describes the use of a solid tissue microbiome or fecal microbiome to detect colon cancer, several types of mutations associated with colon cancer, and a kit for collecting and amplifying the corresponding microorganisms; PCT WO 2019 / 191649 describes the use of cell-free microbial DNA and a machine learning model to distinguish subjects with advanced adenoma and / or colorectal cancer from healthy subjects, the machine learning algorithm relying on DNA sequence reads that map to a reference genome as input for analysis. [Overview of the project]
[0005] This disclosure provided herein describes systems and methods for accurately diagnosing or determining the presence or absence of cancer and other diseases, their subtypes, and their likelihood of responding to a particular treatment, using only non-human nucleic acids obtained from tissue or liquid biopsy samples. Specifically, the invention provides methods for identifying the presence and abundance of microbial functional genes (and their fragments) and biochemical pathways present in a biopsy sample (e.g., liquid or tissue biopsy). In some cases, microbial functional genes and biochemical pathways can be used to train one or more models and / or predictive models described elsewhere in this specification. Such trained models can output a determination of the presence or absence of a cancer of interest, or the likelihood of a treatment response and / or effectiveness if the subject is treated.
[0006] The methods of the present invention disclosed herein provide methods for generating diagnostic models that can diagnose and classify cancer, while also providing information relating to the presence and / or abundance of biochemical capabilities to elucidate the contribution of the intratumor microbiota to tumor-specific biology. In some cases, tumor-specific biology may relate to how the intratumor microbiota contributes to the consumption of metabolites required or produced by the tumor. For example, pathway-based analysis can help shed light on microbial-catalyzed therapeutic small molecule transformations, the enzymatic activity of which can alter the in vivo efficacy of the molecule. To give a concrete example using a therapeutic case where microbial activity is directly related (bacterial-mediated deamination of the cytidine moiety in the chemotherapy drug gemcitabine), it has been shown that bacteria expressing a long isoform of cytidine deaminase (cdd) can convert the active form of gemcitabine to the less therapeutically effective 2'2-difluorodeoxyuridine (PMID:28912244). This biochemical test example demonstrates that the present invention, as disclosed herein, aims to address the unmet need for cancer diagnosis in a subject by detecting the circulating microbial DNA of the subject, as detailed by Poore et al., while simultaneously detecting the presence / absence or amount of cancer-related isoforms of cdd. In consideration of this example, in some embodiments, the methods disclosed herein may not be limited solely to the diagnosis of cancer in a subject, but may also aim to predict that a subject may not respond to gemcitabine treatment if it is found to have a long isoform of cdd.
[0007] Aspects of the disclosure provided herein include, in some embodiments, a method for determining the presence or absence of a target cancer. In some embodiments, the method includes (a) preparing one or more sequencing reads of a target biological sample; (b) filtering the sequencing reads using a genome database to produce a set of filtered non-human sequencing reads; (c) translating the non-human sequencing reads into non-human proteins; (d) mapping the non-human proteins into a protein database to produce a set of protein database relevances; and (e) determining the presence or absence of a target cancer as an output to a trained model, provided the trained model is provided with input to the set of protein database relevances. In some embodiments, the set of protein database relevances includes a set of functional genes, biochemical pathways, or a combination thereof. In some embodiments, the method further includes, prior to (c), a step of removing contaminant non-human sequencing reads by purging the filtered non-human sequencing reads. In some embodiments, the translation step is completed in silico. In some embodiments, the biological sample is a tissue, a fluid biopsy, or a combination thereof. In some embodiments, the subject is a human or a non-human mammal. In some embodiments, the biological sample comprises a nucleic acid composition, which comprises DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or a combination thereof. In some embodiments, the genome database is a human genome database. In some embodiments, the trained model is trained with a set of functional gene and biochemical pathway abundances that are present or absent, along with characteristic abundances for the cancer of interest. In some embodiments, the non-human sequences originate from the biological origins of bacteria, archaea, fungi, viruses, or a combination thereof. In some embodiments, the trained model is configured to determine the category or tissue-specific location of the cancer of interest. In some embodiments, the trained model is configured to determine one or more types of the cancer of interest.In some embodiments, the trained model is configured to determine one or more subtypes of the target cancer. In some embodiments, the trained model is configured to determine the stage of the target cancer, the target cancer prognosis, or any combination thereof. In some embodiments, the trained model is configured to determine the presence or absence of cancer in low-stage (stage I or stage II) tumors. In some embodiments, the trained model is configured to determine the immunotherapy response of one or more subjects in a second set if immunotherapy is provided to one or more subjects in a second set. In some embodiments, the method further includes the step of using the trained model to output a treatment for a subject to treat the target cancer, the subject will respond with positive therapeutic efficacy if administered this treatment. In some embodiments, the cancers covered include acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof. In some embodiments, the liquid biopsy includes plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled breath condensate, or any combination thereof. In some embodiments, the filtering step includes filtering the sequencing reads by computer using a program such as bowtie2, Kraken, or any combination thereof. In some embodiments, the protein database is the UniRef database. In some embodiments, the translation step is achieved by a software package such as BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or any combination thereof.In some embodiments, the step of mapping non-human proteins to biochemical pathways is achieved by mapping non-human proteins to databases such as KEGG, MetaCyc, PANTHER pathway, PathBank, or a combination thereof. In some embodiments, the biochemical pathways are generated using the software package MinPath.
[0008] Aspects of the present disclosure describe a method for providing a determination of the presence or absence of a target cancer, which in some embodiments includes: (a) sequencing a nucleic acid composition of a target biological sample to generate sequencing reads; (b) filtering the sequencing reads using a genome database to produce a set of filtered non-human sequencing reads; (c) translating the non-human sequencing reads into non-human proteins; (d) mapping the non-human proteins into a protein database to produce a set of protein database relevances; and (e) providing a trained model, if the set of protein database relevances is provided as input to the trained model, a determination of the presence or absence of a target cancer as output of the trained model. In some embodiments, the set of protein database relevances includes a set of functional genes, biochemical pathways, or a combination thereof. In some embodiments, the method further includes, prior to (c), a step of removing contaminant non-human sequencing reads by purging the filtered non-human sequencing reads. In some embodiments, the translation step is completed in silico. In some embodiments, the biological sample is a tissue, a fluid biopsy sample, or a combination thereof. In some embodiments, the subject is a human or a non-human mammal. In some embodiments, the biological sample comprises a nucleic acid composition, which comprises DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or a combination thereof. In some embodiments, the genome database is a human genome database. In some embodiments, the trained model is trained with a set of functional gene and biochemical pathway abundances that are present or absent, along with characteristic abundances for the cancer of interest. In some embodiments, the non-human sequences originate from the biological origins of bacteria, archaea, fungi, viruses, or a combination thereof. In some embodiments, the trained model is configured to determine the category or tissue-specific location of the cancer of interest. In some embodiments, the trained model is configured to determine one or more types of the cancer of interest.In some embodiments, the trained model is configured to determine one or more subtypes of the target cancer. In some embodiments, the trained model is configured to determine the stage of the target cancer, the target cancer prognosis, or any combination thereof. In some embodiments, the trained model is configured to determine the presence or absence of cancer in low-stage (stage I or stage II) tumors. In some embodiments, the trained model is configured to determine the immunotherapy response of the target if the target is offered immunotherapy. In some embodiments, the method further includes the step of using the trained model to output a therapy for the target to treat the target cancer, the target would respond with positive therapeutic efficacy if administered this therapy. In some embodiments, the cancers covered include acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof. In some embodiments, the liquid biopsy includes plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled breath condensate, or any combination thereof. In some embodiments, the filtering step includes filtering the sequencing reads by computer using a program such as bowtie2, Kraken, or any combination thereof. In some embodiments, the protein database is the UniRef database. In some embodiments, the translation step is achieved by a software package such as BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or any combination thereof.In some embodiments, the step of mapping non-human proteins to biochemical pathways is achieved by mapping non-human proteins to databases such as KEGG, MetaCyc, PANTHER pathway, PathBank, or a combination thereof. In some embodiments, the biochemical pathways are generated using the software package MinPath.
[0009] Aspects of the disclosure provided herein describe a method for training a model configured to determine the presence or absence of a cancer of interest, which in some embodiments includes: (a) preparing a dataset comprising nucleic acid sequencing reads of nucleic acid compositions of one or more subjects in a first set, and corresponding one or more cancers of one or more subjects in the first set; (b) filtering the nucleic acid sequencing reads using a genome database build to generate non-human sequencing reads; (c) translating the non-human sequencing reads into non-human proteins; (d) mapping the non-human proteins into a protein database to create a set of protein database relevances; and (e) training a model using the set of protein database relevances and corresponding one or more cancer states of one or more subjects in the first set to generate a trained model configured to determine the presence or absence of a cancer of one or more subjects in a second set. In some embodiments, the set of protein database relevances comprises a set of functional genes, biochemical pathways, or a combination thereof. In some embodiments, the method further includes a step of removing contaminant non-human sequencing reads by filtering the non-human sequencing reads for contaminants prior to (c). In some embodiments, the translation step is completed in silico. In some embodiments, the biological sample is a tissue, a liquid biopsy sample, or a combination thereof. In some embodiments, the subject is a human or a non-human mammal. In some embodiments, the biological sample comprises a nucleic acid composition, the nucleic acid composition comprising DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or a combination thereof. In some embodiments, the genome database is a human genome database. In some embodiments, the trained model is trained with a set of functional gene and biochemical pathway abundances that are present or absent, along with characteristic abundances for the cancer of interest.In some embodiments, the non-human sequence originates from a biological origin of bacteria, archaea, fungi, viruses, or any combination thereof. In some embodiments, the trained model is configured to determine the category or tissue-specific location of cancer in one or more subjects from a second set. In some embodiments, the trained model is configured to determine one or more types of cancer in one or more subjects from a second set. In some embodiments, the trained model is configured to determine one or more subtypes of cancer in one or more subjects from a second set. In some embodiments, the trained model is configured to determine the stage, prognosis, or any combination thereof of cancer in one or more subjects from a second set. In some embodiments, the trained model is configured to determine the presence or absence of cancer in one or more subjects from a second set in low-stage (stage I or stage II) tumors. In some embodiments, the trained model is configured to determine the immunotherapy response of a subject if the subject is offered immunotherapy. In some embodiments, the method further includes the step of using a trained model to output a treatment for cancer in one or more subjects of a second set, the one or more subjects of the second set, which would respond with positive therapeutic efficacy if administered with this treatment. In some embodiments, the cancers of one or more subjects in the first and second sets include acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof.In some embodiments, the liquid biopsy includes plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled condensate, or any combination thereof. In some embodiments, the filtering step includes filtering sequencing reads by computer using a program such as bowtie2, Kraken, or any combination thereof. In some embodiments, the protein database is the UniRef database. In some embodiments, the translation step is achieved by a software package such as BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or any combination thereof. In some embodiments, the step of mapping non-human proteins to biochemical pathways is achieved by mapping non-human proteins to a database such as KEGG, MetaCyc, PANTHER pathway, PathBank, or any combination thereof. In some embodiments, the biochemical pathways are generated using the software package MinPath. In some embodiments, the dataset further includes corresponding previous or current treatments administered to one or more subjects in a first set. In some embodiments, the dataset further includes the treatment efficacy of previous or current treatment administrations to one or more subjects in a first set.
[0010] Aspects of the disclosure provided herein describe a computer implementation method for utilizing a trained predictive model to provide therapeutic treatment predictions for one or more subjects, which in some embodiments include: (a) receiving nucleic acid sequencing reads and corresponding cancer classifications of a first set of one or more subjects; (b) filtering the nucleic acid sequencing reads using a genome database build to generate non-human sequencing reads; (c) translating the non-human sequencing reads into non-human proteins; (d) mapping the non-human proteins into a protein database to create a set of protein database relevances; and (e) if the set of protein database relevances is provided to the trained predictive model as input, utilizing the trained predictive model to provide treatment predictions for the first set of one or more subjects. In some embodiments, the trained predictive model is trained on nucleic acid sequencing reads, corresponding cancer classifications, corresponding treatments administered, corresponding treatment responses, or any combination thereof of a second set of one or more subjects. In some embodiments, one or more subjects in the second set are different from one or more subjects in the first set. In some embodiments, the protein database relevance set includes a set of functional genes, biochemical pathways, or a combination of either. In some embodiments, the method further includes the step of removing contaminant non-human sequencing reads by filtering the non-human sequencing reads prior to (c). In some embodiments, the translation step is completed in silico. In some embodiments, the biological sample is a tissue, a liquid biopsy sample, or a combination of either. In some embodiments, one or more subjects in the first and / or second set are human or non-human mammals. In some embodiments, the biological sample nucleic acid composition includes DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or a combination of either. In some embodiments, the genome database is a human genome database.In some embodiments, the non-human sequence originates from a biological origin of bacteria, archaea, fungi, viruses, or any combination thereof. In some embodiments, if immunotherapy is administered to one or more subjects in a first set, the treatment prediction includes the immunotherapy response of one or more subjects in the first set. In some embodiments, the treatment prediction includes the therapeutic efficacy that one or more subjects in the first set will respond with positive efficacy. In some embodiments, the cancer classification includes acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof. In some embodiments, the liquid biopsy includes plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled breath condensate, or any combination thereof. In some embodiments, the filtering step includes filtering sequencing reads by computer using a program such as bowtie2, Kraken, or any combination thereof. In some embodiments, the protein database is the UniRef database. In some embodiments, the translation step is achieved by a software package such as BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or any combination thereof. In some embodiments, the step of mapping non-human proteins to biochemical pathways is achieved by mapping non-human proteins to a database such as KEGG, MetaCyc, PANTHER pathway, PathBank, or any combination thereof. In some embodiments, the biochemical pathways are generated using the software package MinPath.
[0011] Aspects of the disclosure provided herein include, in some embodiments, a method for modifying a target cancer treatment using a trained predictive model. In some embodiments, the method includes (a) preparing one or more sequencing reads of a target biological sample along with the cancer, cancer type, and treatments administered to treat the cancer; (b) filtering the sequencing reads using a genome database to create a set of filtered non-human sequencing reads; (c) translating the non-human sequencing reads into non-human proteins; (d) mapping the non-human proteins into a protein database to create a set of protein database relevances; and (e) modifying the target cancer treatment if the administered treatment differs from the treatment recommendation output by the trained predictive model when input using the set of protein database relevances. In some embodiments, the trained predictive model is trained on nucleic acid sequencing reads of a second set of one or more target biological samples, corresponding cancer classifications, corresponding administered treatments, corresponding treatment responses, or any combination thereof. In some embodiments, the one or more targets in the second set are different from the one or more targets in the first set. In some embodiments, the protein database relevance set includes a set of functional genes, biochemical pathways, or a combination thereof. In some embodiments, the method further includes the step of removing contaminant non-human sequencing reads by filtering the non-human sequencing reads prior to (c). In some embodiments, the translation step is completed in silico. In some embodiments, the biological sample is a tissue, a liquid biopsy sample, or a combination thereof. In some embodiments, the subject is a human or a non-human mammal. In some embodiments, the biological sample nucleic acid composition includes DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or a combination thereof. In some embodiments, the genome database is a human genome database. In some embodiments, the non-human sequences originate from the biological origins of bacteria, archaea, fungi, viruses, or a combination thereof.In some embodiments, if the subject is administered immunotherapy, the treatment recommendation includes the subject's immunotherapy response. In some embodiments, the treatment recommendation includes therapies in which the subject would respond with positive efficacy. In some embodiments, the cancers covered include acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof. In some embodiments, the liquid biopsy includes plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled breath condensate, or any combination thereof. In some embodiments, the filtering step includes filtering sequencing reads by computer using a program such as bowtie2, Kraken, or any combination thereof. In some embodiments, the protein database is the UniRef database. In some embodiments, the translation step is achieved by a software package such as BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or any combination thereof. In some embodiments, the step of mapping non-human proteins to biochemical pathways is achieved by mapping non-human proteins to a database such as KEGG, MetaCyc, PANTHER pathway, PathBank, or any combination thereof. In some embodiments, the biochemical pathways are generated using the software package MinPath.
[0012] Embodiments disclosed herein provide a method for creating a diagnostic model for diagnosing cancer in a subject based on taxonomy-independent non-human functional gene abundances in a biological sample, comprising: (a) sequencing a nucleic acid composition in a biological sample to generate sequencing reads; (b) filtering the sequencing reads using a genome database build to isolate non-human sequencing reads; (c) translating the composition of the non-human sequencing reads in silico to identify non-human proteins represented in the non-human sequencing reads; (d) mapping the non-human proteins to a non-human protein database of non-human functional genes and biochemical pathways; (e) generating a table of functional gene and biochemical pathway abundances using non-human functional genes and biochemical pathways; (f) analyzing the table of biochemical pathway abundances using a trained machine learning algorithm; and (g) providing a diagnosis of the presence or absence of cancer in the subject using the output of the trained machine learning algorithm. In some embodiments, the biological sample is tissue, a fluid biopsy sample, or a combination thereof. In some embodiments, the subject is human or a non-human mammal. In some embodiments, the nucleic acid composition includes the total population of DNA, RNA, cell-free DNA (cfDNA), cell-free RNA (cfRNA), exosomal DNA, exosomal RNA, or a combination thereof. In some embodiments, the genome database is the human genome database. In some embodiments, the output of the trained machine learning algorithm includes the analysis of a table of functional gene and biochemical pathway abundances. In some embodiments, the trained machine learning algorithm is trained using a set of functional gene and biochemical pathway abundances known to be present or absent in characteristic abundances in the cancer of interest. In some embodiments, the diagnostic model utilizes biochemical pathway abundance information from one or more of the following biological domains: bacteria, archaea, and / or fungi.In some embodiments, the diagnostic model diagnoses the category or tissue-specific location of cancer. In some embodiments, the diagnostic model is used to diagnose one or more types of cancer in a subject. In some embodiments, the diagnostic model is used to diagnose one or more subtypes of cancer in a subject. In some embodiments, the diagnostic model is used to predict the stage of cancer in a subject and / or the prognosis of cancer in a subject. In some embodiments, the diagnostic model is used to diagnose the type of cancer in low-stage (stage I or stage II) tumors. In some embodiments, the diagnostic model is used to predict the immunotherapy response of a subject. In some embodiments, the diagnostic model is used to select the optimal treatment for a particular subject. In some embodiments, the diagnostic model is used to model the long-term course of response to one or more cancer treatments and to adjust subsequent treatment regimens. In some embodiments, the diagnostic model diagnoses one or more of the following: acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade brain glioma, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, or uveal melanoma. In some embodiments, the diagnostic model identifies and removes certain non-human features as contaminants, referred to as noise, while selectively retaining other non-human features referred to as signals. In some embodiments, the liquid biopsy sample includes, but is not limited to, one or more of the following: plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, or breath condensate. In some embodiments, the filtering step includes filtering sequencing reads by computer using bowtie2, the Kraken program, or a combination thereof. In some embodiments, the protein database is the UniRef database.In some embodiments, the step of querying a non-human protein database to identify proteins represented in non-human sequencing reads is performed using the software package DIAMOND. In some embodiments, the biochemical pathway database is the KEGG or MetaCyc database. In some embodiments, the step of generating a table of biochemical pathway abundances is performed using the software package MiniPath.
[0013] Aspects disclosed herein are methods for creating a diagnostic model for diagnosing cancer in a subject based on taxonomy-independent non-human functional gene abundances in a biological sample, comprising: (a) sequencing a nucleic acid composition in a biological sample to generate sequencing reads; (b) filtering the sequencing reads using a genome database build to isolate non-human sequencing reads; (c) mapping the non-human sequencing reads to a database of sequenced genomes; and (d) generating a plurality of mapped genome coordinates between the non-human sequencing reads and the database of sequenced genomes. The present invention provides a method comprising the steps of (e) querying a database of known non-human proteins using multiple mapped genomic coordinates to calculate their abundances, (f) mapping non-human proteins to a database of functional genes and biochemical pathways, (g) generating a table of multiple functional gene and biochemical pathway abundances, (h) analyzing the table of functional gene and biochemical pathway abundances using a trained machine learning algorithm, and (i) diagnosing the presence or absence of a target cancer using the output of the trained machine learning algorithm analysis of the table of multiple functional gene and biochemical pathway abundances. In some embodiments, the diagnostic model utilizes biochemical pathway abundance information from one or more of the following biological domains: bacteria, archaea, and / or fungi. In some embodiments, the biological sample is tissue, a liquid biopsy sample, or a combination thereof. In some embodiments, the subject is human or a non-human mammal. In some embodiments, the nucleic acid composition includes the total population of DNA, RNA, cell-free DNA (cfDNA), cell-free RNA (cfRNA), exosomal DNA, exosomal RNA, or a combination thereof. In some embodiments, the genome database is a human genome database. In some embodiments, the output of a trained machine learning algorithm includes the analysis of tables of abundances of multiple functional genes and biochemical pathways.In some embodiments, the trained machine learning algorithm is trained using a set of functional gene and biochemical pathway abundances known to be present or absent in characteristic abundances in the cancer of interest. In some embodiments, the diagnostic model diagnoses the category or tissue-specific location of the cancer. In some embodiments, the diagnostic model is used to diagnose one or more types of cancer in a subject. In some embodiments, the diagnostic model is used to diagnose one or more subtypes of cancer in a subject. In some embodiments, the diagnostic model is used to predict the stage of cancer in a subject and / or the prognosis of cancer in a subject. In some embodiments, the diagnostic model is used to diagnose the type of cancer in low-stage (stage I or stage II) tumors. In some embodiments, the diagnostic model is used to predict the immunotherapy response of a subject. In some embodiments, the diagnostic model is used to select the optimal treatment for a particular subject. In some embodiments, the diagnostic model is used to model the long-term course of response to one or more cancer treatments and to adjust subsequent treatment regimens. In some embodiments, the diagnostic model diagnoses one or more of the following: acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade brain glioma, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, or uveal melanoma. In some embodiments, the diagnostic model identifies and removes certain non-human features as contaminants, referred to as noise, while selectively retaining other non-human features referred to as signals. In some embodiments, the liquid biopsy includes, but is not limited to, one or more of the following: plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, or exhaled condensate.In some embodiments, the filtering step includes filtering the sequenced reads by computer using botwie2, the Kaken program, or a combination thereof. In some embodiments, the database of sequenced genomes is the Web of Life database. In some embodiments, the protein database is the UniRef database. In some embodiments, the biochemical pathway database is the KEGG or MetaCyc database.
[0014] In some embodiments, the present invention provides a method for generating a broad pattern ("signature") of the presence or abundance of microbial functional genes associated with the presence and / or type of cancer using a liquid biopsy sample. Such "signatures" can then be deployed to diagnose the presence, type, and / or subtype of cancer in humans.
[0015] In some embodiments, the present invention provides a method for generating a wide range of patterns of functional genes or abundances of microorganisms associated with the presence and / or type of cancer using primary tumor tissue. Such “signatures” can then be deployed to diagnose the presence, type, and / or subtype of cancer in a human using a human-derived liquid biopsy sample.
[0016] In some embodiments, the present invention provides a method for broadly diagnosing diseases in mammalian subjects, comprising the steps of: detecting the presence or abundance of microorganisms in a liquid biopsy sample derived from the subject; determining that the functional genes or abundance of the detected microorganisms differ from the functional genes or abundance of microorganisms in a normal liquid biopsy sample; and correlating the functional genes or abundance of the detected microorganisms with the functional genes or abundance of known microorganisms associated with the disease, thereby diagnosing the disease.
[0017] In some embodiments, the present invention provides a method for diagnosing a type of disease in a mammalian subject, comprising the steps of: detecting the presence or abundance of a microorganism in a liquid biopsy sample derived from the subject; determining, using previously studied liquid biopsy samples, whether the functional genes or abundance of the detected microorganism are similar to or different from the functional genes or abundance of the microorganism in a population of cancer and / or healthy patients; and correlating the functional genes or abundance of the detected microorganism with the most similar liquid biopsy sample in this cohort, thereby diagnosing a disease and / or type of disease.
[0018] In some embodiments, the present invention provides a method for predicting whether any subject will respond to or not to a particular treatment for a disease, wherein the disease is cancer, the subject is a human, the treatment is immunotherapy, and the immunotherapy is PD-1 blockade (e.g., nivolumab, pembrolizumab).
[0019] In embodiments, the present invention provides a method for diagnosing a disease, further comprising the step of treating a disease in a subject based on identified non-mammalian features of the disease, wherein the disease is cancer, the non-mammalian features are microbial, and the subject is human.
[0020] In some embodiments, the present invention provides a method for diagnosing a disease, further comprising the step of monitoring its non-mammalian features over a long period of time to indicate a response to disease treatment, wherein the disease is cancer, the non-mammalian features are microbial, and the subject is human.
[0021] In some embodiments, the present invention provides an assay for measuring the functional genes or abundance of microorganisms in a designated tissue sample, thereby enabling the diagnosis of a disease.
[0022] In some embodiments, the present invention utilizes a diagnostic model based on a machine learning architecture. In some embodiments, the present invention utilizes a diagnostic model based on a standardized machine learning architecture.
[0023] In some embodiments, the present invention utilizes a diagnostic model based on an ensemble of machine learning architectures. In some embodiments, the present invention selectively retains other non-mammalian features as non-contaminants called signals, while identifying and selectively removing certain non-mammalian features as contaminants called noise, where the non-mammalian features are microbial.
[0024] In some embodiments, the present invention provides a method for diagnosing a disease, wherein information on the functional genes or abundance of a microorganism is combined with additional information on the host (target) and / or the host's (target's) cancer to create a diagnostic model that has better predictive performance than information on the functional genes or abundance of a microorganism alone.
[0025] In some embodiments, the diagnostic model utilizes information combined with information on the functional genes or abundances of microorganisms from one or more of the following sources: cell-free tumor DNA, cell-free tumor RNA, exosome-derived tumor DNA, exosome-derived tumor RNA, circulating tumor cell-derived DNA, circulating tumor cell-derived RNA, methylation patterns of cell-free tumor DNA, methylation patterns of cell-free tumor RNA, methylation patterns of circulating tumor cell-derived DNA, and / or methylation patterns of circulating tumor cell-derived RNA.
[0026] In some embodiments, the functional genes or abundance of a microbial organism are detected by one or more nucleic acid detection methods, including metagenomic shotgun sequencing, targeted microbial sequencing, host whole-genome sequencing, host transcriptome sequencing, cancer whole-genome sequencing, and cancer transcriptome sequencing.
[0027] In some embodiments, the microbial nucleic acids are detected simultaneously with host-derived nucleic acids and then distinguished.
[0028] In some embodiments, prior to measurement (e.g., sequencing) of the combined nucleic acid pool, the host nucleic acids are selectively depleted and the microbial nucleic acids are selectively retained.
[0029] In some embodiments, the present invention provides that the tissue is blood, a blood component (e.g., plasma) or a tissue biopsy, and that the tissue biopsy can be malignant or non-malignant.
[0030] In some embodiments, the functional genes or abundance of cancer-associated microorganisms are determined by measuring the functional genes or abundance of microorganisms at other locations in the host.
[0031] The novel features of the invention are set forth with particularity in the appended claims. A better understanding of the features and advantages of the invention will be obtained from the following detailed description which illustrates exemplary embodiments in which the principles of the invention are utilized, and from the accompanying drawings thereof.
Brief Description of the Drawings
[0032] [Figure 1A] FIG. 1A is a diagram showing an exemplary diagnostic model training scheme incorporating a metagenomic function profiling module to enable discovery of health and disease-related microbial signatures based on metagenomic function. FIG. 1A shows an exemplary training structure of a diagnostic model. [Figure 1B] FIG. 1B is a diagram showing an exemplary diagnostic model training scheme incorporating a metagenomic function profiling module to enable discovery of health and disease-related microbial signatures based on metagenomic function. FIG. 1B shows the use of the trained model of FIG. 1A to provide diagnosis of disease and classification of disease states, as described in some embodiments herein, where the trained model of FIG. 1A provides new subject data of an unknown disease state. [Figure 2A]Figure 2A shows exemplary workflows for two metagenomic function calculation pipelines. Figure 2A illustrates an exemplary metagenomic workflow using the HUMAnN2.0 pipeline to generate abundance tables of genes and pathways that can be input into the machine learning model in Figure 1A. [Figure 2B] Figure 2B shows an exemplary metagenomic workflow that uses the WolTka pipeline to generate gene and pathway abundance tables that can be input into the machine learning model of Figure 1A, as described in some embodiments herein. [Figure 3] This figure shows the breakdown of research populations for health, cancer, and lung disease used to create predictive models. [Figure 4A] Figure 4A shows the pathway classification of non-human cell-free DNA sequences by HUMAnN2.0 (human) and the Web of Life Toolkit App (Woltka), as described in some embodiments of this specification. [Figure 4B] Figure 4B shows the pathway classification of non-human cell-free DNA sequences by HUMAnN2.0 (human) and the Web of Life Toolkit App (Woltka), as described in some embodiments of this specification. [Figure 5A] Figure 5A shows the detailed mean pathway importance of pathways identified by Woltka analysis of sequenced cf-mbDNA samples from cancer vs. health and cancer vs. lung disease, as described in some embodiments of this specification. [Figure 5B] Figure 5B shows the detailed mean pathway importance of pathways identified by Woltka analysis of sequenced cf-mbDNA samples for cancer vs. health and cancer vs. lung disease, as described in some embodiments of this specification. [Figure 6A] Figure 6A shows receiver operating characteristic curves and area under the curve analysis illustrating the accuracy of various trained predictive models, as described in some embodiments of this specification. [Figure 6B]Figure 6B shows receiver operating characteristic curves and area under the curve analysis illustrating the accuracy of various trained predictive models, as described in some embodiments of this specification. [Figure 6C] Figure 6C shows receiver operating characteristic curves and area under the curve analysis illustrating the accuracy of various trained predictive models, as described in some embodiments of this specification. [Figure 6D] Figure 6D shows receiver operating characteristic curves and area under the curve analysis illustrating the accuracy of various trained predictive models, as described in some embodiments of this specification. [Figure 7] This figure shows the breakdown of research populations targeting cancer and lung disease, thereby allowing cell-free DNA nucleic acid genetic pathway data of such subjects to be used to train predictive models, as described in some embodiments herein. [Figure 8A] Figure 8A shows the receiver operating characteristic curves and calculated area under the curve for each predictive model trained on known cancer stages of subjects and corresponding cell-free mbDNA nucleic acid genetic pathway data, as well as for subjects with cell-free mbDNA nucleic acid genetic pathway data for lung disease. [Figure 8B] Figure 8B shows the receiver operating characteristic curves and calculated area under the curve for each predictive model trained on known cancer stages of subjects and corresponding cell-free mbDNA nucleic acid genetic pathway data, as well as for subjects with cell-free mbDNA nucleic acid genetic pathway data for lung disease. [Figure 8C] Figure 8C shows the receiver operating characteristic curves and calculated area under the curve for each predictive model trained on known cancer stages of subjects and corresponding cell-free mbDNA nucleic acid genetic pathway data, as well as for subjects with cell-free mbDNA nucleic acid genetic pathway data for lung disease. [Figure 8D] Figure 8D shows the receiver operating characteristic curves and calculated area under the curve for each predictive model trained on known cancer stages of subjects and corresponding cell-free mbDNA nucleic acid genetic pathway data, as well as for subjects with cell-free mbDNA nucleic acid genetic pathway data for lung disease. [Figure 9] A diagram of a computer system configured to implement the methods of this disclosure, as described in some embodiments of this specification, is shown. [Modes for carrying out the invention]
[0033] The disclosures provided herein describe methods for accurately diagnosing and / or determining the likelihood of one or more cancers, subtypes, and / or cancers in response to treatment in one or more subjects. In some cases, one or more subjects may be human or non-human mammals. The methods described herein may utilize non-human nucleic acids derived from tissue or fluid biopsy samples. This may be achieved by identifying specific patterns of microbial functional units (i.e., proteins, including but not limited to enzymes, transcription factors, and receptors). Table 1 provides exemplary microbial enzymes that can be used for disease classification, as well as the presence and absence of microbial enzymes ("signatures") in a sample to assign probabilities of certain probabilities, or any combination thereof, that are found to be: (1) the individual has cancer; (2) the individual has cancer from a particular body site; (3) the individual has a particular type of cancer; (4) the cancer, at that time, is highly or lowly likely or responsive to a particular cancer treatment; and (5) the cancer, at that time, is found to have microbiological features (e.g., microbial antigens) that can be targeted for developing personalized therapies to treat the cancer in question. Other applications of this method are reasonably conceivable and can be easily implemented by those skilled in the art.
[0034] [Table 1-1]
[0035] [Table 1-2]
[0036] Table 1-3
[0037] Table 1-4
[0038] Table 1-5
[0039] Table 1-6
[0040] Table 1-7
[0041] Table 1-8
[0042] Table 1-9
[0043] Table 1-10
[0044] Table 1-11
[0045] Table 1-12
[0046] [Table 1-13]
[0047] [Table 1-14]
[0048] [Table 1-15]
[0049] [Table 1-16]
[0050] [Table 1-17]
[0051] [Table 1-18]
[0052] [Table 1-19]
[0053] [Table 1-20]
[0054] [Table 1-21]
[0055] Sample handling and model generation method The methods described herein can use non-human nucleic acids to diagnose conditions traditionally considered to be diseases of the human genome (e.g., cancer). In some embodiments, the methods described herein do not necessarily rely on observed tissue structure, cellular dysplasia, or any other subjective measurements traditionally used to diagnose cancer, and therefore may provide better clinical outcomes compared to typical pathological reports. In some cases, the methods may provide a high degree of sensitivity by focusing solely on microbial nucleic acid sources rather than modified human (i.e., cancerous) nucleic acid sources, which are often modified at very low frequencies in the background of "normal" nucleic acid sources. In some embodiments, the methods disclosed herein may achieve such results by either solid tissue and / or liquid biopsy samples, the latter of which may require minimal sample preparation and may be minimally invasive. In some embodiments, liquid biopsy-based assays can overcome the challenges posed by circulating tumor DNA (ctDNA) assays, which are often plagued by sensitivity issues due to cell-free DNA (cfDNA) derived from non-malignant human cells. In some cases, microbial assays based on liquid biopsies can distinguish cancer types that ctDNA assays typically cannot achieve, because the most common cancer genomic abnormalities are shared among cancer types (e.g., TP53 mutations, KRAS mutations). In some cases, the methods described herein can constrain the size of the signature, and the methods are expected to be performed by those familiar with the technique (e.g., normalized machine learning), and the microbial assays can be made clinically available, for example, by using targeted assay panels for multiplexed quantitative polymerase chain reaction (qPCR) and multiplexed amplicon sequencing.
[0056] In some embodiments, the methods described herein can determine the presence or absence of a cancer of interest by utilizing a trained model and / or a trained predictive model, the model and / or predictive model may include a machine learning model trained on non-human functional gene and biochemical pathway abundances (i.e., non-human signatures) that can be deployed on real-time sequencing data or retrospective sequencing data (i.e., sequencing of data from a database or repository). In some examples, non-human signatures may include microbial signatures. In some cases, the method for determining or diagnosing a cancer of interest may include the step of sequencing a nucleic acid composition of interest. Alternatively, the method for determining or diagnosing a cancer of interest may include the step of accessing sequencing reads of a nucleic acid composition of interest from a biological sample.
[0057] In some embodiments, the methods described herein include (a) collecting a blood sample from a patient during a routine clinic visit; (b) preparing plasma or serum from the blood sample and extracting and amplifying sequences of specific microbial genes previously determined through a pre-trained machine learning model as useful signatures for diagnosing cancer; (c) obtaining a digital read of the presence and / or abundance of these microbial signatures; (d) standardizing the presence and / or abundance data on an adjacent computer or cloud computing infrastructure and feeding it to the pre-trained machine learning model; and (e) reading predictive and somewhat reliable information about the likelihood that the sample is associated with (1) the presence or absence of cancer, (2) the likelihood of being associated with cancer of a particular type or physical location, or (3) the likelihood of being moderate or low in response to a range of cancer treatments; and (f) training the model by continuing to train the machine learning model with the microbial information of the sample if additional information is later input by the user.
[0058] In some examples, the methods described herein may include a method for training a model configured to determine the presence or absence of a cancer of interest. In some examples, the method may include the steps of (a) preparing a dataset containing nucleic acid sequencing reads of nucleic acid compositions of one or more subjects from a first set, and corresponding one or more cancers of one or more subjects from the first set; (b) filtering the nucleic acid sequencing reads using a genome database build to generate non-human sequencing reads; (c) translating the non-human sequencing reads into non-human proteins; (d) mapping the non-human proteins into a protein database to create a set of protein database relevances; and (e) training a model using the set of protein database relevances and corresponding one or more cancer states of one or more subjects from the first set to generate a trained model configured to determine the presence or absence of cancer in one or more subjects from a second set. In some examples, the set of protein database relevances may include a set of functional genes, biochemical pathways, or any combination thereof, as described elsewhere herein. In some examples, the method may further include a step prior to step (c) of removing contaminant non-human sequencing reads by filtering the non-human sequencing reads for contaminants. In some cases, contaminant non-human sequencing reads can be determined from a database of contaminant non-human sequencing reads determined a prior to (a prior) or from experimental data analysis. In some cases, the translating step in step (c) can be completed in silico. In some examples, the method may include a step of sequencing nucleic acid compositions of one or more subjects from a first set, instead of or in addition to step (a). In some cases, the method may further include a step of using a trained model to output a therapeutic agent for treating cancer in one or more subjects from a second set, which would respond with positive therapeutic efficacy if administered.In some cases, the dataset may further include corresponding prior or current treatments administered to one or more subjects in the first set. In some cases, the dataset may further include the treatment efficacy of prior or current treatment administrations to one or more subjects in the first set.
[0059] In some cases, one or more subjects in the first and / or second set may be human or non-human mammals. In some cases, the biological sample may include tissue, a liquid biopsy sample, or a combination thereof. In some cases, the biological sample may include a nucleic acid composition, which may include DNA, RNA, cell-free RNA, exosomal DNA, exosomal RNA, or a combination thereof. In some cases, the non-human sequence may originate from the biological origin of bacteria, archaea, fungi, viruses, or a combination thereof. In some examples, the liquid biopsy may include plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, breath condensate, or a combination thereof.
[0060] In some cases, one or more subjects in the first and / or second set may include cancer. In some cases, cancer may include acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof.
[0061] In some cases, the trained model can be trained using sets of functional gene and biochemical pathway abundances that are present or absent, along with characteristic abundances related to the cancer of interest. In some cases, the trained model can be configured to determine one or more subtypes of cancer in one or more subjects from a second set. In some cases, the trained model can be configured to determine the stage, prognosis, or any combination thereof of cancer in one or more subjects from a second set. In some cases, the trained model can be configured to determine the presence or absence of cancer in one or more subjects from a second set in low-stage (stage I or stage II) tumors. In some cases, the trained model can be configured to determine the immunotherapy response of a subject if the subject is offered immunotherapy. In some cases, the trained model can be configured to determine the category or tissue-specific location of cancer in one or more subjects from a second set. In some cases, the trained model can be configured to determine one or more types of cancer in one or more subjects from a second set.
[0062] In some examples, the genome database may be the human genome database. In some cases, the filtering step in step (b) may involve computer filtering of sequencing reads using a program such as bowtie2, Kraken, or a combination thereof. In some examples, the protein database may be the UniRef database. In some cases, the translating step in step (c) may be accomplished by a software package such as BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or a combination thereof. In some cases, the step in step (d) of mapping non-human proteins to biochemical pathways may be accomplished by mapping non-human proteins to a database such as KEGG, MetaCyc, PANTHER pathway, PathBank, or a combination thereof. In some cases, the biochemical pathways may be generated using the software package MiniPath.
[0063] In some embodiments, the methods of the present invention disclosed herein may include (a) sequencing the nucleic acid content of a liquid biopsy sample and (b) generating a diagnostic model. In some embodiments, the sequencing method may include next-generation sequencing or long-read sequencing (e.g., nanopore sequencing) or a combination thereof. In some embodiments, model 110 may include a diagnostic model. In some embodiments, the diagnostic model may include a trained machine learning algorithm 109, as shown in Figure 1A. In some embodiments, the diagnostic model may be a ruled machine learning model. In some embodiments, the trained machine learning model algorithm may include linear regression, logistic regression, decision trees, support vector machines (SVMs), naive Bayes, k-nearest neighbors (kNNs), k-means, random forest algorithm models or a combination thereof. In some embodiments, the machine learning algorithm may include one or more machine learning algorithms.
[0064] In some embodiments, the machine learning algorithm 109 can be trained using nucleic acid sequencing data 103 derived from nucleic acids from a plurality of known healthy subjects 101 and a plurality of known cancer subjects 102. In some embodiments, the machine learning algorithm 109 can be trained using nucleic acid sequencing data 103 processed by a metagenomic functional bioinformatics pipeline 108, which comprises (a) a step 104 in which all sequencing reads that map to the human genome are computer-filtered; (b) a step 106 in which the remaining non-human microbial sequencing reads 105 are processed to remove sequences derived from common microbial contaminants; and (c) a step 107 in which the remaining reads are analyzed for their translated (i.e., protein) content. In some embodiments, computer filtering of all sequencing reads can be achieved using bowtie2, the Kraken program, or an equivalent thereof.
[0065] In some embodiments, a machine learning algorithm 109 can be trained to produce a trained diagnostic model 110, which can determine microbial signatures 111 associated with and / or indicating healthy subjects and microbial signatures 112 associated with / indicating cancerous subjects.
[0066] In some embodiments, the machine learning algorithm 109 shown in Figure 1A can be further trained using data related to the abundance of functional microbial genes 207 (e.g., enzymes) in a sample (one or more), as shown in Figure 2A. In some embodiments, the abundance of a functional microbial gene can be determined using the bioinformatics pipeline HUMAnN208 shown in Figure 2A, which includes: (a) a step of generating next-generation sequencing (NGS) reads from a liquid biopsy of interest 201; (b) a step of filtering human sequencing reads by bowtie, the Kraken filtering method, or an equivalent thereof 202; (c) a step of generating microbial sequencing as a result of filtering the sequencing reads in (b) 203; (d) a step of searching for the sequenced readings translated into a unitProt reference cluster (UniRef) database such as DIAMOND or an equivalent thereof 204; (e) a step of mapping UniRef hits to pathways by Kyoto Encyclopedia of Genes and Genomes (Kegg), the MetaCyc database, or an equivalent thereof 205; (f) a step of generating a pathway abundance table using MiniPath; and (g) a step of outputting the pathway abundance table for machine learning (ML) analysis 207.
[0067] In some embodiments, the abundance of a functional microbial gene is determined using the bioinformatics pipeline Web of Life Toolkit App (WolTka) 212 or an equivalent thereof, shown in Figure 2B, which includes: (a) a step of generating next-generation sequencing (NGS) 201 from a liquid biopsy of interest; (b) a step of filtering human sequencing reads by the bowtie, kraken filtering method, or an equivalent thereof, in step 202; (c) a step of generating microbial sequencing as a result of filtering the sequencing reads in step (b), in step 203; (d) a step of mapping the sequencing reads in (c) to the Web of Life database using the bowtie2 or an equivalent thereof read alignment tool, in step 209; (e) a step of calculating the abundance of the UniREF gene using the mapping coordinates derived in (d), in step 210; (f) a step of mapping UniRef hits to pathways using KEGG, MetaCyc, or an equivalent thereof, in step 211; and (g) a step of outputting a table of pathway abundances for machine learning (ML) analysis in step 207. The use of such bioinformatics pipelines and databases is intended, but not limited to, to serve as an explanation of computer-aided means that can access microbial gene abundance data, and therefore the use of any substantial equivalent of the bioinformatics described above.
[0068] Aspects disclosed herein provide a method for training a diagnostic model (Figure 1A), comprising: (a) step 108 of preparing a training dataset, (i) abundances of one or more sequenced microorganisms of one or more subjects; (b) step 108 of preparing a test set, (i) abundances of one or more sequenced microorganisms of one or more subjects; (c) step of training the diagnostic model in sample ratios of at least about 10:90, 20:80, 30:70, 40:60, 50:50, 60:40, 70:30, 80:20 or 90:10 for training versus validation samples, respectively; and (d) step of evaluating the diagnostic accuracy of the diagnostic model.
[0069] In some embodiments, the diagnosis made by the trained diagnostic model may include a machine learning signature 111 indicating a healthy (i.e., cancer-free) subject, or a machine learning-derived signature 112 indicating a cancer-positive subject, as shown in Figure 1A. In some embodiments, the trained diagnostic model may identify and remove one or more microbial or non-microbial nucleic acids classified as noise, while selectively retaining one or more other microbial or non-microbial sequences named as signals.
[0070] Diagnostic or predictive methods using trained models In some embodiments, as shown in Figure 1B, a trained diagnostic model 110 can be used to analyze nucleic acid samples 113 derived from a subject with an unknown disease condition, providing a diagnosis of the disease and, if applicable, a classification of the disease state 115.
[0071] In some examples, the disclosures provided herein describe methods for determining the presence or absence of a cancer of interest. In some examples, the method may include (a) preparing one or more sequencing reads of a biological sample of interest; (b) filtering the sequencing reads using a genome database to create a set of filtered non-human sequencing reads; (c) translating the non-human sequencing reads into non-human proteins; (d) mapping the non-human proteins into a protein database to create a set of protein database relevances; and (e) determining the presence or absence of a cancer of interest as an output to a trained model, provided that the set of protein database relevances is provided as input to the trained model. In some examples, the set of protein database relevances may include a set of functional genes, biochemical pathways, or any combination thereof, as described elsewhere herein. In some examples, the method may further include, prior to step (c), removing contaminant non-human sequencing reads by purging the filtered non-human sequencing reads. In some cases, the contaminant non-human sequencing reads can be determined from a database of contaminant non-human sequencing reads determined prior to (a prior) or from experimental data analysis. In some cases, the translation step in step (c) can be completed in silico. In some examples, the method may include a step of sequencing the nucleic acid composition of interest instead of or in addition to step (a). In some examples, the method may further include a step of using a trained model to output a therapeutic agent for treating the cancer of interest, in which the subject would respond with positive therapeutic efficacy if administered.
[0072] In some cases, the subject may be human or a non-human mammal. In some cases, the biological sample may include tissue, a fluid biopsy sample, or a combination thereof. In some cases, the biological sample may include a nucleic acid composition, which may include DNA, RNA, cell-free RNA, exosomal DNA, exosomal RNA, or a combination thereof. In some cases, the non-human sequence may originate from the biological origin of bacteria, archaea, fungi, viruses, or a combination thereof. In some examples, the fluid biopsy may include plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, breath condensate, or a combination thereof.
[0073] In some cases, the subjects may include cancer. In some cases, cancer may include acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof.
[0074] In some cases, the trained model can be trained using sets of functional gene and biochemical pathway abundances that are present or absent, along with characteristic abundances of the target cancer. In some cases, the trained model can be configured to determine one or more subtypes of the target cancer. In some cases, the trained model can be configured to determine the stage, prognosis, or any combination thereof of the target cancer. In some cases, the trained model can be configured to determine the presence or absence of the target cancer in low-stage (stage I or stage II) tumors. In some cases, the trained model can be configured to determine the immunotherapy response of the target if the target is offered immunotherapy. In some cases, the trained model can be configured to determine the category or tissue-specific location of the target cancer. In some cases, the trained model can be configured to determine one or more types of the target cancer.
[0075] In some examples, the genome database may be the human genome database. In some cases, the filtering step in step (b) may involve computer filtering of sequencing reads using a program such as bowtie2, Kraken, or a combination thereof. In some examples, the protein database may be the UniRef database. In some cases, the translating step in step (c) may be accomplished by a software package such as BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or a combination thereof. In some cases, the step in step (d) of mapping non-human proteins to biochemical pathways may be accomplished by mapping non-human proteins to a database such as KEGG, MetaCyc, PANTHER pathway, PathBank, or a combination thereof. In some cases, the biochemical pathways may be generated using the software package MiniPath.
[0076] In some examples, the disclosures provided herein describe a method for modifying a target cancer treatment using a trained predictive model. In some examples, the method may include: (a) preparing one or more sequencing reads of a target biological sample along with the cancer, cancer type, and treatments administered to treat the cancer; (b) filtering the sequencing reads using a genome database to create a set of filtered non-human sequencing reads; (c) translating the non-human sequencing reads into non-human proteins; (d) mapping the non-human proteins into a protein database to create a set of protein database relevances; and (e) modifying the target cancer treatment if the administered treatment differs from the treatment recommendation output by the trained predictive model when input using the set of protein database relevances. In some examples, the trained predictive model is trained on nucleic acid sequencing reads of a second set of one or more subjects' biological samples, corresponding cancer classifications, corresponding administered treatments, corresponding treatment responses, or any combination thereof. In some examples, the one or more subjects in the second set are different from the one or more subjects in the first set. In some examples, the protein database relevance set may include a set of functional genes, biochemical pathways, or any combination thereof, as described elsewhere in this specification. In some examples, the method may further include a step prior to step (c) of removing contaminant non-human sequencing reads by filtering the non-human sequencing reads. In some cases, contaminant non-human sequencing reads may be determined from a database of contaminant non-human sequencing reads determined beforehand or from experimental data analysis. In some cases, the translating step in step (c) may be completed in silico. In some examples, the method may include a step of sequencing the nucleic acid composition of interest instead of or in addition to step (a).In some cases, the method may further include a step of using a trained model to output a treatment for the target cancer, which the subject would respond with positive therapeutic efficacy if administered.
[0077] In some cases, the subject may be human or a non-human mammal. In some cases, the biological sample may include tissue, a fluid biopsy sample, or a combination thereof. In some cases, the biological sample may include a nucleic acid composition, which may include DNA, RNA, cell-free RNA, exosomal DNA, exosomal RNA, or a combination thereof. In some cases, the non-human sequence may originate from the biological origin of bacteria, archaea, fungi, viruses, or a combination thereof. In some examples, the fluid biopsy may include plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, breath condensate, or a combination thereof.
[0078] In some cases, the subjects may include cancer. In some cases, cancer may include acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof.
[0079] In some cases, treatment recommendations include therapies in which the subject is likely to respond with positive efficacy. In some cases, if the subject is receiving immunotherapy, treatment recommendations include the subject's immunotherapy response.
[0080] In some examples, the genome database may be the human genome database. In some cases, the filtering step in step (b) may involve computer filtering of sequencing reads using a program such as bowtie2, Kraken, or a combination thereof. In some examples, the protein database may be the UniRef database. In some cases, the translating step in step (c) may be accomplished by a software package such as BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or a combination thereof. In some cases, the step in step (d) of mapping non-human proteins to biochemical pathways may be accomplished by mapping non-human proteins to a database such as KEGG, MetaCyc, PANTHER pathway, PathBank, or a combination thereof. In some cases, the biochemical pathways may be generated using the software package MiniPath.
[0081] Computer system Figure 9 shows a computer system 901 suitable for implementing and / or training the models and / or predictive models described herein. The computer system 901 can process various aspects of the information disclosed herein, such as the target sequence of a biological sample. The computer system 901 may be an electronic device. The electronic device may be a mobile electronic device.
[0082] The computer system 901 may comprise a central processing unit (CPU, also referred to herein as "processor" and "computer processor") 905, which may be a single-core or multi-core processor, or multiple processors for parallel processing. The computer system 901 may further comprise memory or memory locations 904 (e.g., random-access memory, read-only memory, flash memory), electronic storage units 906 (e.g., hard disks), a communication interface 908 for communicating with one or more other devices (e.g., a network adapter), and peripheral devices 907 such as a cache, other memory, data storage, and / or an electronic display adapter. The memory 904, storage units 906, interface 908, and peripheral devices 907 communicate with the CPU 905 via a communication bus (solid line), such as a motherboard. The storage unit 906 may be a data storage unit (or data repository) for storing data. The computer system 901 may be operably coupled to a computer network ("network") 400 with the help of the communication interface 908. Network 400 may be the Internet, the Internet and / or an extranet, or an intranet and / or extranet communicating with the Internet. Network 400 may, in some cases, be a communications and / or data network. Network 400 may consist of one or more computer servers that can enable distributed computing, such as cloud computing. Network 400 may, in some cases, implement a peer-to-peer network with the help of computer system 901, so that devices coupled to computer system 901 can act as clients or servers.
[0083] The CPU 905 can execute a sequence of machine-readable instructions that can be embodied in a program or software. Instructions are directed to the CPU 905, which can then be programmed or otherwise configured to implement the methods of this disclosure. Examples of operations performed by the CPU 905 may include fetching, decoding, executing, and writing back.
[0084] CPU905 may be part of a circuit such as an integrated circuit. One or more other components of system901 may be included in the circuit. In some cases, the circuit is an application-specific integrated circuit (ASIC).
[0085] The storage unit 906 can store files such as drivers, libraries, and saved programs. The storage unit 906 can store one or more sequencing reads of one or more biological samples from one or more subjects, cancer types if present, procedures performed to treat the cancer, the efficacy of the procedures performed, or any combination thereof. The computer system 901 may include one or more additional data storage units located outside the computer system 901, such as being located on a remote server that communicates with the computer system 901 via an intranet or the internet.
[0086] The methods described herein can be implemented by machine-executable code stored in an electronic storage location of a computer device 901, for example, memory 904 or electronic storage unit 906. The machine-executable code or machine-readable code may be provided in the form of software. During use, the code may be executed by the processor 905. In some examples, the code is retrieved from the storage unit 906 and stored in memory 904 for easy access by the processor 905. In some examples, the electronic storage unit 906 may be omitted, and the machine-executable instructions are stored in memory 904.
[0087] The code may be pre-compiled and configured for use in a machine having a processor adapted to run the code, or it may be compiled at runtime. The code may be supplied in a programming language, which may be selected to enable the code to run pre-compiled or in a compiled state.
[0088] Embodiments of systems and methods provided herein, such as computer system 901, can be embodied in programming. Various embodiments of the art can typically be considered “products” or “manufactured articles” in the form of machine (or processor) executable code and / or associated data carried on or embodied on a type of machine-readable medium. Machine-executable code can be stored in electronic storage units, such as memory (e.g., read-only memory, random-access memory, flash memory) or hard disks. The “storage” type medium may include any or all of tangible memory such as that of a computer or processor, or associated modules such as various semiconductor memories, tape drives, disk drives, etc., which can provide non-temporary storage at any time for software programming. All or part of the software can sometimes be communicated over the Internet or various other telecommunication networks. Such communication can, for example, enable the loading of software from one computer or processor to another computer, for example, from a management server or host computer to an application server computer platform. Therefore, other types of media that can carry software elements include optical, electrical, and electromagnetic waves, such as those used across physical interfaces between local devices, via wired and optical landline networks, and via various aerial links. Physical elements that carry such waves, such as wired or wireless links, optical links, etc., can also be considered media that carry software. As used herein, unless limited to tangible “storage” media that are not temporary, terms such as computer or machine-readable media refer to any medium involved in providing instructions to a processor for execution.
[0089] Therefore, machine-readable media such as computer executable code can take many forms, but are not limited to tangible storage media, carrier media, or physical transmission media. Non-volatile storage media can include optical or magnetic disks, such as storage devices in any computer(s) that can be used to implement databases, etc. Volatile storage media include dynamic memory, such as the main memory of a computer platform. Tangible transmission media include coaxial cables, copper wires, and optical fibers, and wires that make up buses in computer devices. Carrier media can take the form of electrical or electromagnetic signals, or sound or light waves generated during radio frequency (RF) and infrared (IR) data communications. Common forms of computer-readable media for this purpose include, for example, floppy disks, flexible disks, hard disks, magnetic tapes, other magnetic media, CD-ROMs, DVDs or DVD-ROMs, other optical media, punched card paper tapes, other physical storage media having a pattern of holes, RAM, ROMs, PROMs and EPROMs, FLASH-EPROMs, any other memory chips or cartridges, carrier wave transport data or instructions, cables or links that transport such carrier waves, or any other media from which a computer can read programming code and / or data. Many of these forms of computer-readable media can be involved in transporting one or more sequences of one or more instructions to a processor for execution.
[0090] The computer system may include, or may communicate with, a user interface (UI) 903 for viewing treatment actions output by a trained predictive model, and / or an electronic display 902 containing recommendations or decisions regarding the presence or absence of cancer in one or more subjects. Examples of UIs include, but are not limited to, graphical user interfaces (GUIs) and web-based user interfaces.
[0091] The methods and systems disclosed herein can be implemented by one or more algorithms and by instructions provided together with one or more processors disclosed herein. The algorithms can be implemented in software when executed by the central processing unit 905. The algorithms may be, for example, random forests, graphical models, support vector machines, or others.
[0092] In some cases, the disclosures provided herein describe computer implementations for utilizing a trained predictive model to provide therapeutic treatment predictions for one or more subjects. In some examples, the method may include: (a) receiving nucleic acid sequencing reads and corresponding cancer classifications of a first set of one or more subjects; (b) filtering the nucleic acid sequencing reads using a genome database build to generate non-human sequencing reads; (c) translating the non-human sequencing reads into non-human proteins; (d) mapping the non-human proteins into a protein database to create a set of protein database relevances; and (e) if the set of protein database relevances is provided to a trained predictive model as input, utilizing the trained predictive model to provide treatment predictions for the first set of one or more subjects. In some cases, the method may further include, prior to (c), a step of removing contaminant non-human sequencing reads by removing contaminant non-human sequencing reads. In some examples, the translation step (c) can be completed in silico.
[0093] In some cases, the trained predictive model can be trained on nucleic acid sequencing reads of one or more subjects in a second set, corresponding cancer classifications, corresponding treatments administered, corresponding treatment responses, or any combination thereof. In some examples, one or more subjects in the second set may be different from one or more subjects in the first set. In some examples, the set of protein database relevances may include a set of functional genes, biochemical pathways, or any combination thereof. In some examples, the biological sample may include tissue, liquid biopsy sample, or any combination thereof. In some examples, the liquid biopsy may include plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, breath condensate, or any combination thereof. In some examples, one or more subjects in the first set may be human or non-human mammals. In some examples, the biological sample nucleic acid composition may include DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or any combination thereof. In some examples, the genome database may be a human genome database. In some cases, non-human sequences may originate from the biological origins of bacteria, archaea, fungi, viruses, or any combination thereof. In some instances, treatment prediction may include the immunotherapy response of one or more subjects in a first set, if immunotherapy is administered to one or more subjects in a first set. In some instances, treatment prediction may include the therapeutic efficacy that one or more subjects in a first set will respond with positive efficacy.In some cases, cancer classifications may include acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasms diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof.
[0094] In some cases, the filtering step in step (b) may involve computer filtering of sequencing reads using a program such as bowtie2, Kraken, or a combination thereof. In some cases, the protein database may be the UniRef database. In some examples, the translating step in step (c) can be accomplished using a software package such as BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or a combination thereof. In some cases, the mapping step in step (d) of non-human proteins to biochemical pathways can be accomplished by mapping non-human proteins to a database such as KEGG, MetaCyc, PANTHER pathway, PathBank, or a combination thereof. In some cases, the biochemical pathways can be generated using the software package MinPath.
[0095] The steps described above illustrate the system's method according to an example, but those skilled in the art will recognize many variations based on the teachings provided herein. The steps can be completed in different orders. Steps can be added or removed. Some steps can include substeps. Many steps can be repeated as many times as is beneficial to the platform.
[0096] definition Unless otherwise defined, all technical terms, notes, and other technical and scientific terms used herein are intended to have the same meaning as those commonly understood by those skilled in the art in which the claimed subject matter pertains. In some cases, terms that have a commonly understood meaning are defined herein for clarity and / or easy reference, and the inclusion of such definitions herein should not necessarily be construed as representing a substantial difference from those commonly understood in the art.
[0097] Throughout this application, various embodiments can be presented in range format. It should be understood that the range format is merely for convenience and conciseness and should not be interpreted as an inflexible limitation to the scope of this disclosure. Therefore, a range description should be considered to specifically disclose all possible subranges within that range, as well as individual numerical values. For example, a range description such as 1-6 should be considered to specifically disclose subranges such as 1-3, 1-4, 1-5, 2-4, 2-6, 3-6, etc., as well as individual numbers within the range, such as 1, 2, 3, 4, 5, and 6. This applies regardless of the width of the range.
[0098] When used in the specification and claims, unless the context explicitly indicates otherwise, the singular forms “a,” “an,” and “the” refer to multiple objects. For example, the term “a sample” refers to multiple samples, including mixtures thereof.
[0099] The terms “determination,” “measurement,” “evaluation,” “assessment,” “assay,” and “analysis” are often used interchangeably herein and refer to forms of measurement. The terms include the determination of whether an element is present or not (e.g., detection). These terms may include quantitative, qualitative, or both quantitative and qualitative determinations. Assessments may be relative or absolute. “Detecting presence” may, depending on the context, include determining the quantity of something that is present, in addition to determining whether it is present or not.
[0100] The terms “subject,” “individual,” or “patient” are often used interchangeably herein. “Subject” may be a biological entity containing expressed genetic material. Biological entities may be plants, animals, or microorganisms, including, for example, bacteria, viruses, fungi, and protozoa. A subject may be a tissue, cell, or offspring of a biological entity obtained in vivo or cultured in vitro. A subject may be a mammal. A mammal may be a human. A subject may be diagnosed or suspected of being at high risk for a disease. In some cases, a subject may not necessarily be diagnosed or suspected of being at high risk for a disease.
[0101] The term "in vivo" is used to describe events that occur within the body of a subject.
[0102] The term "ex vivo" is used to describe events that occur outside the body of the subject. Ex vivo assays are not performed on the subject; rather, they are performed on a sample separate from the subject. An example of an ex vivo assay performed on a sample is an "in vitro" assay.
[0103] The term "in vitro" is used to describe events that occur within a container for holding experimental reagents, resulting in their separation from the biological source from which the material is obtained. In vitro assays can include cell-based assays in which living or dead cells are employed. In vitro assays can also include cell-free assays in which intact cells are not employed.
[0104] As used herein, the term “approximately” of a number refers to plus or minus 10% of that number. The term “approximately” of a range refers to the range minus 10% of its lowest value plus 10% of its highest value.
[0105] The use of absolute words or sequences, such as "will," "will not," "shall," "shall not," "must," "must not," "firstly," "initially," "nextly," "followingly," "before," "after," "finally," and "finally," is disclosed herein but does not limit the scope of this embodiment as an example.
[0106] Any systems, methods, software, compositions, and platforms described herein are modular and not limited to a series of steps. Therefore, terms such as “first” and “second” do not necessarily imply priority, order of importance, or sequence of actions.
[0107] As used herein, the terms “treatment” or “to treat” are used in reference to pharmaceutical or other intervention regimens for obtaining beneficial or desired outcomes in a recipient. Beneficial or desired outcomes include, but are not limited to, therapeutic benefits and / or preventive benefits. Therapeutic benefits may refer to the eradication or improvement of symptoms, or the eradication or improvement of an underlying disease during treatment. A therapeutic benefit may also be achieved by eradicating or improving one or more physiological symptoms associated with an underlying disease so that improvement is observed in the subject, even though the subject may still be suffering from the underlying disease. Preventive effects include delaying, preventing, or eliminating the onset of a disease or condition; delaying or eliminating the onset of symptoms of a disease or condition; slowing, stopping, or reversing the progression of a disease or condition; or any combination thereof. For a preventive benefit, a subject at risk of developing a particular disease, or a subject reporting one or more physiological symptoms of a disease, may receive treatment, even if a diagnosis of the disease has not been made.
[0108] The section headings used herein are for organizational purposes only and should not be construed as limiting the subject matter described herein. [Examples]
[0109] Example 1 Generation and use of trained diagnostic models for genetic pathways in disease diagnosis and classification. A diagnostic model was developed and tested that classified subjects as healthy individuals, lung cancer patients, or lung disease patients based on the abundance of non-mammalian pathways. Cell-free DNA (cfDNA) sequencing libraries were obtained and further processed from 166 healthy individuals, 288 lung cancer patients, and 109 lung disease patients. A further breakdown of subcancer categories is shown in Figure 3. Next, the cfDNA sequencing samples were aligned with biochemical pathway classifications using both the Web of Life Toolkit App (Woltka) and the HUMAnN 3.0 (Humann) pipeline, as shown in Figures 4A-4B. Based on this initial analysis, Woltka classified the samples into more representative pathway distributions than the Human toolkit. From the Woltka classification pathway, the following gene ontology (GO) pathways were found to be the most important features of the machine learning-based classifier: GO:0055085: transmembrane transport; GO:0005975: carbohydrate metabolism processes; GO:0006412: translation; GO:0006313: translocation, DNA-mediated; GO:0006355: transcriptional regulation, DNA template; GO:0006260: DNA replication; GO:00006351: transcription, DNA template; and GO:0000160: phosphate signaling pathway. Other pathways identified as important for cancer vs. health differentiation, and other pathways identified as important for cancer vs. lung disease differentiation, can be seen in Figures 5A-5B. The microbial pathways identified via the WolTka pipeline in Figure 2B were used as input to train predictive models (e.g., 10x cross-validation random forests) enabling cancer vs. health and cancer vs. lung disease differentiation. The performance of each model, as represented by receiver operating area (AUC) analysis (Figures 6A-6B), can be compared to cancer vs. healthy and cancer vs. lung disease predictive models trained on microbial taxonomic abundances, as shown in Figures 6C-D. The predictive model trained on the importance of pathways classified by Woltka was found to be able to distinguish cancer vs. healthy individuals with an AUC of 0.756 and cancer vs. lung disease with an AUC of 0.705, which is comparable to the AUC of 0.818 for cancer vs. healthy individuals and 0.707 for cancer vs. lung disease in the predictive model trained on microbial taxonomy.
[0110] Example 2 We generated and utilized a diagnostic model trained on genetic pathways for determining cancer staging. We constructed and tested a diagnostic model configured to classify the stage of a target cancer based on the abundance of non-mammalian pathways against the abundance of lung disease pathways. Cell-free DNA (cfDNA) sequencing data were obtained from cancer patients of various stages, in addition to lung disease patients. The sequencing data consisted of 288 cancer patients of various known stages and 109 lung disease patients, as shown in Figure 7. Further decomposition of cancer types and the number of subcategories is also shown in Figure 7. As shown in Example 1, multiple Woltka classification pathways for cf-mbDNA sequences were determined and used to train random forests with 10x cross-validation. The accuracy of each trained random forest predictive model was then analyzed by the area under the receiver operating characteristic curve (AUC), as shown in Figures 8A–8D. A predictive model trained on the importance of pathways classified by Woltka was able to distinguish between stage 1 cancer and lung disease, with an AUC of 0.868 for stage 2 cancer vs. lung disease, 0.582 for stage 3 cancer vs. lung disease, 0.793 for stage 4 cancer vs. lung disease, and 0.906 for stage 4 cancer vs. lung disease.
[0111] Embodiment 1. A method for determining the presence or absence of a target cancer, (a) A step of preparing one or more sequencing reads of the target biological sample, (b) A step of filtering sequencing reads using a genome database to create a set of filtered non-human sequencing reads, (c) A step of translating non-human sequencing reads into non-human proteins, (d) A step of mapping non-human proteins to a protein database, thereby creating a set of protein database relevances, (e) If the trained model is provided with inputs of a set of protein database relevances, the step of determining the presence or absence of the target cancer as an output to the trained model. A method that includes this. 2. The method according to Embodiment 1, wherein the set of protein database relevances includes a set of functional genes, biochemical pathways, or a combination thereof. The method according to Embodiment 1, further comprising the step of removing contaminants from the filtered non-human sequencing reads prior to 3.(c). 4. The method according to Embodiment 1, wherein the translation step is completed in silico. 5. The method according to Embodiment 1, wherein the biological sample is tissue, a fluid biopsy, or a combination thereof. 6. The method according to Embodiment 1, wherein the subject is human or a non-human mammal. 7. The method according to Embodiment 1, wherein the biological sample comprises a nucleic acid composition, and the nucleic acid composition comprises DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or any combination thereof. 8. The method according to Embodiment 1, wherein the genome database is the human genome database. 9. The method according to Embodiment 1, wherein the trained model is trained using a set of functional gene and biochemical pathway abundances that are present or absent, along with characteristic abundances related to the cancer of interest. 10. The method according to Embodiment 1, wherein the non-human sequence originates from the biological origin of bacteria, archaea, fungi, viruses, or any combination thereof. 11. The method according to Embodiment 1, wherein a trained model is configured to determine the category or tissue-specific location of the cancer in question. 12. The method according to Embodiment 1, wherein a trained model is configured to determine one or more types of cancer of interest. 13. The method according to Embodiment 12, wherein a trained model is configured to determine one or more subtypes of the target cancer. 14. The method according to Embodiment 1, wherein a trained model is configured to determine the stage of the cancer in question, the prognosis of the cancer in question, or any combination thereof. 15. The method according to Embodiment 1, wherein a trained model is configured to determine the presence or absence of cancer in low-stage (stage I or stage II) tumors. 16. The method according to Embodiment 1, wherein a trained model is configured to determine the immunotherapy response of a subject when the subject is offered immunotherapy. 17. The method according to Embodiment 1, further comprising the step of using a trained model to output a treatment for a subject to treat a cancer of the subject, wherein the subject would respond with positive therapeutic efficacy if the treatment were administered. 18. The method according to Embodiment 1, wherein the target cancer includes acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe cells, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof. 19. The method according to Embodiment 5, wherein the liquid biopsy includes plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled condensate, or any combination thereof. 20. The method according to Embodiment 1, wherein the filtering step includes filtering the sequencing reads by a computer using a program such as bowtie2, Kraken, or a combination thereof. 21. The method according to Embodiment 1, wherein the protein database is the UniRef database. 22. The method according to Embodiment 1, wherein the translation step is achieved by a software package of BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or any combination thereof. 23. The method according to Embodiment 2, wherein the step of mapping non-human proteins to biochemical pathways is achieved by mapping non-human proteins to databases such as KEGG, MetaCyc, PANTHER pathway, PathBank, or any combination thereof. 24. The method according to Embodiment 2, wherein the biochemical pathway is generated using the software package MinPath. 25. A method for providing a determination of the presence or absence of a target cancer, (a) A step of sequencing the nucleic acid composition of the target biological sample and thereby generating sequencing reads, (b) A step of filtering sequencing reads using a genome database to create a set of filtered non-human sequencing reads, (c) A step of translating non-human sequencing reads into non-human proteins, (d) A step of mapping non-human proteins to a protein database, thereby creating a set of protein database relevances, (e) If the trained model is provided with inputs of set protein database relevance, the step of providing a determination of the presence or absence of the target cancer as an output of the trained model. A method that includes this. 26. The method according to Embodiment 25, wherein the set of protein database relevances includes a set of functional genes, biochemical pathways, or a combination thereof. The method according to Embodiment 25, further comprising the step of removing contaminants from the filtered non-human sequencing reads prior to 27.(c). 28. The method according to embodiment 25, wherein the translation step is completed in silico. 29. The method according to Embodiment 25, wherein the biological sample is a tissue, a liquid biopsy sample, or a combination thereof. 30. The method according to Embodiment 25, wherein the subject is human or a non-human mammal. 31. The method according to Embodiment 25, wherein the biological sample comprises a nucleic acid composition, and the nucleic acid composition comprises DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or any combination thereof. 32. The method according to Embodiment 25, wherein the genome database is the human genome database. 33. The method according to Embodiment 25, wherein the trained model is trained using a set of functional gene and biochemical pathway abundances that are present or absent, along with characteristic abundances related to the cancer of interest. 34. The method of Embodiment 25, wherein the non-human sequence originates from a biological origin of bacteria, archaea, fungi, viruses, or any combination thereof. 35. The method according to Embodiment 25, wherein a trained model is configured to determine the category or tissue-specific location of the cancer in question. 36. The method according to Embodiment 25, wherein a trained model is configured to determine one or more types of cancer of interest. 37. The method according to embodiment 36, wherein a trained model is configured to determine one or more subtypes of the cancer in question. 38. The method according to Embodiment 25, wherein a trained model is configured to determine the stage of the cancer in question, the prognosis of the cancer in question, or any combination thereof. 39. The method according to Embodiment 25, wherein a trained model is configured to determine the presence or absence of cancer in low-stage (stage I or stage II) tumors. 40. The method according to Embodiment 25, wherein a trained model is configured to determine the immunotherapy response of a subject when the subject is offered immunotherapy. 41. The method according to Embodiment 25, further comprising the step of using a trained model to output a treatment for a subject to treat a cancer of the subject, wherein the subject would respond with positive therapeutic efficacy if the treatment were administered. 42. The method according to Embodiment 25, wherein the target cancer includes acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof. 43. The method according to Embodiment 29, wherein the liquid biopsy includes plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled condensate, or any combination thereof. 44. The method according to Embodiment 25, wherein the filtering step includes filtering the sequencing reads by a computer using a program such as bowtie2, Kraken, or a combination thereof. 45. The method according to Embodiment 25, wherein the protein database is the UniRef database. 46. The method according to Embodiment 25, wherein the translation step is achieved by a software package of BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or any combination thereof. 47. The method according to Embodiment 26, wherein the step of mapping non-human proteins to biochemical pathways is achieved by mapping non-human proteins to databases such as KEGG, MetaCyc, PANTHER pathway, PathBank, or any combination thereof. 48. The method according to Embodiment 26, wherein the biochemical pathway is generated using the software package MinPath. 49. A method for training a model configured to determine the presence or absence of a target cancer, (a) A step of preparing a dataset containing nucleic acid sequencing reads of nucleic acid compositions of one or more subjects from the first set, and corresponding one or more cancers of one or more subjects from the first set, (b) A step of filtering nucleic acid sequencing reads using a genome database build to generate non-human sequencing reads, (c) A step of translating non-human sequencing reads into non-human proteins, (d) A step of mapping non-human proteins to a protein database, thereby creating a set of protein database relevances, (e) A step of training a model using a set of protein database associations and the corresponding one or more cancer states of one or more subjects from the first set, thereby generating a trained model configured to determine the presence or absence of cancer in one or more subjects from the second set. A method that includes this. 50. The method according to Embodiment 49, wherein the set of protein database relevances includes a set of functional genes, biochemical pathways, or a combination thereof. The method according to Embodiment 49, further comprising the step of removing contaminants from the filtered non-human sequencing reads prior to 51.(c). 52. The method according to embodiment 49, wherein the translation step is completed in silico. 53. The method according to Embodiment 49, wherein the biological sample is a tissue, a liquid biopsy sample, or a combination thereof. 54. The method according to Embodiment 49, wherein one or more subjects in the first set, the second set, or any combination thereof are human or non-human mammals. 55. The method according to Embodiment 49, wherein the biological sample comprises a nucleic acid composition, and the nucleic acid composition comprises DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or any combination thereof. 56. The method according to Embodiment 49, wherein the genome database is the human genome database. 57. The method according to Embodiment 49, wherein the trained model is trained using a set of functional gene and biochemical pathway abundances that are present or absent, along with characteristic abundances for the cancer of interest. 58. The method according to Embodiment 49, wherein the non-human sequence originates from a biological origin of bacteria, archaea, fungi, viruses, or any combination thereof. 59. The method according to embodiment 49, wherein a trained model is configured to determine the category or tissue-specific location of cancer in one or more subjects in a second set. 60. The method according to embodiment 49, wherein a trained model is configured to determine one or more types of cancer in one or more subjects of a second set. 61. The method according to embodiment 60, wherein a trained model is configured to determine one or more subtypes of cancer in one or more subjects of a second set. 62. The method according to Embodiment 49, wherein a trained model is configured to determine the cancer stage, cancer prognosis, or any combination thereof for one or more subjects in a second set. 63. The method according to Embodiment 49, wherein a trained model is configured to determine the presence or absence of cancer in a second set of one or more subjects in low-stage (stage I or stage II) tumors. 64. The method according to Embodiment 49, wherein a trained model is configured to determine the immunotherapy response of a subject when the subject is offered immunotherapy. 65. The method according to Embodiment 49, further comprising the step of using a trained model to output a treatment for cancer in one or more subjects of a second set, wherein one or more subjects of the second set would respond with positive therapeutic efficacy if administered the treatment. 66. The method according to Embodiment 49, wherein the cancer of one or more subjects in the first and second sets includes acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof. 67. The method according to Embodiment 53, wherein the liquid biopsy includes plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled condensate, or any combination thereof. 68. The method according to Embodiment 49, wherein the filtering step includes filtering the sequencing reads by a computer using a program such as bowtie2, Kraken, or a combination thereof. 69. The method according to Embodiment 49, wherein the protein database is the UniRef database. 70. The method according to Embodiment 49, wherein the translation step is achieved by a software package of BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or any combination thereof. 71. The method according to Embodiment 50, wherein the step of mapping non-human proteins to biochemical pathways is achieved by mapping non-human proteins to databases such as KEGG, MetaCyc, PANTHER pathway, PathBank, or any combination thereof. 72. The method according to Embodiment 50, wherein the biochemical pathway is generated using the software package MinPath. 73. The method according to Embodiment 51, wherein the dataset further includes corresponding prior or current treatments performed on one or more subjects of the first set. 74. The method according to Embodiment 73, wherein the dataset further comprises the treatment efficacy of previous or current treatment administrations to one or more subjects of the first set. 75.1 A computer implementation method for utilizing a trained predictive model to provide predictions of therapeutic actions for one or more subjects, (a) A step of receiving nucleic acid sequencing reads of one or more subject biological samples from a first set, and the corresponding cancer classification, (b) A step of filtering nucleic acid sequencing reads using a genome database build to generate non-human sequencing reads, (c) A step of translating non-human sequencing reads into non-human proteins, (d) A step of mapping non-human proteins to a protein database, thereby creating a set of protein database relevances, (e) If a set of protein database relevances is provided to a trained predictive model as input, the step of using the trained predictive model to provide treatment predictions for one or more subjects in the first set. A method that includes this. 76. The method according to Embodiment 75, wherein the trained predictive model is trained on nucleic acid sequencing reads of one or more subject biological samples from a second set, corresponding cancer classifications, corresponding treatments administered, corresponding treatment responses, or any combination thereof. 77. The method according to Embodiment 76, wherein one or more subjects in the second set are different from one or more subjects in the first set. 78. The method according to Embodiment 75, wherein the set of protein database relevances includes a set of functional genes, biochemical pathways, or a combination thereof. The method of Embodiment 75, further comprising the step of removing contaminants from the filtered non-human sequencing reads prior to 79.(c). 80. The method according to embodiment 75, wherein the translation step is completed in silico. 81. The method according to Embodiment 75, wherein the biological sample is tissue, a liquid biopsy sample, or a combination thereof. 82. The method according to Embodiment 75, wherein one or more subjects in the first set are human or non-human mammals. 83. The method according to Embodiment 75, wherein the biological sample nucleic acid composition comprises DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or any combination thereof. 84. The method according to Embodiment 75, wherein the genome database is the human genome database. 85. The method of Embodiment 75, wherein the non-human sequence originates from the biological origin of bacteria, archaea, fungi, viruses, or any combination thereof. 86. The method according to Embodiment 75, wherein, when immunotherapy is administered to one or more subjects in the first set, the treatment prediction includes an immunotherapy response in one or more subjects in the first set. 87. The method of Embodiment 75, wherein the treatment prediction includes therapeutic efficacy, in which one or more subjects in a first set will respond with positive efficacy. 88. The method according to Embodiment 75, wherein the cancer classification includes acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof. 89. The method according to Embodiment 79, wherein the liquid biopsy includes plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled condensate, or any combination thereof. 90. The method according to Embodiment 75, wherein the filtering step includes filtering the sequencing reads by a computer using a program such as bowtie2, Kraken, or any combination thereof. 91. The method according to Embodiment 75, wherein the protein database is the UniRef database. 92. The method according to Embodiment 75, wherein the translation step is achieved by a software package of BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or any combination thereof. 93. The method according to Embodiment 76, wherein the step of mapping non-human proteins to biochemical pathways is achieved by mapping non-human proteins to databases such as KEGG, MetaCyc, PANTHER pathway, PathBank, or any combination thereof. 94. The method according to Embodiment 76, wherein the biochemical pathway is generated using the software package MinPath. 95. A method for modifying a target cancer treatment using a trained predictive model, (a) A step of preparing one or more sequencing reads of the target biological sample along with the cancer, cancer type, and the treatment performed to treat the cancer, (b) A step of filtering sequencing reads using a genome database to create a set of filtered non-human sequencing reads, (c) A step of translating non-human sequencing reads into non-human proteins, (d) A step of mapping non-human proteins to a protein database, thereby creating a set of protein database relevances, (e) If the treatment performed differs from the treatment recommendation output by the trained predictive model when input using a set of protein database relevances, the step of changing the cancer treatment in question. A method that includes this. 96. The method according to Embodiment 95, wherein the trained predictive model is trained on nucleic acid sequencing reads of one or more subject biological samples from a second set, corresponding cancer classifications, corresponding treatments administered, corresponding treatment responses, or any combination thereof. 97. The method according to Embodiment 96, wherein one or more subjects in the second set are different from one or more subjects in the first set. 98. The method according to Embodiment 95, wherein the set of protein database relevances includes a set of functional genes, biochemical pathways, or a combination thereof. The method of Embodiment 95, further comprising the step of removing contaminants from the filtered non-human sequencing reads prior to 99.(c). 100. The method according to Embodiment 95, wherein the translation step is completed in silico. 101. The method according to Embodiment 95, wherein the biological sample is tissue, a liquid biopsy sample, or a combination thereof. 102. The method according to Embodiment 95, wherein the subject is human or a non-human mammal. 103. The method according to Embodiment 95, wherein the biological sample nucleic acid composition comprises DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or any combination thereof. 104. The method according to Embodiment 95, wherein the genome database is the human genome database. 105. The method according to Embodiment 95, wherein the non-human sequence originates from a biological origin of bacteria, archaea, fungi, viruses, or any combination thereof. 106. The method according to Embodiment 95, wherein, if the subject is administered immunotherapy, the treatment recommendation includes the subject's immunotherapy response. 107. The method of Embodiment 95, wherein the treatment recommendation includes a treatment that would result in the subject responding with positive efficacy. 108. The method according to Embodiment 95, wherein the target cancer includes acute myeloid leukemia, adrenocortical carcinoma, urothelial carcinoma of the bladder, low-grade glioma of the brain, invasive breast cancer, squamous cell carcinoma of the cervix and intracervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal cancer, glioblastoma multiforme, squamous cell carcinoma of the head and neck, renal chromophobe, clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, cutaneous melanoma of the skin, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid cancer, uterine carcinosarcoma, endometrial carcinoma of the uterine body, uveal melanoma, or any combination thereof. 109. The method according to Embodiment 101, wherein the liquid biopsy includes plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled condensate, or any combination thereof. 110. The method according to Embodiment 95, wherein the filtering step includes filtering the sequencing reads by a computer using a program such as bowtie2, Kraken, or any combination thereof. 111. The method according to Embodiment 95, wherein the protein database is the UniRef database. 112. The method according to Embodiment 95, wherein the translation step is achieved by a software package of BLASTP, USEARCH, LAST, MMSeqs2, DIAMOND, or any combination thereof. 113. The method according to Embodiment 96, wherein the step of mapping non-human proteins to biochemical pathways is achieved by mapping non-human proteins to databases such as KEGG, MetaCyc, PANTHER pathway, PathBank, or any combination thereof. 114. The method according to Embodiment 96, wherein the biochemical pathway is generated using the software package MinPath.
Claims
1. 1. A method for determining the presence of cancer in a subject, comprising: (a) providing one or more sequencing reads for a biological sample of the subject; (b) filtering the sequencing reads using a genome database to generate a set of filtered non-human sequencing reads; (c) decontaminating the filtered non-human sequencing reads to remove contaminant non-human sequencing reads from the filtered non-human sequencing reads; (d) translating the decontaminated non-human sequencing reads into non-human proteins; (e) mapping the non-human proteins to a protein database, thereby generating a set of protein database relationships; and (f) determining the presence of cancer in said subject by processing the set of protein database associations with the trained model; The method includes:
2. The method of claim 1, wherein the set of protein database associations includes a set of functional genes, biochemical pathways, or a combination thereof.
3. The method of claim 1, wherein the translating step is completed in silico.
4. The method of claim 1, wherein the biological sample is a tissue, a liquid biopsy, or a combination thereof.
5. The method of claim 4, wherein the liquid biopsy comprises plasma, serum, whole blood, urine, cerebrospinal fluid, saliva, sweat, tears, exhaled breath condensate, or a combination thereof.
6. The method of claim 1, wherein the subject is a human or a non-human mammal.
7. The method of claim 1, wherein the biological sample comprises a nucleic acid composition, the nucleic acid composition comprising DNA, RNA, cell-free DNA, cell-free RNA, exosomal DNA, exosomal RNA, or a combination thereof.
8. The method of claim 1, wherein the genome database is a human genome database.
9. The method described in claim 1, wherein the trained model is trained using a set of functional gene and biochemical pathway abundances that are present or absent along with characteristic abundances for the cancer of interest.
10. The method of claim 1, wherein the non-human sequence originates from a bacterial, archaeal, fungal, viral, or combination thereof origin of life.
11. The trained model, (i) the cancer category or tissue-specific location of the subject; (ii) one or more types of cancer in the subject; (iii) one or more subtypes of cancer in the subject; (iv) the stage of the cancer in the subject, the prognosis of the cancer in the subject, or a combination thereof. (v) the presence of cancer in a low-stage (stage I or stage II) tumor; or (vi) the subject's immunotherapy response when the subject is provided with immunotherapy; The method of claim 1 , further configured to determine:
12. The method of claim 1, further comprising using the trained model to output a therapeutic approach for a subject to treat the subject's cancer, wherein the subject, when administered the therapeutic approach, responds with positive therapeutic efficacy.
13. The method of claim 1, wherein the target cancer comprises acute myeloid leukemia, adrenal cortical carcinoma, bladder urothelial carcinoma, brain low-grade glioma, invasive breast cancer, cervical squamous cell carcinoma and endocervical adenocarcinoma, cholangiocarcinoma, colon adenocarcinoma, esophageal carcinoma, glioblastoma multiforme, head and neck squamous cell carcinoma, renal chromophobe, renal clear cell carcinoma of the kidney, papillary renal cell carcinoma of the kidney, hepatocellular carcinoma of the liver, lung adenocarcinoma, lung squamous cell carcinoma, lymphoid neoplasm diffuse large B-cell lymphoma, mesothelioma, ovarian serous cystadenocarcinoma, pancreatic adenocarcinoma, pheochromocytoma and paraganglioma, prostate adenocarcinoma, rectal adenocarcinoma, sarcoma, skin cutaneous melanoma, gastric adenocarcinoma, testicular germ cell tumor, thymoma, thyroid carcinoma, uterine carcinosarcoma, uterine endometrial carcinoma, uveal melanoma, or any combination thereof.
14. 2. The method of claim 1, wherein the filtering step comprises computationally filtering the sequencing reads by bowtie2, Kraken, or a combination thereof.
15. The method of any one of claims 1 to 14, wherein the trained model comprises one or more machine learning algorithms.
