Pharmacogenomic decision support for modulators of NMDA, glycine and AMPA receptors

Abstract

The present invention relates to pharmacogenomic decision support for modulators of NMDA, glycine and AMPA receptors. Used for identifying diagnosed as having therapeutically resistant or refractory depression, pain or other clinical indications and suitable for accepting N-methyl-d-aspartate receptor antagonists, glycine receptor beta (GLRB) modulators or alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPAR) Methods for patients of therapy include determining an appropriate drug, an optimal dose per patient, and determining which patients are not suitable for receiving the therapy. Pharmacogenomic clinical decisions support the determination of a combination of targeted single nucleotide polymorphisms and clinical values, or targeted single nucleotide polymorphisms, targeted ketamine-specific expansion and contraction of topologically associated domains, and clinical values. The methods described herein allow for more efficient determination of which patients will experience drug efficacy and which patients will experience adverse drug events. The method provides personalized patient dose suggestions, drug administration frequency, and drug selection suggestions.

Description

[0001] This application is a divisional application of the invention application with the application date of January 22, 2020, the Chinese national application number of 202080015844.7, and the invention name of "Pharmacogenomic Decision Support for Modulators of NMDA, Glycine, and AMPA Receptors". Technical Field

[0002] The techniques described herein relate to pharmacogenomic clinical decision support assays and can be used to select therapies for depression (especially treatment-resistant or refractory depression) and other clinical indications based on N-methyl-D-aspartic acid (NMDA) receptors, glycine receptors, and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors, including anesthesia and analgesia / pain conditions, neuropsychiatric conditions, and neurological conditions exemplified by ketamine and its enantiomers. More specifically, the techniques relate to specific biomarkers, including genetic markers, clinical values, and disease phenotypes derived from patients, to optimize the selection of pharmaceuticals that affect these receptor networks and individual patient dosing. Background Art

[0003] Existing antidepressant medications are ineffective for many patients. A new class of antidepressant drugs is being developed that target glutamate receptors in the human forebrain. Ketamine (RS-2-chlorophenyl-2-methylamino-cyclohexanone) is a non-competitive antagonist of the glutamate N-methyl-D-aspartic acid receptor (NMDAR) and has been approved by the US Food and Drug Administration (FDA) as an anesthetic and has shown promise as an antidepressant in patients with treatment-resistant depression (TRD). Although the racemic formulation may have potent and undesirable psychotomimetic and other side effects depending on several variables, chemical analogs of ketamine exhibit reduced adverse events. Intravenous and oral formulations have demonstrated efficacy and tolerability in controlled trials and open-label studies in patient populations that typically show little response to traditional antidepressants known to target the serotonin transporter (SLC6A4, also known as 5HTT or SERT1), including serotonin-norepinephrine reuptake inhibitors (SNRIs). There is evidence that ketamine, its enantiomers, and ketamine analogs exert their mechanism of action primarily by modulating NMDA receptors (NMDARs) and downstream receptors in this network in the human brain.

[0004] The pharmacodynamic (PD) target of ketamine drugs is the NMDAR composed of GRIN1 and GRIN2 subunits, which binds glutamate and N-methyl-D-aspartic acid, the binding sites of glycine and D-serine encoded by GLRB, and the sites that bind polyamines, histamine, and cations. Antagonists, partial antagonists, and receptor modulators, such as ketamine and its combinations with other NMDAR and glycine modulators, including phencyclidine, amantadine, dextromethorphan, tiletamine, riluzole, methoxetamine, methoxphenidine, and memantine, block inward Ca +2 influx, thus preventing postsynaptic depolarization. Neuroimaging studies have demonstrated that intravenous infusion of ketamine results in a transient surge in glutamate levels observed in the prefrontal cortex, along with rapid antidepressant effects. It has been shown that after NMDAR blockade, glutamate preferentially binds to the GRIA1, GRIA2, and GRIA4 subunits of the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor. Several NMDAR antagonists and partial antagonists, GLRB modulators, and AMPAR agonists are being developed for the treatment of refractory depression but exhibit dissociative effects in patients. Additionally, NMDAR antagonists, GLRB antagonists, and AMPAR modulators, partial antagonists, and receptor modulators that act in the same network as such drugs due to safety concerns and failures in clinical trials may repurpose using the methods of the present disclosure to increase their chances of success in clinical trials.

[0005] The three-dimensional (3D) architecture of the human regulatory epigenome plays an important role in determining the human phenotype. It is now well known that most of the important single nucleotide polymorphisms (SNPs) associated with disease risk, drug response, and other human traits discovered in genome-wide association studies (GWAS) are located in enhancers, promoters, and other non-coding regulatory elements. Coupled with the recent in-depth understanding of functional 4D nuclear body organization, a large amount of biomedical "big data" has become available from open-source and proprietary resources, enabling the reconstruction of drug regulatory pathways in the human genome. New methods are being applied to mine regulatory variants from genome-wide association studies (GWAS), phenome-wide association studies (PheWAS), and the mining of electronic health record data and clinical trial data. It is now standard practice in the art to re-evaluate single nucleotide polymorphism (SNP)-trait associations from these data sources in the context of pathway analysis, and the SNP-trait associations have been found to be important because they map to the same or related biological networks. Combining large amounts of biomedical data with innovative methods for mining biological networks lays the foundation for detecting gene variants that may affect drug response variability, including adverse drug events. This approach holds promise for making great progress in specialized fields such as psychiatry, where lack of drug efficacy and a high incidence of adverse drug events have proven to be particularly problematic in patient care.

[0006] More than half of Americans will exhibit symptoms of a mental disorder during their lifetime. The most common lifetime mental disorders are stress and anxiety disorders, mood disorders, including major depressive disorder and bipolar disorder, impulse control disorders, substance use disorders, and schizoaffective disorder. The lifetime prevalence of any mental disorder in the United States is 53%, while 28% have two or more lifetime disorders, and 18% have three or more, indicating that the comorbidity of mental disorders is a significant medical challenge. While some mental disorders, such as bipolar disorder type 1, are inherited in families with an approximate 80% penetrance, other mental disorders do not exhibit obvious heritability. For major depressive disorder (MDD), as indicated by family, twin, and adoption studies, genetic factors play an important role in the etiology of the disease. Twin studies suggest a heritability of 50%, and family studies indicate a two- to three-fold increased lifetime risk of developing MDD in first-degree relatives. In studies of control groups, several sociodemographic variables are significantly associated with the lifetime risk of mental disorders. For example, women (biological sex) have a significantly higher risk of anxiety and major depressive disorder than men, while men (biological sex) have a significantly higher risk of impulse control and substance abuse disorders than women. Compared to non-Hispanic whites, non-Hispanic blacks and Hispanics have a significantly lower risk of anxiety, mood, and substance abuse disorders, and low educational attainment is associated with a higher risk of substance abuse disorders. Data show that many factors, including environmental and social factors, biological sex, race, and family inheritance, are related to the etiology of mental illness. Additionally, the complexity of phenotypes within individual patients or patient groups, including comorbidity with other mental disorders and stress-related diseases, requires a range of different algorithmic classification solutions, including machine learning, and multiple statistical analyses, including linear regression, to accurately specify precise therapies beyond the current available scope.

[0007] Mental illnesses have a greater impact on human health than any other disease. For example, major depressive disorder (MDD) causes a greater disability burden worldwide than any other medical condition (including cancer, heart disease, stroke, chronic obstructive pulmonary disease, and HIV / AIDS), yet MDD remains one of the most undiagnosed, misdiagnosed, and untreated or poorly treated diseases known to humanity. In 2013, the National Institutes of Health (NIH) provided 13 times more funding for oncology research—approximately $5.3 billion—than for depression research, which was $415 million. In the United States, from 2009 to 2011, adverse events related to prescription antidepressants resulted in more than 25,000 emergency department visits per year, accounting for 30% of all prescription drug-related hospitalizations annually. Patients with MDD and comorbid medical conditions experience more severe depressive symptoms and lower response and remission rates to antidepressant treatment compared to patients without comorbid symptoms. Treatment-resistant depression (TRD) accounts for 30%–40% of all patients diagnosed with MDD and is defined as “failure to achieve remission after two adequate antidepressant courses of known evidence-based acceptable doses and durations.”

[0008] Modern antidepressant medications are ineffective for many patients, and in patients who do respond or remit, weeks to months of drug therapy are required before symptom remission is achieved. Therefore, newer and more effective antidepressant medications are being developed. For example, both the racemic mixture of ketamine and the S-enantiomer of ketamine are examples of N-methyl-D-aspartic acid receptor (NMDAR) partial antagonists that have been approved by the Food and Drug Administration (FDA) for the treatment of TRD. As measured by the total score on the Montgomery- Depression Rating Scale (MADRS), ketamine induces rapid antidepressant responses and an accompanying increase in cortical glutamate levels in approximately 50% of patients with TRD. Although R,S-ketamine has been used in clinical indications such as chronic pain, perioperative analgesia, and sedation since 1970, adverse drug events (AEs) are common after ketamine treatment, and diversion is restricted by limiting use to inpatient and outpatient treatment settings. For example, in a phase III clinical trial of esketamine for TRD before submission to the FDA, almost one-quarter of patients with TRD experienced severe dissociation, two deaths were reported, and an additional 6.9% of patients with TRD in the treatment group experienced severe psychotomimetic effects, including delirium, delusions, and suicidal ideation, as well as suicide attempts.

[0009] One of the challenges in psychiatry is the precise matching of drug therapies to accurately address the complex symptoms of individual patients. Psychiatric patients present a wide range of comorbid disorders and there are few objective biomarkers that can be used as diagnostic criteria to accurately tailor antidepressant, antipsychotic, and antimanic therapies to patients. Although diagnostic rating scales such as the Hamilton Scale for Depression (HAM-D) demonstrate good inter-rater reliability, mental disorders such as depression present as a variety of different phenotypes. Non-pharmacological therapies can demonstrate improved efficacy in patients with treatment-resistant depression (TRD) or recurrent depression. For example, repetitive transcranial magnetic stimulation (rTMS) shows promise as a non-pharmacological treatment alternative for patients with TRD; however, the best outcomes in TRD occur when rTMS is used as an adjunctive treatment to traditional antidepressant drug therapy, as is the case with antidepressant drug classes that include N-methyl-D-aspartic acid (NMDA) receptor antagonists or partial antagonists, glycine receptor beta (GLRB) modulators, or alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor agonists, which can only be offered in the clinical setting of patients who are already taking another antidepressant drug.

[0010] rTMS therapy requires dozens of clinical visits, remission varies widely in patients with TRD, and rTMS is effective in only about 20%-40% of cases, with remission of depression lasting up to 1-2 years. The latest results of rTMS combined with neuroimaging demonstrate that, based on differential rTMS array placement, patients with depression specifically cluster into four different phenotypes along the axes of anhedonia and anxiety. These results provide substantial evidence that psychiatric patients can be stratified by phenotype based on the activation of different brain connectivity pathways, networks that exhibit considerable inter-individual differences between patients, and greatly improve the chances of precisely matching the best therapy to individual patients.

[0011] Although rTMS offers hope for patients with TRD, its mechanism of action remains elusive until independent studies demonstrated that TMS first acts on the subgenual anterior cingulate cortex and significantly increases glutamate levels as well as biomarker modulation of the N-methyl-D-aspartic acid receptor (NMDAR).

[0012] These are remarkable findings as they demonstrate that the mechanism of rTMS brain activation is almost indistinguishable from the mechanism of action of ketamine's pharmacological treatment. Thus, rTMS and ketamine exhibit similar mechanisms of action in alleviating TRD, but approximately half of all TRD patients do not achieve remission after treatment with either therapeutic option. Additionally, both rTMS and ketamine exhibit transient but severe AEs, including dissociation (the putative basis for ketamine's analgesic effect), psychotomimetic effects, and neurocognitive impairments. This indicates that matching individual patients to one of these therapies or other antidepressant medications is crucial if one can select which patients will benefit from these treatments and which patients will unnecessarily suffer severe AEs without sufficient antidepressant efficacy.

[0013] Recent studies combining transcranial magnetic stimulation (TMS) for depression alleviation followed by neuroimaging have demonstrated that patients with TRD can be clearly stratified into 4 subtypes based on their response to TMS device placement, where 4 different intrinsic neuroanatomical pathways are activated, along with distinct symptom clusters. As demonstrated in the present disclosure, these 4 subtypes can now be independently determined using a combination of clinical and molecular data, thus providing a paradigm for other psychiatric and stress-related disorders where improved pharmacophenomics decision support will deliver better treatment options to patients. Similarly, NMDAR antagonist therapies can be used as adjuvant treatments for age-related degenerative medical conditions. SUMMARY OF THE INVENTION

[0014] Different methods can be used to accurately determine the precise treatment requirements of an individual patient phenotype or phenotype group. The present disclosure describes methods for configuring a pharmacophenomics assay for clinical decision support or for companion diagnostics for psychotropic medications, which optimize the fit of treatment interventions to individual patients or groups of patients diagnosed with a psychiatric disorder or related disorder, such as treatment-resistant depression, chronic pain, migraine, fibromyalgia, inflammatory disorders, and other conditions for which ketamine or an analogue thereof includes an effective therapeutic agent. In the context of the present disclosure, the drug response and adverse event phenotypes of a patient comprise multiple sets of variables as described herein, ranging from the intrinsic configuration of the patient's pharmaco-genomic network, including its mutational profile configuration, to the behavioral phenotypes that can be obtained from clinical data.

[0015] The methods for patient stratification in the present disclosure use different data sources, some of which may be incomplete and require data cleaning and / or governance, or may not exist. The scope of different methods as described herein ranges from methods that can accommodate different combinations of limited data to more extensive computational solutions, or methods that can bridge missing data elements using probabilistic methods.

[0016] The present disclosure includes a series of serial and distinct methods to provide accurate pharmacophenomics decision support for diagnosing patients with mental disorders. The output provides a quantitative score for ranking treatment interventions, including recommendations such as drug selection and dosage, transcranial magnetic stimulation, electroconvulsive therapy, and behavioral interventions. The present disclosure includes pharmacophenomics methods for classifying patients diagnosed with mental disorders into subtypes to optimize treatment interventions. These methods can be used to configure diagnoses to recommend the best treatment match for individual patients. In another embodiment, these methods can be used to enhance patient selection based on pharmacophenomics stratification prior to clinical trials. In another embodiment, these methods can be used to configure companion diagnostics for psychotropic drugs to ensure patient safety during drug development, marketing, and post-marketing periods.

[0017] In another embodiment, clinical values are obtained from an EHR or similar source, and SNPs in PD and PK genes are obtained from the patient's genotype, and these are input as quantitative values into a regression equation (nomogram) to determine the drug dosage, such as for ketamine, for the individual patient. In this embodiment, the optimal treatment dose is developed in a stepwise regression model equation containing genetic and clinical values, and the regression model is retested and validated using a patient population to ensure the accuracy of the regression equation output, as can be judged by the area under the curve (AUC) of the receiver-operator (ROC) curve.

[0018] In another embodiment, clinical values are obtained from an EHR or similar source, and SNPs in PD and PK genes are combined with disease risk SNPs obtained from genome-wide association studies (GWAS) to differentially annotate adverse event and efficacy-specific subnetworks of a pharmaco-genomics network, such as ketamine, to predict whether a patient will benefit from the drug and, if so, to determine the dose suitable for the patient.

[0019] In another embodiment, clinical values are obtained from an EHR or similar source, and SNPs in PD and PK genes are combined with disease risk SNPs obtained from GWAS and PheWAS to differentially annotate adverse event and efficacy-specific subnetworks of a pharmaco-genomics network, such as ketamine, to predict whether a patient will benefit from the drug and, if so, to determine the dose suitable for the patient. In this embodiment, therapeutic drug monitoring by pharmacometabolomics is used to collect more accurate data on pre-existing prescription and over-the-counter drugs and their metabolites used by the patient by analyzing biological samples (blood, cheek swabs, urine, or other body fluids) obtained from the patient or a group of patients.

[0020] In yet another embodiment, clinical values are obtained from an EHR or similar source, and SNPs in PD and PK genes are combined with disease risk SNPs obtained from GWAS and PheWAS to differentially annotate adverse event and efficacy-specific subnetworks of a pharmacogenomics network, such as ketamine, and this data is matched to one of four phenotypes, which may or may not be from rTMS and neuroimaging data, and scored using the Hamilton Depression Rating Scale (HAMD) in the context of an antidepressant drug, such as ketamine.

[0021] In another embodiment, inputs from a model comprising: (1) molecular profiling of drug-induced subnetworks in a patient or group of patients, (2) clinical variables such as those derived from an electronic health record or equivalent measurements made by a clinician, (3) patient subtypes based on clinical variables and neuroimaging studies, and (4) PD and PK SNPs that stratify patients by drug response are used to determine pharmacophenomics decision support. Additionally, drug-drug and drug-gene interactions objectively measured using pharmacometabolomics methods can be used to minimize adverse drug events in individual patients or groups of patients.

[0022] Another embodiment of this system is the configuration of a companion diagnostic, which can be used for patient selection in clinical trials, as well as during the pre-market and post-market periods of a drug, such as antidepressant drugs that are NMDAR antagonists, partial antagonists, GLRB modulators, and AMPAR agonists.

[0023] Another embodiment of this system is to determine and select therapeutic drugs to add to an NMDAR modulator to improve outcomes. Another embodiment of the methods and systems described herein can be used to reevaluate drugs for clinical trials and for drug repurposing. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1A A block diagram of a computer network and system on which an exemplary companion diagnostic system according to the presently described embodiments can operate is shown;

[0025] Figure 1B is a block diagram of an exemplary drug and dose decision server that can operate in a Figure 1A system according to the presently described embodiments;

[0026] Figure 1C is a block diagram of an exemplary client device that can operate in a Figure 1A system according to the presently described embodiments;

[0027] Figure 1DDisclosed is a flowchart which represents an exemplary method for determining a drug and dosage to be administered to a patient suffering from depression or other neuropsychiatric disorders based on comparing data from a biological sample of the patient with a reference drug-specific pharmacogenomics network and a constituent sub-network of the drug of interest;

[0028] Figure 2 Disclosed are exemplary measurements collected from a biological sample of a patient using chromosome conformation capture, bioinformatics analysis, and / or similar measures;

[0029] Figure 3 Illustrated is a simple example of how an SNP located within an enhancer in a network can disrupt the contact between the enhancer and one of its target gene promoters in a TAD, resulting in an adverse drug event in a patient within the drug response cohort. Figure 3 A shows how measurements are obtained from a three-dimensional chromatin spatial interactome using different laboratory methods and the data is analyzed as a two-dimensional map of enhancer-gene promoter interactions. Figure 3 B depicts how an SNP can disrupt the chromatin loop between an enhancer and one of the two gene promoters it regulates within a TAD. This disruption eliminates the spatial connection between the enhancer and gene promoter 1, leading to dysregulation of gene 1 and resulting in an adverse event in this patient and its cohort in response to the administration of a specific drug of interest;

[0030] Figure 4 Disclosed is a flowchart and a scoring system which represent an exemplary method for determining a suitable drug and dosage to be administered to a patient suffering from depression or other neuropsychiatric disorders while avoiding adverse events (AE) based on comparing data from a biological sample of the patient with a reference drug-specific pharmacogenomics network and a constituent sub-network of the drug of interest;

[0031] Figure 5 Disclosed is a flowchart which represents an exemplary method for safely prescribing a given drug at a given dosage, continuing with caution, or stopping, using the efficacy and adverse event sub-networks derived from an individual patient's pharmacogenomics network and prescribing based on a quantitative determination of a molecular phenotype generated from the drug sub-network;

[0032] Figure 6 Disclosed are two different methods based on post hoc bioinformatics analysis and disease risk SNP annotation from GWAS to determine that patients of a specific ancestry should receive ketamine based on disease gene risk variant analysis and the presence of disease risk SNPs associated with the ketamine adverse event sub-network 3 and efficacy sub-network 2;

[0033] Figure 7Shows how, after selecting a specific drug from the reference set database of the PharmacoGenomics Network and its constituent subnetworks, a model for fine-tuning a series of subnetwork types across the range of human drug response phenotypes is used with machine learning compared to the model for the patient input sample data, for models created from a set of highly heterogeneous hierarchical biomedical and biological data types and elements;

[0034] Figure 8 Shows a flowchart of an exemplary method for matching the pharmacodynamic efficacy and adverse events of a patient with those of a reference PharmacoGenomics Network using a similarity score, for discovering novel pharmacodynamic targets using TAD matrix plotting and deep learning based on computer vision algorithms, to allow the study of drug similarity using TAD patterns and / or the use of TAD matching methods in clinical trials studying the drug TAD profile;

[0035] Figure 9A Shows an exemplary model of how the system integrates heterogeneous hierarchical biomedical and biological data and processes these multi-scale data using machine learning and deep learning for PharmacoGenomics Network topology and subnetwork reconstruction. This strategy for mapping the drug network provides insights into on-target and off-target effects. Using the discovered PharmacoGenomics Network topology provides a basis for advanced PharmacoGenomics decision support, laying the foundation for refining this ability for subsequent preclinical and clinical studies, which can also be used for drug mechanism discovery and drug repurposing prediction;

[0036] Figure 9B Shows a flowchart representing a method for generating a reconstructed PharmacoGenomics Network and corresponding subnetworks for a drug of interest, including a human PharmacoGenomics SNP input filter, a drug space network reconstruction engine, and an iterative gene set optimization engine. A reference set of such PharmacoGenomics Networks, drug efficacy, and drug adverse event information can also be created;

[0037] Figure 10 Shows a flowchart representing an exemplary method for integrating drug-gene set optimization to construct a PharmacoGenomics Network and determine its subnetworks;

[0038] Figure 11 Shows a flowchart for post hoc validation of a reconstructed PharmacoGenomics Network and a set of such networks using bioinformatics data / software and an integrated drug informatics pipeline;

[0039] Figure 12An example of a comparison of the numerical outputs of six different machine learning algorithms as used in usage analysis and the significance test results of SNP rs12967143-G (located in an intragenic enhancer within the TCF4 gene) relative to other GWAS SNPs as described for various neural and non-neural cell types is shown;

[0040] Figure 13 Characteristics of two different ketamine pharmacogenomic subnetworks are shown, as determined by post hoc validation of the ketamine pharmacogenomic network in the human brain. Figure 13 A is the gene enrichment of the ketamine pharmacogenomic subnetwork in the human brain that mediates efficacy and neuroplasticity. Figure 13 B is the gene enrichment of the ketamine pharmacogenomic subnetwork in the human brain that mediates glutamate receptor signaling and adverse events in the human brain;

[0041] Figure 14A A graphical depiction of the ketamine pharmacogenomic subnetwork in the human brain that mediates efficacy and neuroplasticity and the diseases and conditions associated with efficacy and neuroplasticity is shown;

[0042] Figure 14B A graphical depiction of the ketamine pharmacogenomic subnetwork in the human brain that mediates glutamate receptor signaling and adverse events and the diseases and conditions associated with this signaling and these adverse events is shown;

[0043] Figure 15 Genes and regulatory RNAs located in the ketamine efficacy and neuroplasticity subnetwork are listed;

[0044] Figure 16 Genes located in the glutamate-ketamine receptor signaling and adverse events subnetwork are listed;

[0045] Figure 17 Genes located in the ketamine pharmacokinetics and hormonal regulation subnetwork are listed;

[0046] Figure 18A - 18B Example values of a linear regression analysis for ketamine dose determination and the accuracy of ketamine doses in a validation cohort are shown;

[0047] Figure 19 Definition of four subtypes of treatment-resistant depression (TRD) patients as determined by transcranial magnetic stimulation (TMS) coupled with neuroimaging of the resting connectivity network in the human brain is shown;

[0048] Figure 20 Example neurographs of four different subtypes of TRD depressive patients are shown;

[0049] Figure 21Shows example drug prescribing / non-prescribing recommendations and alternative drug options for each of four different subtypes of TRD depression patients;

[0050] Figure 22 Shows examples of synergistic histone modification-mediated beneficial combination mechanisms and therapeutics discovered using the methods described herein, with valproic acid and ketamine used for H3K9 acetylation and deacetylation, respectively, and the combination resulting in neurogenesis and neural differentiation;

[0051] Figure 23 Shows Figure 23 The complementary pharmacogenomics network of valproic acid for A, and Figure 23 The pharmacogenomics network of ketamine for B, respectively showing neurogenesis and neural differentiation; and

[0052] Figure 24 Shows the combination and biological synergy of valproic acid and ketamine pharmacogenomics networks in neurogenesis, neuronal proliferation, and terminal neuronal differentiation. Detailed Description

[0053] Although the following text sets forth a detailed description of many different embodiments, it should be understood that the legal scope of this description is defined by the words of the claims set forth at the end of this disclosure. The detailed description should be construed as merely exemplary and does not describe every possible embodiment, as it would be impractical, if not impossible, to describe every possible embodiment. Many alternative embodiments can be implemented using current technology or technology developed after the filing date of this patent application, and such embodiments will still fall within the scope of the claims.

[0054] It should also be understood that unless a term is explicitly defined in this patent by the sentence "As used herein, the term '______' is defined herein to mean..." or a similar sentence, there is no intent to limit the meaning of such term, whether expressly or by implication, beyond its ordinary or common meaning, and such term should not be construed as being limited in scope by any statement made in any section of this patent (except the language of the claims). To the extent that any term is referred to in this patent in a manner consistent with a single meaning in the claims recited at the end of this patent, this is done merely for clarity so as not to confuse the reader and is not intended to limit such claim term by implication or otherwise to the single meaning. Finally, unless a claim element is defined by reference to the word "means" and the recitation of a function without any structure, the scope of any claim element is not intended to be construed in accordance with the provisions of 35 U.S.C. § 112, paragraph 6.

[0055] The present disclosure includes systems and methods for stratifying patients or patient populations diagnosed with a mental disorder or in need of these drugs for other clinical indications to accurately select drugs and doses of NMDAR antagonists. Ketamine is used as an example, but these methods can be used for NMDAR antagonists or α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor modulators in clinical trials for the clinical indication of treatment-resistant depression. For patients diagnosed with TRD or specific patient subgroups, ketamine and its enantiomers are novel antidepressants that exhibit greater efficacy and fewer side effects compared to other antidepressants. A pharmacogenomics decision support system can determine which patients diagnosed with TRD should receive ketamine or one of its enantiomers as an antidepressant drug, and if so, what the appropriate dose should be for an individual patient or patient population to maximize efficacy, maximize adverse drug events and drug-drug interactions, and reduce the harmful effects of drug-gene and drug-drug interactions. Adverse psychotropic drug side effects, including the psychotomimetic and neurocognitive effects of glutamate receptor-targeted drugs used to relieve depression, and the heterogeneity of patient populations that may exhibit treatment-resistant depression (TRD) are based on multiple variables, including demographic and sociological variables, trauma history, genotype, and clinical variables.

[0056] The system includes several methods that can be used to configure clinical decision support diagnostics for the selection and administration of ketamine or its enantiomers as antidepressants for treatment-resistant depression and can be generalized to other psychotropic drugs. One embodiment includes an integrated multi-scale measurement system that includes Figure 1D the components shown, where patients' biological samples can be analyzed as broadly as Figure 2 shown and other relevant clinical data can also be obtained.

[0057] In some embodiments, the system is configured to analyze the results of the minimum amount of biological sample required to match a stored population subnet reference set. In this case, a learning machine pre-trained with a population response range for a specific drug is used to compare patient data with a reference set of drug subnets spanning the entire range of the human drug response population, and matching the reference set prompts clinical decision recommendations. For example, for a specific drug, the reference set of drug subnets can include a set of reference drug pharmacodynamic efficacy subnets, reference drug pharmacodynamic adverse event subnets, reference chromatin remodeling subnets, and reference pharmacokinetic enzyme and hormone subnets.

[0058] An electronic encryption proxy is first used to protect health information through de-identification and prevent patient identification. A biological sample (e.g., blood, cheek swab, saliva, urine, or other bodily fluid) is obtained from a patient or group of patients, with clinical data from a medical record, such as an electronic health record (EHR) or other source, attached thereto. An initial pharmacometabolomics analysis is performed on a small blood sample or its plasma fraction collected from the patient or group of patients to identify potential drug-drug and drug-gene interactions that may alter subsequent pharmacogenomics decision support. These objective measurements augment self-reported, clinician-reported, or other data included in the EHR or other patient records.

[0059] Generally, techniques for determining whether to administer a drug, such as a glutamate NMDAR antagonist or partial antagonist, a GLRB modulator, or an AMPAR agonist, to a patient and / or determining an appropriate dose of a drug administered to a patient can be implemented in one or more client devices, one or more network servers, or a system comprising a combination of these devices. However, for clarity, the examples below mainly focus on embodiments in which a healthcare professional obtains a biological sample from a patient and provides the biological sample to a testing laboratory for analysis.

[0060] Biological samples can include a subject's skin, blood, urine, sweat, lymph fluid, bone marrow, cheek cells, saliva, cell lines, tissues, etc. Cells are then extracted from the biological sample and reprogrammed into stem cells (such as induced pluripotent stem cells (iPSCs)). The iPSCs are then differentiated into various tissues such as neurons, cardiomyocytes, etc., and assays are performed to obtain genomic data, chromosomal data, pharmacometabolomics data, etc. of the patient. In some embodiments, the iPSCs can be assayed for loci associated with or causally related to the phenotypic response to the drug of interest. The iPSCs comprise a part of a reference set for deriving variables for evaluating an individual patient.

[0061] The drug and dose decision server relies on a drug and dose decision support engine. This engine receives a numerical score representing the overlap between an input patient sample as shown Figure 2 and a relevant drug-specific reference pharmacogenomics network stored in a database 154 of such references. The system uses a learning machine trained over the entire human drug response population, which consists of a reference set of pharmacogenomics networks for a particular drug, covering the range of human drug response variations to that particular drug, including the reference set that matches the input patient sample.

[0062] The drug and dose decision server analyzes laboratory results to determine a subnetwork representation of a patient for a drug gene set of an NMDAR antagonist or partial antagonist, a GLRB modulator, or an AMPAR agonist (such as ketamine). Additionally, the drug and dose decision server retrieves a reference pharmacogenomics network and a constituent reference subnetwork of an NMDAR antagonist or partial antagonist, a GLRB modulator, or an AMPAR agonist from, for example, a reference drug pharmacogenomics network database. Then, the drug and dose decision server compares the patient's drug subnetwork representation with the reference pharmacogenomics network and the constituent reference subnetwork of the drug to determine whether to administer the drug to the patient. For example, the drug and dose decision server can compare the patient's efficacy drug-specific (e.g., ketamine) subnetwork with the reference efficacy drug-specific (e.g., ketamine) subnetwork, and can compare the patient's adverse event drug-specific (e.g., ketamine) subnetwork with the reference adverse event drug-specific (e.g., ketamine) subnetwork. If the similarity between the patient's efficacy drug-specific subnetwork and the reference efficacy drug-specific subnetwork is greater than a threshold (indicating that the drug may be effective for the patient), then the drug and dose decision server can then determine that the drug should be administered to the patient. If the similarity between the patient's adverse event drug-specific subnetwork and the reference adverse event drug-specific subnetwork is below a threshold (indicating that the patient is less likely to experience an adverse event) or based on some combination of the two, then the drug and dose decision server can also determine that the drug should be administered to the patient.

[0063] Accordingly, the drug and dose decision server can provide a recommendation indicating that the patient should receive the drug to the client device of a healthcare professional, enabling the healthcare professional to administer the drug to the patient. Thus, the healthcare professional can administer the drug to the patient. In some embodiments, the drug and dose decision server can determine the drug dose to be administered to the patient according to a dosing algorithm. The dosing algorithm can be determined using machine learning techniques (such as linear regression) and can be based on the patient's demographic data, the patient's clinical data, the patient's biological data, and the like.

[0064] A drug and dosage decision server can determine the drug dosage to be administered to a patient and perform other methods described herein using various machine learning techniques, including but not limited to regression algorithms (e.g., ordinary least squares regression, linear regression, logistic regression, stepwise regression, multivariate adaptive regression splines, locally estimated scatterplot smoothing, etc.), instance-based algorithms (e.g., k-nearest neighbor, learning vector quantization, self-organizing maps, locally weighted learning, etc.), regularization algorithms (e.g., ridge regression, least absolute shrinkage and selection operator, elastic net, least angle regression, etc.), decision tree algorithms (e.g., classification and regression trees, C4.5, C5, chi-squared automatic interaction detection, decision stumps, M5, conditional decision trees, etc.), clustering algorithms (e.g., k-means, k-medians, expectation maximization, hierarchical clustering, spectral clustering, mean shift, density-based spatial clustering of applications with noise, ordering points to identify the clustering structure, etc.), association rule learning algorithms (e.g., Apriori algorithm, Eclat algorithm, etc.), Bayesian algorithms (e.g., naive Bayesian, Gaussian naive Bayesian, multinomial naive Bayesian, averaged one-dependence estimators, Bayesian belief networks, Bayesian networks, etc.), artificial neural networks (e.g., perceptrons, Hopfield networks, radial basis function networks, etc.), deep learning algorithms (e.g., multi-layer perceptrons, deep Boltzmann machines, deep belief networks, convolutional neural networks, stacked autoencoders, generative adversarial networks, etc.), dimensionality reduction algorithms (e.g., principal component analysis, principal component regression, partial least squares regression, Sammon mapping, multidimensional scaling, projection pursuit, linear discriminant analysis, mixture discriminant analysis, quadratic discriminant analysis, flexible discriminant analysis, factor analysis, independent component analysis, non-negative matrix factorization, t-distributed stochastic neighbor embedding, etc.), ensemble algorithms (e.g., boosting, bagging, AdaBoost, stacking generalization, gradient boosting machines, gradient boosting regression trees, random decision forests, etc.), reinforcement learning (e.g., temporal difference learning, Q-learning, learning automata, state-action-reward-state-action, etc.), support vector machines, mixture models, evolutionary algorithms, probabilistic graphical models, etc.

[0065] Reference Figure 1A, an example pharmacogenomics decision support system 100 determines whether to administer a psychotropic drug to a patient with depression, such as a patient with TRD, and the appropriate dose of the psychotropic drug. The pharmacogenomics decision support system 100 includes a drug and dose decision server 102 and a plurality of client devices 106 - 116 that can be communicatively connected via a network 130, as described below. In one embodiment, the drug and dose decision server 102 and the client devices 106 - 116 can communicate via a wireless signal 120 over a communication network 130, which can be any suitable local area network or wide area network, including a WiFi network, a Bluetooth network, a cellular network (such as 3G, 4G, Long Term Evolution (LTE), 5G), the Internet, etc. In some cases, the client devices 106 - 116 can communicate with the communication network 130 via an intermediate wireless or wired device 118, which can be a wireless router, a wireless repeater, a base transceiver station of a mobile phone provider, etc. For example, the client devices 106 - 116 can include a tablet computer 106, a smartwatch 107, a network-enabled cellular phone 108, a wearable computing device (such as Google Glass TM or 109), a personal digital assistant (PDA) 110, a mobile device smart phone 112 (also referred to herein as a "mobile device"), a laptop computer 114, a desktop computer 116, a wearable biosensor, a portable media player (not shown), a phablet, any device configured to perform wired or wireless RF (radio frequency) communication, etc. Additionally, any other suitable client device that records patient clinical data can also communicate with the drug and dose decision server 102.

[0066] Each of the client devices 106 - 116 can interact with the drug and dose decision server 102 to receive recommendations regarding whether to administer a psychotropic drug to a patient and the dose of the psychotropic drug. The client devices 106 - 116 can present the recommendations via a user interface for display to a healthcare professional.

[0067] In an example implementation, the drug and dose decision server 102 can be a cloud-based server, an application server, a web server, etc., and includes a memory 150, one or more processors (CPUs) 142 (such as a microprocessor coupled to the memory 150), a network interface unit 144, and an I / O module 148, which can be, for example, a keyboard or a touch screen.

[0068] The drug and dose decision server 102 can also be communicatively connected to a reference drug pharmacogenomics network and a database 154 that constitutes a sub-network (such as a sub-network of drug efficacy and adverse events).

[0069] Memory 150 may be a tangible non-transitory memory and may include any type of suitable memory module, including random access memory (RAM), read only memory (ROM), flash memory, other types of persistent memory, etc. Memory 150 may store, for example, instructions for an operating system (OS) 152 that can be executed on processor 142, and the operating system may be any type of suitable operating system, such as a modern smartphone operating system. Memory 150 may also store, for example, instructions for a drug and dose decision support engine 146 that can be executed on processor 142. The drug and dose decision server 102 is described in more detail below with reference to Figure 1B In some embodiments, the drug and dose decision support engine 146 may be part of one or more of the client devices 106-116, the drug and dose decision server 102, or a combination of the drug and dose decision server 102 and the client devices 106-116.

[0070] In any case, the drug and dose decision support engine 146 may obtain laboratory results from a patient biological sample only as necessary to match one of the sets of pharmacogenomic networks and their constituent subnetworks that define the human drug response variation for a particular drug. These include molecular data, which includes genomic variations defined by SNPs in PD and PK genes, pharmacogenomic interactions between regulatory elements, gene and / or direct topologically associating domain (TAD)-specific measurements in the patient genome that can be defined using chromosome conformation data (such as Hi-C), which includes differential gene expression determined using RNA sequencing (RNA-Seq) or expression microarray profiling and patient-specific TAD contactome measurements in relevant or alternative cell types using chromosome conformation capture (e.g., 3C, 4C, 5C, Hi-C, ChIA-PET, and GAM). Molecular data may be assayed for loci associated with or causally related to the phenotypic response of the drug of interest. Additionally, the drug and dose decision support engine 146 may obtain the reference pharmacogenomic network and constituent reference subnetworks for the psychotropic drug of interest (e.g., ketamine) from the reference pharmacogenomic network database 154.

[0071] Then, the medication and dosage decision support engine 146 can analyze the patient's laboratory results to determine the sub-network representations of the psychotropic medications of interest, such as the efficacy sub-network and the adverse event sub-network. The medication and dosage decision support engine 146 can compare the patient's efficacy sub-network and adverse event sub-network with reference efficacy and adverse event sub-networks to determine whether to administer the psychotropic medication of interest to the patient. If the similarity between the patient's efficacy drug-specific sub-network and the reference efficacy drug-specific sub-network is greater than a threshold and / or the similarity between the patient's adverse event drug-specific sub-network and the reference adverse event drug-specific sub-network is less than a threshold, then the medication and dosage decision support engine 146 can determine that the psychotropic medication of interest should be administered to the patient. The medication and dosage decision support engine 146 can then provide a recommendation to the client devices 106-116 of the healthcare professional, the recommendation indicating that the patient should receive the psychotropic medication of interest. Otherwise, the medication and dosage decision support engine 146 can provide a recommendation for an alternative medication administered to the patient to treat depression. Additionally, the medication and dosage decision support engine 146 can determine the dosage of the psychotropic medication of interest administered to the patient according to a dosing algorithm. The medication and dosage decision support engine 146 can also provide the recommended dosage of the psychotropic medication of interest to the client devices 106-116 of the healthcare professional.

[0072] The medication and dosage decision server 102 can communicate with the client devices 106-116 via the network 130. The digital network 130 can be a private network, a secure public internet, a virtual private network, and / or some other type of network, such as a dedicated access line, an ordinary conventional telephone line, a satellite link, a combination of these, etc. In the case where the digital network 130 includes the Internet, data communication can be carried out on the digital network 130 via the Internet communication protocol.

[0073] Now turning to Figure 1B , the medication and dosage decision server 102 can include a controller 224. The controller 224 can include a program memory 226, a microcontroller or microprocessor (MP) 228, a random access memory (RAM) 230, and / or input / output (I / O) circuitry 234, all of which can be interconnected via an address / data bus 232. In some embodiments, the controller 224 can also include a database 239, or otherwise be communicatively connected to the database or other data storage mechanisms (e.g., one or more hard disk drives, optical storage drives, solid state storage devices, etc.). The database 239 can contain data such as pharmacogenomic network reference data for medications, medication recommendation display templates, web page templates, and / or web pages, and other data necessary for interacting with users via the network 130. The database 239 can contain data similar to the database 154 described above with reference to Figure 1A description.

[0074] It should be understood that although Figure 1B only one microprocessor 228 is depicted, the controller 224 may include multiple microprocessors 228. Similarly, the memory of the controller 224 may include multiple RAMs 230 and / or multiple program memories 226. Although Figure 1B the I / O circuit 234 is described as a single block, the I / O circuit 234 may include many different types of I / O circuits. The controller 224 may implement, for example, the RAM 230 and / or the program memory 226 as semiconductor memories, magnetically readable memories, and / or optically readable memories.

[0075] As Figure 1B shown, the program memory 226 and / or the RAM 230 may store various applications for execution by the microprocessor 228. For example, the user interface application 236 may provide a user interface to the drug and dose decision server 102, and the user interface may, for example, allow a system administrator to configure, troubleshoot, or test various aspects of the server operation. The server application 238 may operate to receive molecular data of a patient, analyze the molecular data to determine a sub-network of the patient related to a specific drug of interest, compare the sub-network of the patient with a reference sub-network of the specific drug of interest, determine to administer the specific drug of interest to the patient based on the comparison, and transmit a recommendation to administer the specific drug of interest to the patient to the client devices 106 - 116. The server application 238 may be a single module 238, such as the drug and dose decision support engine 146 or multiple modules 238A, 238B.

[0076] Although in Figure 1B the server application 238 is depicted as including two modules 238A and 238B, the server application 238 may include any number of modules that complete tasks related to the implementation of the drug and dose decision server 102. In addition, it should be understood that although in Figure 1B only one drug and dose decision server 102 is depicted, multiple drug and dose decision servers 102 may be provided for distributing server loads, serving different web pages, etc. These multiple drug and dose decision servers 102 may include web servers, entity-specific servers (such as servers, etc.), servers located in retail or private networks, etc.

[0077] Now refer to Figure 1C, the laptop computer 114 (or any one of the client devices 106 - 116) may include a display 240, a communication unit 258, a user input device (not shown), and a controller 242 such as the drug and dose decision server 102. Similar to the controller 224, the controller 242 may include a program memory 246, a microcontroller or microprocessor (MP) 248, a random access memory (RAM) 250, and / or input / output (I / O) circuits 254, all of which may be interconnected via an address / data bus 252. The program memory 246 may include an operating system 260, a data storage device 262, a plurality of software applications 264, and / or a plurality of software routines 268. For example, the operating system 260 may include Microsoft OS and so on. The data storage device 262 may include data such as application data of the plurality of applications 264 as described, routine data of the plurality of routines 268, and / or other data necessary for interacting with the drug and dose decision server 102 via the digital network 130. In some embodiments, the controller 242 may also include other data storage mechanisms (e.g., one or more hard disk drives, optical storage drives, solid state storage devices, etc.) residing within the laptop computer 114, or otherwise communicatively connected to the other data storage mechanisms.

[0078] The communication unit 258 may communicate with the drug and dose decision server 102 via any suitable wireless communication protocol network such as a wireless telephone network (e.g., GSM, CDMA, LTE, etc.), a Wi-Fi network (802.11 standard), a WiMAX network, a Bluetooth network, etc. The user input device (not shown) may include a "soft" keyboard displayed on the display 240 of the laptop computer 114, an external hardware keyboard (e.g., a Bluetooth keyboard) communicating via a wired or wireless connection, an external mouse, a microphone for receiving voice input, or any other suitable user input device. As discussed with reference to the controller 224, it should be understood that although Figure 1C only one microprocessor 248 is depicted, the controller 242 may include multiple microprocessors 248. Similarly, the memory of the controller 242 may include multiple RAMs 250 and / or multiple program memories 246. Although Figure 1C the I / O circuits 254 are described as a single block, the I / O circuits 254 may include many different types of I / O circuits. The controller 242 may implement one or more RAMs 250 and / or program memories 246 as, for example, semiconductor memories, magnetically readable memories, and / or optically readable memories.

[0079] In addition to other software applications, one or more processors 248 may be adapted and configured to execute any one or more of a plurality of software applications 264 and / or any one or more of a plurality of software routines 268 residing in program memory 246. One of the plurality of applications 264 may be a client application 266, which may be implemented as a series of machine-readable instructions for performing various tasks associated with receiving information at the laptop computer 114, displaying information on the laptop computer, and / or transmitting information from the laptop computer.

[0080] One of the plurality of applications 264 may be a native application and / or a web browser 270 (such as Apple's Google Chrome TM , Microsoft Internet and Mozilla ), which may be implemented as a series of machine-readable instructions for receiving, interpreting, and / or displaying web page information from the drug and dose decision server 102 while also receiving input from a user such as a healthcare professional or researcher. Another of the plurality of applications may include an embedded web browser 276, which may be implemented as a series of machine-readable instructions for receiving, interpreting, and / or displaying web page information from the drug and dose decision server 102.

[0081] One of the plurality of routines may include a drug recommendation display routine 272, which presents on the display 240 a recommendation as to whether to administer a psychotropic drug of interest to a patient and / or a recommended dose.

[0082] Preferably, a user may initiate the client application 266 from a client device (such as one of the client devices 106 - 116) to communicate with the drug and dose decision server 102 to implement the companion diagnostic system 100. Additionally, the user may also initiate or instantiate any other suitable user interface application (e.g., a native application or web browser 270, or any other application among the plurality of software applications 264) to access the drug and dose decision server 102 to implement the companion diagnostic system 100.

[0083] Figure 1D A flowchart is shown that represents an exemplary method 160 for determining a drug and dose to administer to a patient suffering from depression based on comparing data from a biological sample of the patient with a reference pharmacogenomics network and a constituent subnetwork of the drug of interest. Method 160 may be executed by the drug and dose decision server 102.

[0084] In some embodiments, as Figure 1D shown, analyzing a patient's biological sample for personalized therapy to quantify the relative activation of different pathways mediating the mechanism of action of ketamine in the human CNS, the pathways determined using the methods described herein. A drug metabolomics assay can be used to analyze a patient or patient cohort biological sample to identify drugs or metabolites in the sample that may cause unwanted drug-drug interactions, thereby affecting efficacy, adverse events, and drug dosing. Biological sample measurements include: (1) genotyping of pharmacokinetic SNPs, which in the case of the ketamine paradigm consisting of CYP2B6 gene mutations have been shown to be determinants of metabolizer status (poor, below normal, normal, or ultrarapid subtypes), (2) pharmacodynamic SNP targeting as input to a pharmacogenomic network and subnet profiling analysis that is decisive for both efficacy and adverse events as analyzed using a pharmacogenomic genomic classifier and a pharmacodynamic subnet profiling system, (3) direct topologically associating domain (TAD) specific measurements, which include differential gene expression determined using RNA sequencing (RNA-Seq) or expression microarray profiling and patient-specific TAD contactome measurements in relevant or alternative cell types using chromosome conformation capture (e.g., 3C, 4C, 5C, Hi-C, ChIA-PET, and GAM), and (4) drug metabolomics analysis.

[0085] Figure 1D The data processing pipeline shown in includes parallel routes for analyzing biological samples using multiple methods and analyzing available clinical data for the same patient that may be collected from an electronic health record (EHR). This example demonstrates the analysis for determining whether an N-methyl-D-aspartic acid receptor (NMDAR) modulator should be administered to a particular patient. Adverse events associated with NMDAR modulators such as ketamine can be severe and include severe dissociation, hallucinations, and nightmares. Thus, the first decision made from these parallel analyses is to ensure that a patient does not receive an NMDAR modulator such as ketamine if the system predicts that the individual will experience moderate to severe adverse events.

[0086] Figure 1D The data processing pipeline shown in may make a "no-go" decision for administering an NMDAR modulator such as ketamine based on a molecular network representation in the patient, certain disease risk SNPs the patient may carry, as Figure 23 shown, the patient's treatment-resistant depression phenotype, as Figure 20 and Figure 22 shown, and dose adjustment based on the patient's drug metabolism profile. If the system decides that the patient can receive an NMDAR modulator such as ketamine, then a dose determination algorithm is initiated.

[0087] Method 160 is used as a clinical decision support diagnosis to determine whether the NMDAR module should be prescribed to a patient as an antidepressant (block 162) and the optimal dosage for the patient. In some embodiments, a patient biological sample is analyzed (block 164a) for personalized therapy to quantify the relative activation of different pathways that mediate the mechanism of action of ketamine in the human CNS, the pathways being determined using the methods described herein. A drug metabolism assay can be used to analyze a patient or patient cohort biological sample to determine drugs or metabolites in the sample that may result in unwanted drug-drug interactions, which can affect efficacy, adverse events, and drug administration. At block 166, the biological sample measurements include: (1) genotyping of pharmacokinetic SNPs, which in the case of the ketamine paradigm consisting of CYP2B6 gene mutations have been shown to be determinants of metabolic status (poor, below normal, normal, or ultra-rapid subtypes), (2) pharmacodynamic SNP targeting as input to a pharmacogenomics network and subnetwork profiling, which profiling is decisive for both efficacy and adverse events as analyzed using a pharmacogenomics genome classifier and a pharmacodynamic subnetwork profiling system, (3) direct topologically associating domain (TAD) specific measurements, which include differential gene expression determined using RNA sequencing (RNA-Seq) or expression microarray profiling and patient-specific TAD contactome measurements in relevant or alternative cell types using chromosome conformation capture (e.g., 3C, 4C, 5C, Hi-C, ChIA-PET, and GAM), and (4) drug metabolomics analysis (block 164c).

[0088] As described above, the biological sample measurements include pharmacodynamic SNP targeting as input to a pharmacogenomics network and subnetwork profiling (blocks 168, 170b), which profiling is decisive for both efficacy and adverse events as analyzed using a pharmacogenomics genome classifier and a pharmacodynamic subnetwork profiling system. At block 170a, a reference pharmacogenomics network and subnetwork for the drug of interest are retrieved from a reference database. Then, the patient subnetwork for the specific drug of interest that includes efficacy and adverse event subnetworks is compared to the reference pharmacogenomics network and subnetwork for the drug of interest (block 172). To determine the similarity to the reference set, two pairs of different reference-patient metrics include an accurate measure of similarity and an output similarity score for each of the efficacy and adverse event subnetworks. At block 172, the similarity scores for the efficacy and adverse event subnetworks for the drug of interest can be used to determine whether to administer the drug of interest to the patient. For example, if the similarity score for the efficacy subnetwork is higher than a threshold similarity score, method 160 can determine that the drug of interest should be administered to the patient (block 176). Otherwise, method 160 determines to select a different drug product (block 174).

[0089] In addition to comparing the patient's sub-network of the drug of interest with a reference sub-network of the drug of interest, the patient's clinical data is collected and analyzed at block 164b to determine whether to administer a drug and / or the dosage of the drug to the patient. More specifically, the patient's HAMD score and / or the patient's symptoms can be analyzed to classify the patient into one of four TRD patient subtypes (block 180). The TRD patient subtypes are described in more detail below with reference to Figure 15 - 17 If the patient is classified as TRD subtype 3 (block 182), then method 160 determines to select a different drug product (block 184). Other clinical data can also be analyzed, such as the patient's drug-drug interactions, age, weight, biological sex, body mass index, race, family history, patient substance abuse history, diagnosis codes, hospitalization history, drug-gene interactions, psychiatric history, whether the patient smokes or uses nicotine, etc. (block 186).

[0090] Then the dosage of the drug of interest to be administered to the patient is determined (block 178). The dosage can be determined based on a dosing algorithm with a predetermined constant to be applied to each of a number of patient characteristics, such as biological characteristics, demographic characteristics, clinical characteristics, etc. In other embodiments, machine learning techniques can be used to generate the dosing algorithm. The patient characteristics used in the dosing algorithm can include biological data, such as SNPs reported to stratify the response to ketamine in humans. The patient characteristics can also include the patient's demographic data, such as the patient's sex, height and weight, age and race. In addition, the patient characteristics can include clinical data, such as family history, drug-drug interactions, psychiatric history, whether the patient smokes or uses nicotine, and Hamilton Depression Scale (HAM-D) score.

[0091] Figure 2 Illustrated are various measures that can be taken from the patient's biological samples. In one embodiment, blood and oral swab samples are obtained and processed. For ketamine and other drugs that undergo first-pass metabolism by proteins encoded by the ultra-induced and ultra-variable CYP2B6 gene, targeted SNP genotyping is performed using a 4-SNP panel, but most importantly, the splicing variant SNP rs3745274 is preferentially examined because it is relatively common in the population (>10% frequency), and carriers of this SNP include ultra-weak metabolizers of any drug primarily metabolized by this enzyme and are likely to experience adverse drug events from NMDAR modulators such as ketamine.

[0092] In Figure 1D the methods shown and Figure 2 the measurements collected, it is recommended to perform a drug metabolomics analysis on the patient's blood to rule out the possibility of any negative drug-drug or drug-gene interactions.

[0093] Figure 2 shows the assays that can be performed on a patient's biological sample to obtain the minimum information required to allow a machine or deep learning algorithm to match a drug-specific pattern to a comprehensive subnetwork reference set, the comprehensive subnetwork including the activation of topologically associated domains (TADs) contained in database 154. These methods can include measuring TAD-specific changes in gene expression using RNA-seq or expression microarrays, pharmacogenomic contacts between genes using chromosome conformation capture assays such as Hi-C, and / or genotyping of genes targeted after a geometric change is induced in the pharmacogenomic genome by a drug, either after administration of a specific drug to the patient or prior to deciding whether a drug should be administered, by actively analyzing oral swabs obtained from the patient.

[0094] Figure 2 Biological sample measurements in include genotyping of targeted pharmacokinetic and pharmacodynamic SNPs (block 292), direct TAD-specific measurements including differential gene expression and pharmacogenomic contacts (block 294), and drug metabolomics (block 296). Genotyping of targeted pharmacokinetic and pharmacodynamic SNPs includes identifying mutations in the CYP2B6 gene that have been shown to determine metabolizer status (block 282), and classifying efficacy and adverse event subnetworks based on pharmacogenomic network and subnetwork profiling (block 284). Direct chromatin contact-specific measurements include differential gene expression determined using restricted RNA sequencing (RNA-Seq) (block 286), and restricted chromosome conformation capture assays for identifying pharmacogenomic contacts (block 288). In addition, drug metabolomics analysis is used to identify pre-existing drugs and metabolites in a patient's biological sample to assess, for example, potential drug-drug interactions (block 290), and can be used to routinely monitor patient drug-drug interactions and drug compliance.

[0095] Figure 3 shows a simple example of how an SNP located within an enhancer in the network can disrupt the contact between the enhancer or super-enhancer and one of its target gene promoters in a TAD, resulting in an adverse drug event in a patient within a drug response cohort. Figure 3 A shows how measurements can be obtained from a three-dimensional chromatin pharmacogenomic interactome using different laboratory methods and the data analyzed as a two-dimensional map of enhancer-gene promoter interactions. Figure 3 B depicts how an SNP can disrupt the chromatin loop between an enhancer and one of the two gene promoters it regulates within a TAD. This disruption eliminates the pharmacogenomic connection between the enhancer and gene promoter 1, resulting in dysregulation of gene 1 and leading to an adverse event in this patient and their cohort in response to administration of the specific drug of interest.

[0096] Figure 4 shows a first example among several examples for matching reference data with patient inputs for clinical decision support. In this example, a pharmacogenomic drug subnetwork reference set for a particular drug can be combined with a biological sample input of sparse results from a patient, and a combined co-training method can be used to derive a drug efficacy score for decision-making.

[0097] For example, as Figure 5 shown, the drug and dose decision server 102 compares the efficacy and adverse event subnetworks of the patient's mental drug of interest with the reference efficacy and adverse event subnetworks of the mental drug of interest. In this example, reference subnet 1 (reference number 502) and reference subnet 2 (reference number 504) are adverse event subnetworks, while reference subnet 3 (reference number 506) is an efficacy subnetwork. The subnetwork 510 of patient X is different from the adverse event subnetworks (reference subnet 1 and 2 (reference numbers 502, 504)) and is similar to the efficacy subnetwork (reference subnet 3 (reference number 506)). Therefore, the drug and dose decision server 102 determines that the mental drug of interest should be administered to patient 520. The subnetwork 512 of patient Y is different from one of the adverse event subnetworks (reference subnet 1 (reference number 502)) and is similar to the other adverse event subnetwork and the efficacy subnetwork (reference subnet 2 and 3 (reference numbers 504, 506)). Due to the adverse event in subnet 2 (reference number 504), the drug and dose decision server 102 determines that the mental drug of interest should be administered to the patient but at a reduced dose 522. The subnetwork 514 of patient Z is similar to the adverse event subnetworks (reference subnet 1 and 2 (reference numbers 502, 504)) and is different from the efficacy subnetwork (reference subnet 3 (reference number 506)), so the drug and dose decision server 102 determines not to administer the mental drug of interest to the patient 524.

[0098] Figure 5 provides an example of how to match the antidepressant efficacy reference set of a drug with the characteristics of an input patient biological sample relative to the adverse event characteristics from a gene set optimizer. In Figure 5 , ketamine can be administered to patient X based on the antidepressant efficacy of subnet 2 from the reference pharmacogenomic network matching the antidepressant efficacy the individual will experience from the pharmacogenomic network and the matching of the adverse event subnet 3 with the antidepressant efficacy the individual will experience from the pharmacogenomic network being reduced. However, in patient Z, since this individual carries some disease risk SNPs of GWAS found in the subnet of the pharmacogenomic network, the ketamine dose for patient Z should be adjusted.

[0099] Figure 6Shows how post - hoc bioinformatics analysis or disease - risk SNP annotation from GWAS can be performed and two different methods derived therefrom to determine that a patient should not receive ketamine based on the analysis of disease - gene risk variants and the presence of disease - risk SNPs associated with the adverse - event sub - network and efficacy sub - network of the pharmacogenomic drug network of ketamine. These simple and preliminary methods can be used for the initial screening of the potential negative consequences of ketamine administration to patients.

[0100] Biological samples 2002 and 2004 are collected from patient A and patient B. The biological sample 2002 of patient A is analyzed to perform pharmacodynamic SNP targeting as an input for pharmacogenomic network and sub - network profiling to determine the efficacy and adverse - event sub - network of patient A (box 2006). The biological sample 2004 of patient B is analyzed to identify pharmacokinetic SNPs associated with ketamine response (box 2008). Then the efficacy and adverse - event sub - network of patient A are compared with the reference pharmacogenomic network of ketamine and the reference efficacy and adverse - event sub - networks (box 2010). The SNPs of patient B are compared with the SNPs contained in the reference pharmacogenomic network of ketamine and the reference efficacy and adverse - event sub - networks (box 2012). In Figure 6 it, the dose of patient A must be adjusted before ketamine administration based on the match of the ketamine sub - network mediating adverse events with the antidepressant efficacy of the pharmacogenomic network that the individual will experience and the reduced match of the antidepressant efficacy sub - network of the reference pharmacogenomic network (box 2014). In addition, in patient B, since this individual carries many disease - risk SNPs from GWAS found in the sub - networks of the pharmacogenomic network, the dose of ketamine administered to patient B should be adjusted (box 2016).

[0101] Figure 7 Shows an integrated strategy in which a library spanning the phenotypic profiles of the human drug - response population stored in a reference database is used to make a preliminary "run or no - run" decision regarding whether ketamine should be administered to a specific patient and how other clinical data from a specific patient can be fine - tuned to make an informed clinical decision.

[0102] Figure 8 Shows a flowchart that uses well - known pattern - matching algorithms from computer - vision deep learning to match the drug efficacy and adverse events of a patient with the drug efficacy and adverse events of a reference drug pharmacogenomic network using a similarity score. Figure 8 Also shows how this strategy can be used to discover novel similar drugs.

[0103] Figure 9A Shows an overview of the integrated, multi - scale data analysis used in the system. Figure 9BA flowchart showing an exemplary method for generating a pharmacogenomics network for a drug for reconstruction and a corresponding subnetwork for the drug of interest, including a human pharmacogenomics SNP input filter, a pharmacogenomics network reconstruction engine, and an iterative gene set optimization engine.

[0104] As described above, the biological sample measurements include pharmacodynamic SNP targeting as input for pharmacogenomics network and subnetwork profiling, which analysis determines both efficacy and adverse events, as analyzed using a pharmacogenomics genomic classifier and a pharmacodynamic subnetwork profiling system. The patient's subnetwork for a particular drug of interest is then compared to a reference pharmacogenomics network and its constituent subnetworks for the drug of interest. Figure 9B Method 300 for identifying a reference pharmacogenomics network and constituent subnetworks for a particular drug of interest, such as ketamine, is shown. In some embodiments, the drug and dose decision server 102 executes method 300 to identify the pharmacogenomics network and constituent subnetworks for a particular drug of interest, and stores the pharmacogenomics network and constituent subnetworks in the reference pharmacogenomics network database 154. In other embodiments, method 300 is executed by another computing device and the output of the method is provided to the drug and dose decision server 102 and stored in the reference pharmacogenomics network database 154.

[0105] Select SNPs

[0106] In any case, at block 302, SNPs are obtained from human clinical studies that have been shown to be significantly associated with response to and adverse events of the drug of interest. Since the positions of SNPs associated with the studied traits have in most cases been inaccurately assigned to the nearest gene or nearby candidate genes in the published literature and GWAS based on the linear sequence assembled from the reference human genome, accurate mapping using imputation and annotation techniques is used to determine the actual positions of the reported SNPs.

[0107] The new study has several important implications for pharmacogenomics network identification. First, new drug target mechanisms can be identified by collecting pharmacogenomics network outputs in the training set using computer vision-based TAD matching, using deep learning (machine learning), and using correspondence with known drug-induced whole-genome TAD matrices for validation. Second, aggregating new drug target mechanisms in previously defined but not fully informed biological pathways will increase the likelihood of success. Third, the insights gained from determining drug targets from pharmacogenomics GWAS using three-dimensional (3D) genomic architecture will generate the next generation of candidate drugs and greatly improve the accuracy of pharmacogenomics clinical decision support diagnostics.

[0108] At block 304, pharmacodynamics, pharmacokinetics, and other SNPs are evaluated using a pharmacogenomics informatics pipeline. The pharmacogenomics informatics pipeline uses lead SNPs reported by GWAS and candidate gene studies to discover permissive candidate SNPs for genetic linkage using TAD boundaries rather than measures of linkage disequilibrium. These SNPs are evaluated using two independent workflows: an enhancer regulation workflow for regulatory SNPs and a coding SNP workflow. The enhancer regulation SNP workflow evaluates DNA methylation, transcription factor binding, histone marks, DNase I hypersensitivity, chromatin state, quantitative trait loci (QTL), and transcription factor binding site disruption of permissive candidate SNPs in disease-relevant tissues using tissue-specific omics datasets. The coding SNP workflow discovers common nonsynonymous coding SNPs within the pool of permissive candidate SNPs and then examines their histone modifications excluding exonic enhancer SNPs. Both sets of SNPs are mapped back to their TADs and host genes and screened for expression in relevant tissues. The final output SNPs are then evaluated using open-source machine learning algorithms to determine whether the SNPs are causal (block 306), and the causal variants are retained for further analysis in the workflow (block 308). Exonic SNPs are also evaluated as splice donors or splice acceptors using the Altrans algorithm. If they are found to be involved in alternative splicing, they are stored as such.

[0109] Query using fortuitous enhancer SNPs

[0110] At block 310, enhancer SNPs are used as probes to identify target genes located within the same TAD as the enhancer and to determine pharmacogenomic trans-interactions with other TADs using Hi-C chromosome conformation capture and ChIA-PET datasets generated from the cell types and tissues in which the drug of interest acts. If the TAD has strong boundaries as predicted by the number of CTCF bindings and a significant association with super-enhancers (block 312), genes containing other functional elements such as long non-coding RNAs are located within the same TAD, which is the target of enhancers that significantly alter the population drug response selected for the drug pharmacogenomics network. Among the top 3 statistically significant pharmacogenomic contacts of the first set of pharmacogenomic TADs in the same cell and / or tissue type in which the drug of interest acts, then an evaluation is performed, and if the genes in these "trans-TADs" are controlled by the same cell and / or tissue-specific enhancer in which the drug of interest acts, the genes are selected (block 316).

[0111] At block 318, the connectivity of the combined gene set is evaluated, where the combined gene set is selected from a first set of TADs carrying pharmacogenomic SNPs and genes selected from "trans TADs", including genes co-controlled with the first set of TAD genes. For example, third-party software such as Ingenuity Pathway Analysis TM , can be used to examine the connectivity of the combined gene set. Using Fisher's right-exact test, if there is significant connectivity within the combined gene set based on published literature, the genes are placed into a preliminary gene set of the pharmacogenomic network including the drug of interest. Any genes that do not form a connected network are discarded as non-candidates for the pharmacogenomic network (block 320).

[0112] Revision of the preliminary pharmacogenomic network of knowledge-based drug-specific interconnected genes

[0113] Then at block 322, each gene in this gene set including the preliminary drug pharmacogenomic network is subjected to manual, semi-automated or automated curation or a combination thereof to remove genes whose functions are not related to the drug of interest in the cell and / or tissue types in which they act or to add other genes that are not part of the preliminary set of this drug pharmacogenomic network. If it is determined that the gene is specifically affected by the drug of interest in the cell and / or tissue types in which they act, they should be added to the set. The interrogation step involves defining the function of individual genes, the phenotypic consequences of gene mutation damage, and the human cells and tissues expressing the genes to determine whether it becomes a candidate for membership in the pharmacogenomic network of the specific drug of interest.

[0114] In one embodiment, manual, semi-automated, or automated strategies can be used to combine the curation of each gene, its mutation profile, and its expression localization within human tissues to make these determinations. These are achieved through various web-based search tools, including gene definitions, genome browser annotations, GWAS catalogs, and other bioinformatics resources. For example, application programming interfaces (APIs) may have executables written in R, Python, PERL, or other programming languages to facilitate data access, data cleaning, and data analysis. This embodiment is an enhanced model of manual curation, but can become time-limited if there are many genes within the gene set of a pharmacogenomics network or a gene subset of a subnetwork, and especially if functional genomic elements may contain regulatory RNAs or functional RNAs, such as long non-coding RNAs, or if little is known about the function of a gene. Listing and analyzing the mutations of a given gene (±10Kb upstream and downstream) is the simplest of the three interrogation steps to perform because these databases are the most comprehensive. There are other resources for analyzing the tissue distribution of gene expression patterns. In cases where these patterns are compared to the sites where a particular drug of interest acts, results from imaging modalities can be analyzed, including those from radiological studies, optical microscopy analysis in pathology, or even more sophisticated methods. In some embodiments, this analysis is performed using machine learning techniques, such as neural networks.

[0115] In another embodiment, a Bayesian probability classifier based on machine learning or using Bayesian probability calculations can be used. Automated methods can be used to reduce the complexity of data analyzed from different data resources, where the functional knowledge profile of a gene, its mutation profile, and its tissue expression map are input into a learning machine that has been trained on many such instances and independently tested on another set of instances to determine accuracy. The predictive features selected by the trained neural network can be implemented on a support vector machine classifier to build a functional and mutation prediction model for the gene, where subsequent machine states determine the adequacy of the statistical fit of the pharmacogenomics network.

[0116] In some cases, machine learning suffers from overfitting, false positives, or false negatives in the output. In another embodiment, machine learning can be used in parallel to perform semi-automated and naive Bayesian classification to improve the accuracy of the final output.

[0117] Knowledge-based curation can be performed through the following steps. First, check the gene definition from multiple databases to see if it is specifically but not generally affected by the drug of interest. Additionally, when using, for example, Google Scholar TMAfter a thorough Internet search in and / or PubMed, the published literature (containing text strings with gene names or precursor gene names or equivalent protein names and any function related to the drug of interest) is evaluated. These may include binding affinity studies of molecules that can be repeatedly found, and the binding affinity is within 10 times of the binding affinity of the drug of interest to the same pharmacodynamic target. Secondly, the drug and dose decision server 102 examines all mutations of each gene, including SNPs, variable number tandem repeats, duplications, and all other known mutation alterations, extending linearly ±10 kb from the transcription start site and stop codon of the examined gene in a genomic browser (such as the UCSC Genome Browser or the Ensembl Genome Browser). If any of these mutations are found in the published literature or sources such as unpublished clinical trial data and they are related to the action of the drug of interest, including efficacy, adverse events, or first-pass metabolism, they are added to the preliminary gene set (box 324) including the pharmacogenomics network. Thirdly, especially for complex tissues such as the brain, skin, and cardiovascular system, the drug and dose decision server 102 qualitatively performs a concordance plot to compare the expression of all genes in this final set with the location where the drug of interest exerts its action, if known. Genes with expression that does not match the pharmacodynamic substrate of the drug of interest are discarded (box 324). Finally, third-party software such as Ingenuity PathwayAnalysis TM is used to examine the connectivity of this gene set (box 326). Using the Fisher's exact test, if the drug and dose decision server 102 determines that there is significant interconnectivity based on the published literature, they are placed into the preliminary gene set of the pharmacogenomics network including the drug of interest. Any genes that do not form a connected network are discarded as non-candidate genes for the drug pharmacogenomics network (box 328).

[0118] Iterative gene set optimization

[0119] As shown at box 330 and more detailedly in Figure 10 an iterative gene set optimization is performed on the identified candidate gene set in the pharmacogenomics network of the specific drug of interest. Figure 10The flowchart shows an example method 400 of iterative gene set optimization for deconstructing a pharmacogenomics network into subnetworks. Iterative gene set optimization can be performed to identify subnetworks of a pharmacogenomics network. More specifically, iterative gene set optimization involves using its API to convert all input molecular terms into genes or long non-coding RNA names, such as from the Human Gene Nomenclature Committee (HGNC) names (block 402). Iterative gene set optimization is different from gene set enrichment methods in that it not only combines multiple statistical methods, but also does not rank genes in a hierarchical manner as in threshold-dependent methods, and iterative gene set optimization does not rely on the comparison of experimental results, as in whole-distribution tests. Instead, iterative gene set optimization groups genes or long non-coding RNAs from a pharmacogenomics network (block 404) using the Jaccard distance to measure the similarity between two genes or long non-coding RNAs based on the dissimilarity of user-selected terms, where the Jaccard distance is the ratio of the size of the symmetric difference gene A Δ gene B = A ∩ B - A ∪ B to the union (block 406). This can be extended to clusters of related different gene names. The drug and dose decision server 102 then automatically sorts these sets or sorts them into subsets of functionally related gene clusters using a minimum entropy sorting algorithm, such as the COOLCAT algorithm, or using a user-defined number of clusters (block 408). After gene subset optimization using entropy minimization, manual curation can be employed to assign efficacy, adverse events, or functional mechanism subnetworks based on the known properties of the drug action mechanisms under consideration (blocks 410, 412).

[0120] Performing post hoc validation using third-party bioinformatics tools

[0121] To scientifically validate the deconstruction of a pharmacogenomics network into mechanism subnetworks based on functional gene subset optimization, each subnetwork of a pharmacogenomics network is post hoc evaluated against top gene ontology terms (molecular function and biological process), such as top canonical pathways, as determined using other proprietary or open-source pathway analysis software, such as disease risk gene variant analysis, as determined using other proprietary or open-source pathway analysis software, and upstream xenobiotic regulators are determined using different bioinformatics resources (block 332). Additionally, the GWAS catalogs of the European Bioinformatics Institute, the National Human Genome Research Institute, and the National Institutes of Health can be searched to discover significant SNP-trait associations for each gene in each subnetwork gene set. By providing examples of SNPs with statistical significance from GWAS, additional evidence can be provided that mutations in the genes contained in each subnetwork provide insights into the normal, unimpaired function of the subnetwork.

[0122] In some embodiments, after performing post hoc validation, as Figure 11As shown, the resulting pharmacogenomics network and the constituent sub-networks for the specific drugs of interest are stored, for example, in database 154, as Figure 1A shown.

[0123] For example, to map the causal SNPs that discretize ketamine response in the population, their target genes within their TADs, and the pharmacogenomics interactomes of these TADs, Hi-C chromosome conformation capture data can be used from publicly available datasets that are mapped to the A735 astrocyte cell line, H1 neuron cell line, SK-N-SH cell line, and postmortem human brain samples.

[0124] Figure 12 An example of a comparison of the results of eight different algorithms is shown, which tested the predicted causality of the GWAS SNP rs12967143-G (located in an intragenic enhancer within the TCF4 gene, a member of the ketamine pharmacogenomics efficacy sub-network) relative to other GWAS SNPs, as described by the numerical outputs of the machine learning algorithms used in the analysis (*p ≤ 0.05; **p ≤ 0.01; ANOVA).

[0125] Using H-GREEN, a user-adjustable binning software package, trans-TADs are mapped to the overlaps between different data sources. The top 3 trans-TAD pharmacogenomics interactions can be selected for each causal SNP TAD locus of origin. Using prior knowledge of ketamine as an anesthetic and analgesic, and in light of recent research and clinical trials on ketamine and other glutamate receptor modulators as antidepressants, the methods described herein can be used to score the top trans-TAD interactions, and the top 3 can be selected for each causal SNP to be included in the pharmacogenomics network. The recent availability of databases of validated enhancers and their targets can be used for the origin and targeted TADs in this workflow to reconstruct the ketamine pharmacogenomics network.

[0126] Intra-TAD and trans-TAD gene sets can be used as seeds to initiate pathway analysis. Filters and thresholds can be applied to eliminate genes expressed in the cell types, neurons and astrocytes, and brain regions where ketamine exerts its mechanism of action. These include not only PD genes but also PK genes, the latter of which have recently been shown to be expressed at high levels in relevant human brain regions where ketamine acts, and in the case of the CYP2B6 gene, are induced to higher expression levels by this psychotropic drug compared to the liver, gastrointestinal tract, or kidney.

[0127] After the output of automated pathway analysis, the rationality of the drug pharmacogenomics network gene set is evaluated, and genes not selected by the pathway analysis program can be added to the pathway. From early studies of binding affinity using molecular pharmacology methods, genes were added back whose products exhibited 10-fold affinity for the NMDAR for the racemate R,S-ketamine or the enantiomers, and molecular interconnectivity was demonstrated. Additional expression studies and investigations of ketamine metabolism may yield additional genes that are added back to the ketamine pharmacogenomics network.

[0128] The ketamine pharmacogenomics network was analyzed into 3 sub-networks by gene set optimization, where 2 sub-networks are subsets of genes and regulatory RNAs that are significantly different using iterative analysis. The 3 sub-networks include: (1) antidepressant efficacy and neuroplasticity, (2) glutamate receptor signaling, chromatin remodeling, and adverse events, and (3) pharmacokinetics and hormonal regulation associated with the drug. The second sub-network (2) glutamate receptor signaling, chromatin remodeling, and adverse events may contain two independent sub-networks: a chromatin remodeling sub-network and a drug pharmacodynamic adverse event sub-network. To understand and validate the pharmacogenomics network and its mechanistic sub-networks, four additional types of analysis were performed. First, the genes in the pharmacogenomics network and each sub-network were interrogated for the presence of enhancer SNPs associated with traits in GWAS. Second, pathway enrichment including biological processes and molecular functions was performed using gene ontology criteria to determine the most significant top pathways for these gene sets. Third, disease gene risk variant analysis was performed, which analyzed the importance of the entire mutational contribution of these sets in humans for each gene superset and subset to appropriately assign to both the parent pathway and its constituent sub-networks, and to assign top diseases (most importantly, Fisher's exact test) to the superset and sub-network sets. Fourth, the top (most important, Fisher's exact test) xenobiotic drugs regulating the superset of genes including the pharmacogenomics network were determined. In the last case, the pharmacogenomics network gene set should be regulated by drugs that mediate the pharmacogenomics network mechanism, but for some of the sub-networks within the network, drugs that are more relevant to a specific sub-network of the pharmacogenomics network mechanism may be most significantly associated, depending on the mechanistic properties of the network.

[0129] Figure 13 The top gene ontology terms for 2 significantly different ketamine pharmacogenomics sub-networks in the human brain are shown. More specifically, Figure 13 A shows the sub-network mediating efficacy and neuroplasticity. Figure 13 B shows the ketamine sub-network mediating glutamate receptor signaling and adverse events in the human CNS.

[0130] Figure 14AGraphical depiction of the ketamine pharmacogenomic subnetwork in the human brain mediating efficacy and neuroplasticity. Figure 14B Graphical depiction of the ketamine pharmacogenomic subnetwork in the human brain mediating glutamate receptor signaling and adverse events. More specifically, as Figure 14A and 15As shown, the genes and regulatory RNAs localized in the ketamine efficacy and neuroplasticity subnetworks include one or more of the following: activity-regulated cytoskeleton-associated protein (ARC) gene, achaete-scute family bHLH transcription factor 1 (ASCL1) gene, brain-derived neurotrophic factor (BDNF) gene, BDNF antisense RNA (BDNF-AS) gene, calcium / calmodulin-dependent protein kinase IIα (CAMK2A) gene, cyclin-dependent kinase inhibitor 1A (CDKN1A) gene, cAMP response element modulator (CREM) gene, cut like homeobox 2 (CUX2) gene, deleted in colorectal cancer (DCC) gene, dopamine receptor D2 (DRD2) gene, eukaryotic translation elongation factor 2 kinase (EEF2K) gene, fragile X mental retardation 1 (FMR1) gene, ganglioside-induced differentiation-associated protein 1-like 1 (GDAP1L1) gene, glutamate metabotropic receptor 5 (GRM5) gene, Homer scaffolding protein 1 (HOMER1) gene, serotonin receptor 1B (HTR1B) gene, serotonin receptor 2A (HTR2A) gene, Kruppel-like factor 6 (KLF6) gene, Lin-7 homolog C, component of the Par complex (LIN7C) long non-coding RNA, LOC105379109 long non-coding RNA, myocyte enhancer factor 2D (MEF2D) gene, myosin VI (MYO6) gene, myelin transcription factor 1-like (MYT1L) gene, neurogenic differentiation 1 (NEUROD1) gene, neurogenic differentiation 2 (NEUROD2) gene, nonsense helix-loop-helix 2 (NHLH2) gene, neuromedin B (NMB) gene, NMDA receptor synaptonuclear signaling and neuronal migration factor (NSMF) gene, neurotrophic receptor tyrosine kinase 2 (NTRK2) gene, phosphatase and tensin homolog (PTEN) gene, prostaglandin-endoperoxide synthase 2 (PTGS2) gene, Rac family small GTPase 1 (RAC1) gene, Ras protein-specific guanine nucleotide-releasing factor 2 (RASGRF2) gene, Ras homolog family member A (RHOA) gene, roundabout guidance receptor 2 (ROBO2) gene, RP11_360A181 long non-coding RNA, semaphorin 3A (SEMA3A) gene, SH3 and multiple ankyrin repeat domains 1 (SHANK1) gene, SH3 and multiple ankyrin repeat domains 2 (SHANK2) gene, SH3 and multiple ankyrin repeat domains 3 (SHANK3) gene, solute carrier family 22 member 15 (SLC22A15) gene, solute carrier family 6 member 2 (SLC6A2) gene, slit guidance ligand 1 (SLIT1) gene, slit guidance ligand 2 (SLIT2) gene, synaptosome-associated protein 25 (SNAP25) gene, synapsin I (SYN1) gene, synapsin II (SYN2) gene, synapsin III (SYN3) gene, T-box, brain 1 (TBR1) gene or transcription factor 4 (TCF4) gene.,

[0131] In addition, as Figure 14B and 16As shown, the genes and regulatory RNAs located in the ketamine glutamate receptor signaling and adverse event subnetwork include one or more of the following: acetylcholinesterase (ACHE) gene, activating transcription factor 7 interacting protein (ATF7IP) gene, activating transcription factor 7 interacting protein 2 (ATF7IP2) gene, ATPase Na+ / K+ transporting subunit alpha 1 (ATP1A1) gene, BLOC-1 related complex subunit 7 (BORCS7) gene, bromodomain containing 4 (BRD4) gene, calcium voltage-gated channel subunit alpha 1C (CACNA1C) gene, calcium voltage-gated channel auxiliary subunit beta 1 (CACNB1) gene, calcium voltage-gated channel auxiliary subunit beta 2 (CACNB2) gene, calcium voltage-gated channel auxiliary subunit gamma 2 (CACNG2) gene, cholinergic receptor muscarinic 2 (CHRM2) gene, cholinergic receptor nicotinic alpha 3 subunit (CHRNA3) gene, cholinergic receptor nicotinic alpha 5 subunit (CHRNA5) gene, cholinergic receptor nicotinic alpha 7 subunit (CHRNA7) gene, cannabinoid receptor 1 (CNR1) gene, disks large homolog 3 (Disks large homolog 3,Dlg3) gene, Discs, large homolog 4 (Dlg4) gene, DNA methyltransferase 1 (Dnmt1) gene, euchromatic histone lysine methyltransferase 1 (Ehmt1) gene, gamma-aminobutyric acid type A receptor subunit alpha2 (Gabra2) gene, gamma-aminobutyric acid type A receptor subunit alpha5 (Gabra5) gene, glutamate decarboxylase 1 (Gad1) gene, glycine receptor alpha1 (Glra1) gene, glycine receptor alpha2 (Glra2) gene, glycine receptor beta (Glrb) gene, glutamate ionotropic receptor AMPA type subunit 1 (Gria1) gene, glutamate ionotropic receptor AMPA type subunit 2 (Gria2) gene, glutamate ionotropic receptor AMPA type subunit 4 (Gria4) gene, glutamate ionotropic receptor NMDA type subunit 1 (Grin1) gene, glutamate ionotropic receptor NMDA type subunit 2A (Grin2A) gene, glutamate ionotropic receptor NMDA type subunit 2B (Grin2B) gene, glutamate ionotropic receptor NMDA type subunit 2C (Grin2C) gene, glutamate ionotropic receptor NMDA type subunit 2D (Grin2D) gene, glutamate ionotropic receptor NMDA type subunit 3A (Grin3A) gene, glutamate ionotropic receptor NMDA type subunit 3B (Grin3B) gene, hyperpolarization-activated cyclic nucleotide-gated potassium channel 1 (Hcn1) gene, histone deacetylase 5 (Hdac5) gene, methyl-CpG-binding domain protein 1 (Mbd1) gene, M-phase phosphoprotein 8 (Mphosph8) gene, neural cell adhesion molecule 1 (Ncam1) gene, nitric oxide synthase 1 (Nos1) gene, nitric oxide synthase 1 (Nos2) gene, nitric oxide synthase 3 (Nos3) gene, NAD(P)H quinone dehydrogenase 1 (Nqo1) gene, opioid receptor kappa1 (Oprk1) gene, opioid receptor mu1 (Oprm1) gene, roundabout guidance receptor 2 (Robo2) gene, SET domain bifurcated 1 (Setdb1) gene, SH3 and multiple ankyrin repeat domains 2 (Shank2) gene, sigma non-opioid intracellular receptor 1 (Sigmar1) gene, solute carrier family 6 member 9 (Slc6a9) gene, transcriptional activator silencer (Tasor) gene, axonemal microtubule TOG array regulator 2 (TOGarray regulator of axonemal microtubules 2, Togoram2) gene, tripartite motif-containing 28 (Trim28) gene or zinc finger protein 274 (Znf274) gene.,

[0132] Such as Figure 17As shown, the genes and regulatory RNAs localized in the pharmacokinetic enzyme and hormone sub-networks include one or more of the following: anaphase promoting complex subunit 2 (ANAPC2) gene, cytochrome P450 family 2 subfamily A member 6 (CYP2A6) gene, cytochrome P450 family 2 subfamily B member 6 (CYP2B6) gene, cytochrome P450 family 3 subfamily A member 4 (CYP3A4) gene, discs large homolog 4 (DLG4), eukaryotic elongation factor 2 kinase (EEF2K) gene, estrogen receptor 1 (ESR1) gene, glutamate ionotropic receptor AMPA type subunit 1 (GRIA1) gene, glutamate ionotropic receptor AMPA type subunit 4 (GRIA4) gene, glutamate ionotropic receptor NMDA type subunit 1 (GRIN1) gene, glutamate ionotropic receptor NMDA type subunit 2B (GRIN2B) gene, myosin VI (MYO6) gene, roundabout guidance receptor 2 (ROBO2) gene, SH3 and multiple ankyrin repeat domains 2 (SHANK2) gene, and transcriptional elongation regulator 1 (TCERG1) gene.

[0133] Automated iterative gene set optimization of the psychopharmacogenomics network into sub-networks may be subject to user limitations to study other features of the pharmacogenomics network. As referenced above Figure 10 As described, the ketamine pharmacogenomics network was iteratively deconstructed until certain specific genes were not significantly associated with any sub-network or were associated with several sub-networks deconstructed from the drug pharmacogenomics network. The gene ESR1 encodes the nuclear hormone receptor for estrogen and it regulates the expression of several genes in the ketamine pharmacogenomics network. It is well known that estrogen greatly induces the CYP2B6 gene in the human brain and elsewhere.

[0134] The learning architecture for training the pattern matching sub-network includes a pre-trained reference set (reference number 710). More specifically, at block 704, the drug and dose decision server 102 develops a patient pattern matching sub-network derived from the patient input biological sample, and jointly develops a separate training pattern metric (block 712) containing the features of the efficacy and adverse event sub-networks into a joint feature representation metric. To determine the similarity to the reference set (blocks 706, 708), two pairs of different reference-patient metrics include an accurate measurement of the similarity and an output similarity score for each of the efficacy and adverse event sub-networks (blocks 714, 716). At block 702, the biological sample obtained from the patient, which can be a cheek swab, saliva, blood, or urine sample, undergoes targeted enhancer SNP genotyping, as well as combined chromosome conformation capture and RNA-seq. Then at block 704, the drug and dose decision server 102 performs the necessary analysis to construct an input patient-specific map for the efficacy and adverse event sub-networks for the specific drug of interest. These patient-specific, drug-induced sub-network patterns can be further processed using Bayesian probability calculations to fill in sparse or missing data. When a new patient is input as input, the pre-trained reference set of the drug-specific efficacy and adverse event sub-networks for pattern matching is optimized again for subsequent patients, thus more accurately measuring the pharmacogenomic variability in humans, thereby enhancing clinical utility. This matching task assumes that the patches are encoded with the same features before calculating and outputting the similarity scores, greatly improving efficiency while reducing the computational requirements.

[0135] Thus, using probability calculations based on Bayesian distributions, each set of inputs (reference set (reference number 710) and patient set (reference number 720)) is constructed in different ways through feature set extraction and sparse data inference to increase the accuracy of the reference and patient maps. The trained feature network is based on the "Siamese" network method, where the constraint is that the two sets must share the same parameters. When completed, the patient's drug-induced training pattern network is coupled with those obtained from the reference database, paired efficacy feature sets, and adverse event feature sets. These provide the basis for developing a trained efficacy metric and a trained adverse event metric that attempt to match all the features from the patient and the reference set of the drug of interest. These paired matching scores produce separate efficacy and adverse event similarity scores between the reference and the patient.

[0136] In an additional embodiment, a reference pattern matching set can be developed for each patient, which can be used to create a patient-specific database of such reference maps and is updated regularly as additional biological samples are obtained from the patient longitudinally in a clinical setting or outpatient pharmacy over time.

[0137] In any case, the drug and dosage decision server 102 can then use the similarity scores of the efficacy and adverse event subnetworks of the psychotropic drug of interest generated by method 700 to determine whether to administer the psychotropic drug of interest to a patient. For example, the similarity score of the efficacy subnetwork can be compared with a threshold similarity score. If the similarity score of the efficacy subnetwork is higher than the threshold similarity score, the drug and dosage decision server 102 can determine that the psychotropic drug of interest should be administered to the patient. The similarity score of the adverse event subnetwork can also be compared with the threshold similarity score. If the similarity score of the adverse event subnetwork is lower than the threshold similarity score, the drug and dosage decision server 102 can determine that the psychotropic drug of interest should be administered to the patient. In another embodiment, the similarity scores of the efficacy subnetwork and the adverse event subnetwork can be combined or aggregated in any suitable manner. For example, the similarity score of the adverse event subnetwork can be subtracted from the similarity score of the efficacy subnetwork. If the combined score is greater than the threshold similarity score, the drug and dosage decision server 102 can determine that the psychotropic drug of interest should be administered to the patient.

[0138] In other ways, the drug and dosage decision server 102 can determine not to administer the psychotropic drug of interest and can provide advice to the client devices 106-116 of healthcare professionals to administer another drug for treating the patient's depression.

[0139] In some embodiments, the drug and dosage decision server 102 can determine whether to administer the psychotropic drug of interest to a patient by generating a machine learning model based on training data of the drug responses of patients previously prescribed the psychotropic drug of interest. The machine learning model can be generated based on several features of the previous patients, including the drug-induced subnetworks of the previous patients, PD and PK SNPs that stratify patients by drug response, neuroimaging data, direct TAD-specific measurements including differential gene expression, and clinical variables of the previous patients, such as age, weight, biological sex, body mass index, race, family history, patient substance abuse history, diagnostic codes, hospitalization history, drug-drug interactions, history of mental illness, whether the patient smokes or uses nicotine, and Hamilton Depression Rating Scale (HAM-D) scores. The drug and dosage decision server 102 can obtain the same features of the current patient, including molecular and clinical data, and apply the features of the current patient to the generated machine learning model to determine whether to administer the psychotropic drug of interest to the patient.

[0140] In addition to determining whether to administer the psychotropic drug of interest to a patient, the drug and dosage decision server 102 also determines the dosage to be administered to the patient. Figure 18A and 18BShows an analysis of an independent cohort dataset for developing a regression model for ketamine dose estimation. Published literature and other sources provided both pharmacokinetic SNP data and clinical values, which helped determine the dose based on CYP2B6 SNPs and clinical data. As can be seen from the differential analysis, the greatest contribution to the ketamine dose was the presence or absence of the poor metabolizer phenotype rs3745274, a variant that causes exon skipping and loss of first-pass metabolism of CYP2B6 ultra-induction of ketamine. The quantitative intranasal ketamine dose range in this model derivation cohort was from 0.4 mg / kg to 0.8 mg / kg. The drug and dose decision server 102 can generate a model based on published ketamine clinical trial results obtained from clinicaltrials.org. The following equation shows a general example based on the sum value of regression variables, which may be included in the ketamine dosing algorithm:

[0141] Dose = exp[2.00 × rs3745274 + 0.25 × Female (biological sex) + 0.22 × rs3786547 + 0.22 × Clopidogrel + 0.19 × rs11083595 + 0.11 × BSA + 0.20 × Smoking + 0.17 × History of suicide attempt + 0.07 × Age per decade + 0.09 × Non-Hispanic white race + 020 × Ticlopidine + 0.15 × Previous psychiatric hospitalization]

[0142] More specifically, the drug and dose decision server 102 can generate a dosing algorithm based on published literature and can have predetermined constants to apply to each of several patient characteristics, such as biological characteristics, demographic characteristics, clinical characteristics, etc., as in the equation above. In other embodiments, the drug and dose decision server 102 can use machine learning techniques to generate a dosing algorithm. For example, the drug and dose decision server 102 can collect dose information about patients for whom ketamine has been previously prescribed as training data. The dosing information can include the dose prescribed to each patient and an indication of whether the patient's dose was adjusted during treatment and / or whether the patient experienced an adverse event. The drug and dose decision server 102 can then analyze the training data to generate a machine learning model (e.g., neural network, decision tree, hyperplane, regression model, etc.) to determine the dose for a new patient based on the biological, demographic, and clinical characteristics of the new patient. Patient characteristics used in the dosing algorithm can include biological data, such as SNPs reported to stratify the response to ketamine in humans. Patient characteristics can also include the patient's demographic data, such as the patient's sex, height and weight, age, and race. In addition, patient characteristics can include clinical data, such as family history, drug-drug interactions, history of mental illness, whether the patient smokes or uses nicotine, and the Hamilton Depression Scale (HAM-D) score.

[0143] In any case, the drug and dose decision server 102 applies the patient's characteristics to the drug administration algorithm to determine the dose of ketamine to be administered to the patient. The drug and dose decision server 102 then provides the recommended dose to the client devices 106-116 of the healthcare professionals.

[0144] Although regression analysis optimized for patient-specific dosing cannot account for nearly half of the pharmacogenomic and clinical variables required for accuracy, few published studies have reported variables included for algorithmic determination of antidepressant selection and dose estimation. Clinical values obtained from medical records are also crucial for determining a reduction in ketamine administration, as Figure 13 shown. These include individuals with a body mass index (BMI) of over 30, a family history of alcohol use disorder (first degree), a history of suicide attempt, previous psychiatric hospitalization, female biological sex (premenopausal), non-Hispanic white race, and smoking. As indicated, these values contribute to potentially substantial ketamine administration but do not preclude the use of this medication.

[0145] In addition to comparing the subnetwork of the patient's psychotropic drug of interest with the reference subnetwork of the psychotropic drug of interest, the drug and dose decision server 102 also analyzes the patient's clinical data and neuroimaging data to determine whether to administer the drug to the patient. For example, the drug and dose decision server 102 can analyze the patient's HAMD score and / or the patient's symptoms to classify the patient into one of four TRD patient subtypes.

[0146] Figure 19 Four TRD patient subtypes determined by transcranial magnetic stimulation (TMS) coupled with neuroimaging of the resting connectivity network in the human brain are shown. In patients with TRD, subtypes have been identified and replicated using transcranial magnetic stimulation (TMS) and neuroimaging studies that discretize the subtypes by differentially activating and inhibiting the resting-state connectivity network in the human brain. TMS has four different anatomical locations on the outer side of the human head that can activate different neuroanatomical structures that are part of the limbic-cortical circuit and include some of the following: the resting-state connectivity network, the default mode network, the defensive response involving the amygdala, the reward circuit located in the basal forebrain that includes the nucleus accumbens (NA) or the orbitofrontal cortex (OFC) in depressive subtype 3, the sensory stimulus gating through the thalamus to the cortex, the cortical areas S1 and the insula, the inhibition of impulses involving the dorsolateral prefrontal cortex, and the inhibition of the activation of the limbic cortex amygdala by the dorsomedial prefrontal cortex, including memory consolidation in the entire anterior cingulate cortex and hippocampal formation.

[0147] Figure 19It is also shown that clinical data obtained from EHRs or other clinical records can be used to identify TRD subtypes from notes using structured values or using natural language processing.

[0148] Figure 20 Brain regions that are consistently activated during ketamine antidepressant response are shown, and how they differentially map to four TRD depressive subtypes. Although brain regions are involved, these are consistent with different depressive subtypes defined by, e.g., TMS and clinical values, and the results of neuroimaging studies that examined ketamine-induced activation and inhibition.

[0149] Figure 21 Based on defining the HAMD-17 rating (Hamilton Depression Scale), neuroimaging meta-analyses, associated clinical values as Figure 19 shown, and available information on the efficacy, indications, and recommendations of current available antidepressants and professional guidelines from the American Psychiatric Association, recommendations for switching medications from ketamine are provided for 4 different TRD subtypes.

[0150] Figure 21 Example drug recommendations and alternative medication choices are shown for each of the four different subtypes of TRD depression patients.

[0151] To identify which depressive subtypes should or should not be provided with ketamine and ketamine analogs, 24 publicly available neuroimaging datasets were analyzed to determine the neuroanatomical regions activated by ketamine and its analogs, depressive and TRD patients, and healthy controls. Since patients with TRD subtype 3 consistently exhibit extremely active subgenual anterior cingulate cortex (sgACC), dorsolateral and dorsomedial prefrontal (executive) cortex (dlPFC, dmPFC), and extremely active orbitofrontal cortex (OFC), it is recommended that these patients not receive ketamine drug therapy because this patient group will not respond or remit, but instead may experience exaggerated adverse psychiatric drug events. Independent analysis of the neuroanatomical localization of all genes found in the ketamine pharmacogenomics network showed that they are all highly expressed in the anterior cingulate gyrus, prefrontal cortex, supplementary motor cortex, orbitofrontal cortex, temporal cortex, amygdala, hippocampal formation, anterior caudate nucleus, and nucleus accumbens, but not in other cerebral cortical regions, hypothalamus, or brainstem. This is the same pattern as 24 functional neuroimaging studies that examined where ketamine first acts on the human brain to exert its antidepressant effect (Table 1).

[0152]

[0153]

[0154]

[0155]

[0156] Table 1. Functional neuroimaging studies have shown human CNS substrates in which ketamine and other NMDAR modulators act and in which all genes in the ketamine pharmacogenomics network are expressed at their highest levels.

[0157] Neuroimaging modality (MOD.):

[0158] 1H-MRS: 4-T 1H proton magnetic resonance spectroscopy

[0159] fMRI: Functional magnetic resonance imaging.

[0160] MEG: Magnetoencephalogram;

[0161] PET: FDG positron emission tomography;

[0162] phMRI; Pharmacological magnetic resonance imaging.

[0163] rs-fcMRI: Resting-state functional connectivity magnetic resonance imaging.

[0164] In another embodiment, disease risks and pharmacogenomic SNPs that distinguish two significantly different ketamine subnetworks in the human brain can be used to determine patient response and adverse events when treated with ketamine. Table 2A lists enhancer and super-enhancer SNPs found in the ketamine efficacy subnetwork, which can be used to determine the representation of mutations significantly associated with a valid response to ketamine. In contrast, Table 2B lists enhancer and super-enhancer SNPs found in the ketamine adverse event subnetwork, which can be used to determine the representation of mutations significantly associated with adverse CNS events in response to ketamine.

[0165]

[0166]

[0167]

[0168]

[0169]

[0170] Table 2A, Part 1. Enhancer and super-enhancer SNPs found in the ketamine efficacy subnetwork.

[0171]

[0172]

[0173]

[0174]

[0175]

[0176]

[0177]

[0178]

[0179]

[0180]

[0181]

[0182]

[0183]

[0184]

[0185]

[0186]

[0187]

[0188] Table 2A, part 2. Enhancers and super-enhancer SNPs associated with regulatory elements found in the ketamine efficacy subnetwork.

[0189]

[0190]

[0191]

[0192]

[0193] Table 2A, part 3. Chromatin interactions of enhancers and super-enhancer SNPs found in the ketamine efficacy subnetwork.

[0194]

[0195]

[0196]

[0197]

[0198]

[0199]

[0200] Table 2B, Part 1. Enhancer and super-enhancer SNPs found in the ketamine adverse event subnetwork.

[0201]

[0202]

[0203]

[0204]

[0205]

[0206]

[0207]

[0208]

[0209]

[0210]

[0211]

[0212]

[0213]

[0214]

[0215]

[0216]

[0217]

[0218]

[0219]

[0220] Table 2B, Part 2. Enhancer and super-enhancer SNPs associated with regulatory elements found in the ketamine adverse event subnetwork.

[0221]

[0222]

[0223]

[0224]

[0225]

[0226] Table 2B, Part 3. Chromatin interactions of enhancer and super-enhancer SNPs found in the ketamine adverse event subnetwork.

[0227] In another embodiment of the methods of the present disclosure, the generalization of this method can be used to reveal combinations of FDA-approved pharmaceuticals that can be used to enhance the treatment of neuropsychiatric disorders through their corresponding network or subnetwork mechanisms in biology. Figure 22 An example is shown of how the valproic acid pharmacogenomics network and the ketamine pharmacogenomics network work in a complementary manner to support neurogenesis. Valproic acid induces the transformation of neural progenitor cells into committed neural progenitor cells through the npBAF complex (1, top), and ketamine can act on committed neural progenitor cells through the human silencing complex (HUSH) to transform the progenitor cells into differentiated neurons (1, bottom).

[0228] Figure 23 Shows how Figure 22 the process shown in Figure 23 acts through the stepwise deacetylation of the histone 3 lysine 9 (H3K9) moiety ( Figure 23 A) caused by the valproic acid pharmacogenomics network and the acetylation of the H3K9 moiety after activation of the ketamine pharmacogenomics network (

[0229] Figure 24 Shows how the complementary pharmacogenomics networks of valproic acid and ketamine enable neural progenitor cells to become mature, differentiated neurons.

[0230] In another embodiment of the present disclosure, these methods can be used for other antidepressant drugs that target the NMDAR network. For example, other partial NMDAR antagonists, including AVP-786 and GLYX-13 (Rapastinel), are undergoing clinical trials as antidepressant drugs. In addition, GLRB blockers on NMDAR are also being developed as antidepressants, which include AV101 and D-cycloserine (Seromycin). Selective antagonists of GRIN2B of NMDAR are also being developed as antidepressants, such as those including EVT103, CP101, and MK-0657. Downstream of this pathway is AMPAR, and several antidepressants are being developed as agonists at GRIA1 and GRIA2, such as ORG 265576.

[0231] In another embodiment, these methods can be used to optimize the drug selection of other antidepressants, which provide greater power compared to commercially available pharmacogenomic clinical decision support assays that rely solely on coding SNPs to classify patients by drug. The methods covered by the techniques disclosed herein utilize the knowledge of the pharmacogenomic epigenome, including organizing it into TADs and TAD-TAD pharmacogenomic linkages, which provides in-depth insights into the mechanisms of CNS drugs. Additionally, these methods allow for the objective monitoring of drug-drug interactions and dosages, as well as the measurement of the parent drug and its metabolites from serum, such as the case of S-ketamine and its active metabolite norketamine, to provide additional insights into individual metabolizer subtypes.

[0232] Throughout the specification, multiple instances may implement components, operations, or structures described as a single instance. Although the separate operations of one or more methods are shown and described as separate operations, one or more of the separate operations may be performed simultaneously, and the operations need not be performed in the order shown. Structures and functions presented as separate components in an example configuration may be implemented as a combined structure or component. Similarly, structures and functions presented as a single component may be implemented as separate components. These and other variations, modifications, additions, and improvements fall within the scope of the subject matter herein.

[0233] Additionally, certain embodiments are described herein as including logic or a number of routines, sub-routines, applications, or instructions. This can constitute software (e.g., code embodied on a machine-readable medium or in a transmitted signal) or hardware. In hardware, a routine, etc. is a tangible unit capable of performing certain operations and can be configured or arranged in a certain manner. In an example embodiment, one or more computer systems (e.g., stand-alone client or server computer systems) or one or more hardware modules of a computer system (e.g., a processor or group of processors) can be configured by software (e.g., an application or part of an application) to operate as a hardware module that performs certain operations as described herein.

[0234] In various embodiments, the hardware module can be implemented mechanically or electronically. For example, the hardware module can include dedicated circuitry or logic that is permanently configured to perform certain operations (e.g., a dedicated processor such as a field programmable gate array (FPGA) or an application specific integrated circuit (ASIC)). The hardware module can also include programmable logic or circuitry that is temporarily configured by software to perform certain operations (e.g., as included in a general purpose processor or other programmable processor). It should be appreciated that the decision to implement the hardware module mechanically in dedicated and permanently configured circuitry or in temporarily configured circuitry (e.g., configured by software) may be driven by cost and time considerations.

[0235] Accordingly, the term "hardware module" should be understood to encompass a tangible entity that refers to an entity that is physically constructed, permanently configured (e.g., hard-wired) or temporarily configured (e.g., programmed) to operate in a certain manner or to perform certain operations as described herein. Considering embodiments in which the hardware module is temporarily configured (e.g., programmed), it is not necessary to configure or instantiate every hardware module at any given time. For example, in cases where the hardware module includes a general purpose processor that is configured by software, the general purpose processor can be configured to a corresponding different hardware module at different times. Thus, software can configure a processor, for example, to constitute a particular module at one time and a different module at a different time.

[0236] Hardware modules can provide information to other hardware modules or receive information from other hardware modules. Thus, the hardware modules can be considered to be communicatively coupled. In cases where multiple such hardware modules are present simultaneously, communication can be achieved through signal transmission that connects the hardware modules (e.g., via appropriate circuitry and buses). In embodiments where multiple hardware modules are configured or instantiated at different times, communication between such hardware modules can be achieved, for example, by storing and retrieving information in a memory structure accessible to the multiple hardware modules. For example, one hardware module can perform an operation and store the output of such operation in a memory device to which it is communicatively coupled. Then, another hardware module can access this memory device at a later time to retrieve and process the stored output. Hardware modules can also initiate communication with input or output devices and can operate on resources (e.g., a collection of information).

[0237] The various operations of the example methods described herein can be performed, at least in part, by one or more processors that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors can constitute processor-implemented modules that operate to perform one or more operations or functions. In some example embodiments, the modules referred to herein can include processor-implemented modules.

[0238] Similarly, the methods or routines described herein can be implemented, at least in part, by processors. For example, at least some of the operations of a method can be performed by one or more processors or processor-implemented hardware modules. The execution of some of the operations can be distributed among one or more processors that reside not only within a single machine but also across multiple machines. In some example embodiments, one or more processors can be located at a single location (e.g., in a home environment, in an office environment, or as a server farm), but in other embodiments, the processors can be distributed across multiple locations.

[0239] The execution of some of the operations can be distributed among one or more processors that reside not only within a single machine but also across multiple machines. In some exemplary embodiments, one or more processors or processor-implemented modules can be located at a single geographical location (e.g., in a home environment, an office environment, or a server farm). In other example embodiments, one or more processors or processor-implemented modules can be distributed across multiple geographical locations.

[0240] Unless otherwise explicitly stated, discussions using terms such as "processing", "computing", "caculating", "determining", "presenting", "displaying", etc. in this document can refer to actions or processes of a machine (e.g., a computer) to manipulate or transform data represented as physical (e.g., electrical, magnetic, or optical) quantities in one or more memories (e.g., volatile memory, non-volatile memory, or a combination thereof), registers, or other machine components that receive, store, transmit, or display information.

[0241] As used herein, any reference to "an embodiment" or "embodiments" means that the particular elements, features, structures, or characteristics described in connection with that embodiment are included in at least one embodiment. The phrase "in an embodiment" appearing in various places in the specification does not necessarily all refer to the same embodiment.

[0242] For example, some embodiments may use the term "coupled" to describe two or more elements in direct physical contact or electrical contact. However, the term "coupled" can also mean that two or more elements are not in direct physical contact with each other, but still cooperate or interact with each other. Embodiments are not limited in this context.

[0243] As used herein, the terms "comprises / comprising", "includes / including", "has / having", or any other variation thereof are intended to cover non-exclusive inclusion. For example, a process, method, article, or apparatus that includes a list of elements is not necessarily limited to those elements, but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Further, unless expressly stated to the contrary, "or" refers to an inclusive or rather than an exclusive or. For example, any of the following satisfies the condition A or B: A is true (or present) and B is false (or absent), A is false (or absent) and B is true (or present), and both A and B are true (or present).

[0244] In addition, "a / an" is used to describe elements and components of embodiments herein. This is for convenience only and provides a general description. This description and the appended claims should be understood to include one or at least one, and the singular also includes the plural unless clearly indicated otherwise. This detailed description should be construed as merely providing examples, and not describing every possible embodiment, as it would be impractical, if not impossible, to describe every possible embodiment. Many alternative embodiments can be implemented using current technology or technology developed after the filing date of this application.

Claims

1. A computing device for determining a drug to be administered to a patient suffering from a neuropsychiatric disorder, the computing device comprising: A communication network; One or more processors; And A non-transitory computer-readable memory coupled to the one or more processors and storing instructions thereon, the instructions when executed by the one or more processors cause the computing device to: Obtain data from a biological sample of the patient; For a drug selected from glutamate N-methyl-d-aspartate receptor (NMDAR) antagonists or partial antagonists, obtain a reference drug pharmacogenomics network representation from a reference database; Analyze the data from the biological sample based on the reference drug pharmacogenomics network representation; Determine a patient pharmacogenomics network representation for the drug based on the analysis; Determine a similarity score according to a comparison of the patient pharmacogenomics network representation for the drug with the reference drug pharmacogenomics network representation; Analyze the data from the biological sample using pharmacometabolomics analysis of drug pharmacokinetics to determine pre-existing drugs, metabolites, and doses of the pre-existing drugs in the patient's body; Determine the administration of the drug to the patient based on (i) the similarity score and (ii) a drug-gene or drug-drug interaction between the pre-existing drug in the patient's body and the drug; And Cause the drug to be administered to the patient.

2. The computing device according to claim 1, wherein the instructions further cause the computing device to: For the drug, obtain a constituent sub-network from the reference database, the constituent sub-network at least including a reference drug pharmacodynamic efficacy sub-network and a reference drug pharmacodynamic adverse event sub-network, wherein a first subset of variants of the reference drug pharmacodynamic adverse event sub-network is causally associated with adverse events of the drug, and a second subset of variants of the reference drug pharmacodynamic efficacy sub-network is causally associated with the efficacy of the drug, Wherein the patient pharmacogenomics network representation includes a patient drug pharmacodynamic efficacy sub-network and a patient drug pharmacodynamic adverse event sub-network, and wherein to determine the similarity score, the instructions cause the computing device to determine the similarity score according to at least one of the following: A comparison of the reference drug pharmacodynamic efficacy sub-network with the patient drug pharmacodynamic efficacy sub-network, or A comparison of the reference drug pharmacodynamic adverse event sub-network with the patient drug pharmacodynamic adverse event sub-network.

3. The computing device according to claim 2, wherein: The patient pharmacogenomic network representation for the drug includes a patient pharmacodynamic efficacy sub-network, a patient pharmacodynamic adverse event sub-network, a chromatin remodeling sub-network, and a pharmacokinetic enzyme and hormone sub-network, and the constituent sub-networks from the reference database for the drug include a reference pharmacodynamic efficacy sub-network, a reference pharmacodynamic adverse event sub-network, a reference chromatin remodeling sub-network, and a reference pharmacokinetic enzyme and hormone sub-network spanning human drug response variations, and To determine the similarity score, the instructions cause the computing device to: Assign a first score to the patient pharmacodynamic efficacy sub-network based on the amount of similarity between the patient pharmacodynamic efficacy sub-network and the reference pharmacodynamic efficacy sub-network; Assign a second score to the patient pharmacodynamic adverse event sub-network based on the amount of similarity between the patient pharmacodynamic adverse event sub-network and the reference pharmacodynamic adverse event sub-network.

4. The computing device according to claim 3, wherein to determine administration of the drug to the patient based on the similarity score, the instructions cause the computing device to: Determine administration of the drug to the patient when the first score is higher than a first threshold score or the second score is lower than a second threshold score.

5. The computing device according to claim 4, wherein to determine administration of the drug to the patient when the first score is higher than a first threshold score or the second score is lower than a second threshold score, the instructions cause the computing device to: Combine the first score and the second score to determine the similarity score; and Determine administration of the drug to the patient when the similarity score is higher than a third threshold score.

6. The computing device according to claim 1, wherein the instructions further cause the computing device to: Determine a dose of the drug to be administered to the patient based on the similarity score; and Cause the drug at the determined dose to be administered to the patient.

7. The computing device according to claim 6, wherein to determine a dose of the drug to be administered to the patient, the instructions cause the computing device to: Use a regression model to determine the dose based on a combination of two or more of: the gender of the patient, the age of the patient, whether the patient smokes, the race of the patient, the height of the patient, the weight of the patient, and the patient's history of mental illness.

8. The computing device according to claim 1, wherein the reference pharmacogenomic network representation for the drug obtained from the reference database is a ketamine pharmacogenomic network representation and includes one or more of the following: activity-regulated cytoskeleton-associated protein (ARC) gene, achaete-scute family bHLH transcription factor 1 (ASCL1) gene, brain-derived neurotrophic factor (BDNF) gene, BDNF antisense RNA (BDNF-AS) gene, calcium / calmodulin-dependent protein kinase IIα (CAMK2A) gene, cyclin-dependent kinase inhibitor 1A (CDKN1A) gene, cAMP response element modulator (CREM) gene, cut like homeobox 2 (CUX2) gene, deleted in colorectal carcinoma (DCC) gene, dopamine receptor D2 (DRD2) gene, fragile X mental retardation 1 (FMR1) gene, ganglioside-induced differentiation-associated protein 1-like 1 (GDAP1L1) gene, glutamate metabotropic receptor 5 (GRM5) gene, Homer scaffold protein 1 (HOMER1) gene, 5-hydroxytryptamine receptor 1B (HTR1B) gene, 5-hydroxytryptamine receptor 2A (HTR2A) gene, Kruppel-like factor 6 (KLF6) gene, Lin-7 homolog C, component of the crumbs cell polarity complex (LIN7C) long non-coding RNA, LOC105379109 long non-coding RNA, myocyte enhancer factor 2D (MEF2D) gene, myosin VI (MYO6) gene, myelin transcription factor 1-like (MYT1L) gene, neurogenic differentiation 1 (NEUROD1) gene, neurogenic differentiation 2 (NEUROD2) gene, nonsense helix-loop-helix 2 (NHLH2) gene, neuromedin B (NMB) gene, NMDA receptor synaptonuclear signaling and neuronal migration factor (NSMF) gene, neurotrophic receptor tyrosine kinase 2 (NTRK2) gene, phosphatase and tensin homolog (PTEN) gene, prostaglandin-endoperoxide synthase 2 (PTGS2) gene, Rac family small GTPase 1 (RAC1) gene, Ras protein-specific guanine nucleotide-releasing factor 2 (RASGRF2) gene, Ras homolog family member A (RHOA) gene, roundabout guidance receptor 2 (ROBO2) gene, RP11_360A181 long non-coding RNA, semaphorin 3A (SEMA3A) gene, SH3 and multiple ankyrin repeat domains 1 (SHANK1) gene, SH3 and multiple ankyrin repeat domains 2 (SHANK2) gene, SH3 and multiple ankyrin repeat domains 3 (SHANK3) gene, solute carrier family 22 member 15 (SLC22A15) gene, solute carrier family 6 member 2 (SLC6A2) gene, slit guidance ligand 1 (SLIT1) gene, slit guidance ligand 2 (SLIT2) gene, synaptosome-associated protein 25 (SNAP25) gene, synapsin I (SYN1) gene, synapsin II (SYN2) gene, synapsin III (SYN3) gene, T-box, brain 1 (TBR1) gene, transcription factor 4 (TCF4) gene, acetylcholinesterase (ACHE) gene, activating transcription factor 7 interacting protein (ATF7IP) gene, activating transcription factor 7 interacting protein 2 (ATF7IP2) gene, ATPase Na+ / K+ transporting subunit alpha 1 (ATP1A1) gene, BLOC-1 related complex subunit 7 (BORCS7) gene, bromodomain-containing 4 (BRD4) gene, calcium voltage-gated channel subunit alpha 1C (CACNA1C) gene, calcium voltage-gated channel auxiliary subunit beta 1 (CACNB1) gene, calcium voltage-gated channel auxiliary subunit beta 2 (CACNB2) gene, calcium voltage-gated channel auxiliary subunit gamma 2 (CACNG2) gene, cholinergic receptor muscarinic 2 (CHRM2) gene, cholinergic receptor nicotinic alpha 3 subunit (CHRNA3) gene, cholinergic receptor nicotinic alpha 5 subunit (CHRNA5) gene, cholinergic receptor nicotinic alpha 7 subunit (CHRNA7) gene, cannabinoid receptor 1 (CNR1) gene, discs large homolog 3 (Diskslargehomolog3,Dlg3) gene, discs large homolog 4 (Dlg4) gene, DNA methyltransferase 1 (Dnmt1) gene, euchromatic histone lysine methyltransferase 1 (Ehmt1) gene, gamma-aminobutyric acid type A receptor subunit alpha2 (Gabra2) gene, gamma-aminobutyric acid type A receptor subunit alpha5 (Gabra5) gene, glutamate decarboxylase 1 (Gad1) gene, glycine receptor alpha1 (Glra1) gene, glycine receptor alpha2 (Glra2) gene, glycine receptor beta (Glrb) gene, glutamate ionotropic receptor AMPA type subunit 1 (Gria1) gene, glutamate ionotropic receptor AMPA type subunit 2 (Gria2) gene, glutamate ionotropic receptor AMPA type subunit 4 (Gria4) gene, glutamate ionotropic receptor NMDA type subunit 1 (Grin1) gene, glutamate ionotropic receptor NMDA type subunit 2A (Grin2A) gene, glutamate ionotropic receptor NMDA type subunit 2B (Grin2B) gene, glutamate ionotropic receptor NMDA type subunit 2C (Grin2C) gene, glutamate ionotropic receptor NMDA type subunit 2D (Grin2D) gene, glutamate ionotropic receptor NMDA type subunit 3A (Grin3A) gene, glutamate ionotropic receptor NMDA type subunit 3B (Grin3B) gene, hyperpolarization-activated cyclic nucleotide-gated potassium channel 1 (Hcn1) gene, histone deacetylase 5 (Hdac5) gene, methyl-CpG-binding domain protein 1 (Mbd1) gene, M-phase phosphoprotein 8 (Mphosph8) gene, neural cell adhesion molecule 1 (Ncam1) gene, nitric oxide synthase 1 (Nos1) gene, nitric oxide synthase 2 (Nos2) gene, nitric oxide synthase 3 (Nos3) gene, NAD(P)H quinone dehydrogenase 1 (Nqo1) gene, opioid receptor kappa1 (Oprk1) gene, opioid receptor mu1 (Oprm1) gene, roundabout guidance receptor 2 (Robo2) gene, SET domain bifurcated 1 (Setdb1) gene, SH3 and multiple ankyrin repeat domains 2 (Shank2) gene, sigma non-opioid intracellular receptor 1 (Sigmar1) gene, solute carrier family 6 member 9 (Slc6a9) gene, transcriptional activator silencer (Tasor) gene, axonemal microtubule TOG array regulator 2 (TOG array regulator of axonemal microtubules 2,TOGORAM2) gene, tripartite motif-containing 28 (TRIM28) gene, zinc finger protein 274 (ZNF274) gene, anaphase promoting complex subunit 2 (ANAPC2) gene, cytochrome P450 family 2 subfamily A member 6 (CYP2A6) gene, cytochrome P450 family 2 subfamily B member 6 (CYP2B6) gene, cytochrome P450 family 3 subfamily A member 4 (CYP3A4) gene, eukaryotic elongation factor 2 kinase (EEF2K) gene, estrogen receptor 1 (ESR1) gene or transcriptional elongation regulator 1 (TCERG1) gene.,

Images