Patient digital twins and patient biomimetic twins for precision medicine

Abstract

Embodiments described herein provide a healthcare ecosystem of computing devices configured to collect different types of subject data (e.g., multimodal) and develop patient digital twins as predictive models that predict the experimental outcomes of subjects. The system can inform clinicians of subject attributes for preparing microphysiological systems called "patient biomimetic twins" (PBTs), which have cell-derived attributes or diseases that can be used to experimentally test the predictions generated by the predictive models of PDT.

Description

Technical Field

[0001] Explanation of government support The present invention was made with government support under grant numbers DK117881, TR003289, DK119973, and TR004124 awarded by the National Institutes of Health. The United States government has certain rights in the invention.

[0002] Cross - Reference to Related Applications This application claims the benefit of U.S. Provisional Application No. 63 / 464,106, filed May 4, 2023, which is hereby incorporated by reference in its entirety.

[0003] This application generally relates to developing and implementing patient biomimetic models and patient digital twins for predicting and validating proposed therapies.

Background Art

[0004] The FDA and other state and federal agencies impose a strict regulatory science - based framework to manage efforts to modernize, develop, evaluate, and deploy therapeutic agents (e.g., drugs) and treatment procedures for precision medicine. Conventionally, randomized controlled trials have been the gold standard for establishing the efficacy and safety of new chemical entities for decades. However, these trials are costly and often difficult to conduct, especially for diseases with low incidence rates. Furthermore, the high level of scientific research observed by agencies such as the FDA and universities tends to homogenize study designs and does not reflect the heterogeneity among patients in the clinical real - world, which is a common cause of many late - stage development failures and poor response rates to drug candidates.

[0005] Thus, the implementation of precision medicine, especially for complex and heterogeneous diseases, faces many barriers. These barriers include, for example, the selection of an optimal subject cohort for clinical trials for treatment, the approval of therapeutic agents by regulatory agencies such as the FDA, and the selection of patient - specific treatment strategies.

Summary of the Invention

[0006] This specification discloses systems and methods that can address the aforementioned shortcomings and may provide any number of additional or alternative benefits and advantages. Embodiments disclosed herein address the challenges of delivering precision medical treatments (e.g., drugs, therapies). Embodiments described herein provide a healthcare ecosystem of computing devices for hosting and implementing systems for therapeutic services. As described herein, computing devices integrate and extend various technologies to address shortcomings in the art, including: creating subject (e.g., individuals participating in a study, medical patients) study groups in a clinical setting and collecting different types (e.g., multimodal) subject data from various data sources; developing data-centric representations of patients, such as patient digital twins (PDTs), fed into machine learning models trained to predict patient-specific disease progression and responses to therapeutic drugs; developing patient-specific microphysiological systems called patient biomimetic twins (PBTs) of disease derived from patient cells, which can be used to experimentally test predictions made by machine learning models using PDTs; and (d) implementing scalable databases and analytical platforms designed to access, manage, selectively share, and computationally model patient-specific PDT and PBT data.

[0007] In one embodiment, a method performed by a computer includes: a step of obtaining subject data from one or more subject databases, the subject data comprising a plurality of types of subject data; a step of generating a predictive output by applying a predictive model of a patient digital twin to the subject data, the patient digital twin comprising a predictive model configured to produce a first predictive result of a trial configuration; a step of updating the parameters of the predictive model by updating the parameters of the predictive model according to user input; a step of generating a second predictive result by applying a patient digital twin to subject data and a trial configuration, the patient digital twin for a subject comprising biomimetic twin configuration data related to a patient biomimetic twin for a subject, the patient biomimetic twin comprising an experimental representation of a subject according to the biomimetic twin configuration data; a step of obtaining experimental result data and a trial configuration for a subject related to a patient biomimetic twin for a subject by updating the parameters of a patient digital twin by applying the patient digital twin model to subject data and experimental result data of a subject by updating the parameters of a patient digital twin. For example, a computer can update one or more parameters of a patient's digital twin by applying one or more types of computational models of the patient's digital twin (e.g., machine learning, mechanical, stochastic, and Bayesian networks) to the subject's data and the subject's experimental results data.

[0008] In another embodiment, the system comprises at least one processor, the processor configured to perform the following: a step of obtaining subject data relating to a subject from one or more subject databases, the subject data comprising a plurality of types of subject data; a step of generating a predictive output by applying a patient digital twin model to the subject data relating to a subject, the patient digital twin configured to generate a first predictive result for a trial configuration; a step of updating the model parameters according to user input; a step of generating a second predictive result by applying the patient digital twin to the subject data and the trial configuration, the patient digital twin relating to the subject includes biomimetic twin configuration data relating to the subject biomimetic twin relating to the subject, and the patient biomimetic twin includes an experimental representation of the subject according to the biomimetic twin configuration data; a step of obtaining experimental outcome data of the subject relating to the subject and the trial configuration, the patient digital twin being associated with the patient biomimetic twin; and a step of training the subject digital twin by updating the parameters of the subject digital twin by applying various computational models, including machine learning, mechanistic, probabilistic, and Bayesian networks, to the subject data and experimental outcome data relating to the subject.

[0009] In another embodiment, a method for developing and administering a precision therapy includes: acquiring subject data and a patient digital twin of a subject according to one or more therapy configurations suggesting a therapy; acquiring predicted therapeutic outcomes of the therapy according to a predictive model of the patient digital twin; constructing a patient biomimetic twin for the subject based on biomimetic twin configuration data indicated by the subject data of the subject; acquiring experimental result data by subjecting the patient biomimetic twin to therapy; inputting the experimental result data into a computer, the computer updating the predictive model of the patient digital twin based on the level of error between the experimental result data and the predicted therapeutic outcome; and administering the therapy to the subject.

[0010] In another embodiment, a method for identifying a subject cohort comprises: a step of obtaining subject data of a plurality of subjects, wherein the subject data of each subject includes one or more attributes of the subject; a step of obtaining a cohort of subjects for a treatment trial according to one or more treatment configurations, wherein a computer applies a cohort prediction engine to the subject data of the plurality of subjects using one or more treatment configurations; a step of constructing a patient biomimetic twin for each subject in the cohort based on biomimetic twin configuration data indicated by the subject data for the subject; a step of obtaining experimental result data related to the patient biomimetic twin of the subject from a laboratory test device by subjecting the patient biomimetic twin to treatment for a treatment trial; and a step of administering the treatment to the cohort of subjects.

[0011] Please understand that both the general description above and the detailed description below are illustrative and explanatory, and are intended to provide a further explanation of the claimed invention. [Brief explanation of the drawing]

[0012] This disclosure can be better understood by referring to the following drawings. The components in the drawings are not necessarily to scale, and the focus is on illustrating the principles of this disclosure. In the drawings, reference numbers indicate corresponding parts across different drawings.

[0013] [Figure 1] One embodiment illustrates the components of a system for therapeutic modeling that uses a machine learning architecture for processing various types of subject data.

[0014] [Figure 2] This describes the data flow between devices in a system for PBT development and experimentation using predictive PDT according to one embodiment.

[0015] [Figure 3]This describes the operation of a method for developing PDTs for predicting and using PBTs in experimental analysis, according to one embodiment.

[0016] [Figure 4] One embodiment illustrates the inter-device data flow of a system for developing and using predictive PDT and PBT for experiments. [Modes for carrying out the invention]

[0017] Hereinafter, the drawings and claims will be described using the specific language and with reference to the exemplary embodiments shown in the drawings and claims. It will be understood that no limitation of the scope of the invention is intended therein. Modifications and further alterations of the features of the invention described herein, and additional applications of the principles of the invention described herein, will be conceivable to those skilled in the relevant art and owners of this disclosure and will be considered within the scope of the invention.

[0018] This specification describes a system and method for developing a patient digital twin (PDT) as a computational model representing a patient or subject from different subject data sources. A PDT includes a predictive machine learning model trained to predict treatment outcomes based on the biocharacteristics of a subject or subject cohort. A PDT may include trained or tuned hyperparameters or weights in various layers of a machine learning architecture. The layers of the PDT's machine learning model can implement, among other things, any number of machine learning techniques, artificial intelligence techniques, and handcrafted machine techniques. A computing device can apply the PDT's machine learning architecture to subject data to predict the outcomes of a proposed treatment (e.g., drug, clinical treatment). It should be understood that the terms machine learning and artificial intelligence are intended to refer to similar data-driven inference capabilities or features and are not intended to be mutually exclusive. References to machine learning capabilities are not intended to exclude artificial intelligence capabilities.

[0019] The embodiment can implement various handcrafted models for generating and developing PDTs. The computing device develops the handcrafted model using prior knowledge and expertise of the therapeutic trial scenario to be modeled. The clinician user can submit or input user inputs to the user interface to calibrate various parameters or weights of the handcrafted model over time.

[0020] As an example, a handcrafted model can include a machine model. A machine model contains algorithmic functions to represent or model the interactions between different components of a clinical study. The purpose of a machine model is, among other advantages, to explain how components interact with each other and / or predict how components will behave under desired test conditions. While machine learning models or artificial intelligence models rely on statistical patterns in the data to generate predictions, unlike machine learning models or artificial intelligence models, machine models are based on fundamental principles and do not require large amounts of training data. Machine models can be used in clinical research, for example, to model the effect of a treatment (e.g., a drug or therapy) on a biological system, to predict the outcome of a clinical trial, and to optimize treatment plans for individual patients. Machine models can be implemented to model and study disease progression, and to identify and model new drug targets.

[0021] As another example, a computing device can implement a hand-drawn probabilistic model. The probabilistic model may be a hand-drawn model or a machine learning model. A hand-drawn probabilistic model is developed using prior knowledge and expertise of the therapeutic trial scenario being modeled, and typically involves specifying a set of probability distributions and algorithmic functions that govern the behavior of the interacting components. A machine learning probabilistic model is developed using a data-driven approach, where the computing device identifies statistical patterns and correlations used to output predictions that model a particular set of inputs in the therapeutic test scenario.

[0022] The computing device can also output information notifying clinicians about creating a patient bio - mimicking twin (PBT) as a physical test target called an "organ - on - chip" (OoC) corresponding to the PDT. The computing device can perform machine learning and artificial intelligence programming to derive causal inferences from experimental result data from experimental analyses (e.g., experiments) performed using the subject PBT. The use of a finely - tuned PBT enabled by continuously adjusted predictive PDTs can overcome the limitations of animal models and reduce dependence on animal models.

[0023] PBT (which may also be referred to as a micro - physiological system (MPS) or a biomedical model) serves as a subject experimental representation used to understand normal and abnormal functions and provide a basis for preventive or therapeutic interventions in human diseases. In some embodiments, the experimental representation may include 3D layered cells. The 3D layered cells are generated by a combination of continuous cell layering and inter - cellular self - organization of specific cell types. The cell types move and self - assemble into distinct layers forming tissue structures.

[0024] The combination of the computational model (PDT) and the subject experimental model (PBT) described herein provides an efficient and safer understanding, for example, of disease mechanisms, characterization of potential drug activity and toxicity (e.g., dosing prediction and mechanism of action), and identification of potential clinical biomarkers in various levels of people (e.g., population level, specific cohort subgroups, or individual subjects). Embodiments further enable predicting and testing drugs for the effects of chronic exposure and for drugs on specific cohort populations, which cannot be achieved (in the real world) before administering to subjects.

[0025] As used herein, the terms "individual," "subject," and "patient" are used interchangeably herein and refer to any individual mammal being treated according to the disclosed methods or uses, e.g., cows, dogs, cats, horses, monkeys, pigs, camels, bats, or humans. In a preferred embodiment, the subject is a human. As used herein, the term "treatment" or "treating" refers to reducing or eliminating a disease and / or ameliorating or alleviating one or more symptoms of the disease.

[0026] Examples of system components

[0027] FIG. 1 shows components of a system 100 for treatment modeling using a machine learning architecture for processing various types of subject data, according to an exemplary embodiment. System 100 includes a treatment analysis system 101 having an analysis server 102 and an analysis database 104 (sometimes referred to as a "federated database"), subject databases 106a-106n (generally referred to as subject database 106), conditional subject databases 108a-108n (generally referred to as conditional database 108), and a user device 114. One or more networks 112 interconnect the components of system 100 and enable the devices to communicate with each other.

[0028] Embodiments may include additional or alternative components or may omit certain components from the components of FIG. 1 and still be within the scope of the present disclosure. For example, it may be common to include multiple analysis servers 102. Embodiments may include or be implemented with any number of devices capable of performing the various features and tasks described herein. For example, FIG. 1 shows analysis server 102 as a computing device separate from analysis database 104. In some embodiments, analysis database 104 includes integrated analysis server 102.

[0029] Therapeutic modeling services may host, implement, operate, or otherwise supply the analysis system 101 to provide various benefits related to the therapeutic modeling functions and features described herein. Non-limiting examples of therapeutic modeling include prediction and evaluation of drug efficacy, ADME-Tox, clinical trials, and therapeutic strategies. The analysis system 101 can perform PDT to evaluate subject attributes and predict subject-specific mechanisms of biological response to a trial configuration, such as predicting disease progression and / or therapeutic strategies. As one example, the analysis system 101 can predict a cohort for testing a drug or treatment using a PDT representing subjects. As another example, the analysis system 101 can use a PDT on a specific subject (such as a patient) to predict the efficacy of a drug or treatment for that subject.

[0030] A clinician interacting with the analysis system 101 using a user device 114 can refer to a PBT that logically corresponds to a PDT. The PBT contains physical test biological material created by the clinician using samples extracted by the clinician (or other actor) from a subject represented by the corresponding PDT. The clinician prepares and conducts experiments on the PBT to test or validate predictions (e.g., predicted treatment strategies) generated by the PDT. In some cases, the PDT may be manually optimized according to user input by manipulating the user device 114 to submit updates to the PDT. As an addition or alternative, in some cases, the PDT may be algorithmically optimized by the analysis server 102 according to various machine learning techniques implemented by the PDT and / or the analysis server 102 using experimental result data.

[0031] System 100 includes one or more networks 112, which may include any number of internal networks, external networks, private networks (e.g., intranets, VPNs), and public networks (e.g., the Internet). Network 112 comprises various hardware and software components for hosting and performing communications between components of System 100. Non-limiting examples of such internal or external networks may include local area networks (LANs), wireless local area networks (WLANs), metropolitan area networks (MANs), wide area networks (WANs), and the Internet. Communications over network 112 may be performed according to various communication protocols, among others, such as the Transmission Control Protocol and Internet Protocol (TCP / IP), User Datagram Protocol (UDP), and IEEE Communication Protocol.

[0032] In some cases, the analysis system 101 may include a computing network infrastructure comprising physically and logically related software and electronic devices managed or operated by, for example, a therapeutic modeling service provider, wherein the devices of the infrastructure 101 are configured to provide the intended therapeutic modeling services. The analysis system 101 may include an internal network (not shown) comprising networking hardware and software components for hosting and carrying out communications between components of the analysis system 101, including communications between the analysis server 102 and the analysis database 104.

[0033] User device 114 enables a user (e.g., a clinician, healthcare provider, or researcher) to interact with the therapeutic modeling services of the analysis system 101. User device 114 may include any computing device comprising hardware (e.g., a non-temporary machine-readable storage medium, a processor) and machine hardware execution software components capable of performing the processes and tasks described herein. Non-limiting examples of user device 114 may include personal computers (PCs) (workstation computers, laptops), tablets, and smartphones, among other types of electronic devices capable of performing the functions of user device 114 described herein.

[0034] The user device 114 includes or is coupled to peripheral devices for receiving user input, such as user I / O devices (e.g., keyboard, mouse, monitor), enabling the user to interact with the user device 114 and the analysis system 101 via the network 112. As an example, the user may input user input indicating experimental test configuration, experimental test results, PDT configuration, and PDT updates, among other types of input. In some cases, peripheral devices may include instruments that acquire (e.g., receive, generate) experimental result data when performing experimental analysis on the PBT and transmit the experimental result data to the user device 114 or the analysis system 102, either connected to or wirelessly. Non-limiting examples of such instruments include image readers, microscopes, live cell image readers, sequencing systems, thermal cyclers, and gel and blot readers. Additionally or alternatively, the user may input experimental result data into the user interface of the user device 114 and then transmit the experimental result data to the analysis system 101 via the network 112.

[0035] The user device 114 executes various software programs for accessing the analysis system 101 via one or more networks 112, enabling the user to provide user input to the analysis system 101 and receive outputs and data returned from the analysis system 101. In some implementations, the user device 114 executes locally installed software associated with the analysis system 101 to access and interact with the services of the analysis system 101 and to perform various functions and features described herein. In some implementations, the user device 114 executes web browser programming to access a website or web application hosted by a web server program executed by the analysis server 102 of the analysis system 101. The user can interact with the services of the analysis system 101 and operate the web browser as a user interface for performing various functions and features described herein.

[0036] The user can operate the user device 114 to submit requests to the analysis server 102 for performing therapeutic services, such as requests for predicted cohorts or requests for predicted therapeutic effects. Requests include machine-readable instructions that instruct the analysis server 102 to perform the requested activity, such as determining a cohort of subjects to test a proposed treatment or determining therapeutic outcomes for a proposed treatment, among other potential methods. The user device 114 receives user input indicating requests for specific processes and various configuration inputs associated with specific requests, and transmits the requests to the analysis system 101 via the network 112.

[0037] System 100 includes any number of subject databases 106 containing various types of subject data. The subject databases 106 may be hosted on one or more computing devices equipped with hardware components (e.g., non-temporary machine-readable storage media, processors) and software components (e.g., database management systems (DBMS)) capable of performing the various tasks and processes described herein. Similarly, a computing device may host one or more subject databases 106. In some cases, a particular subject database 106 may be a component of the analysis system 101. Additionally or alternatively, in some cases, a particular subject database 106 may be external to the analysis system 101 and hosted by a separate enterprise infrastructure network, thereby allowing the analysis server 102 (or other components of the analysis system 101) to retrieve subject data from the subject databases 106 via the network 112.

[0038] The subject database 106 contains various types of data used to generate or update PDTs. Non-exclusive examples of subject database 106 include clinical record databases 106a, omics data 106b, MRI images 106c, and mobile health data 106d.

[0039] Clinical database 106a may include demographic data showing the subject's age, sex, race, and ethnicity. Clinical database 106a may include medical history and comorbidity data showing the subject's disease and surgical history. Clinical database 106a may include body composition data showing the subject's height, weight, and body fat percentage. Clinical database 106a may include drug data showing the subject's current and past medications. Clinical database 106a may include biochemical data showing the subject's blood chemistry characteristics. Clinical database 106a may include patient-reported outcome measures showing the subject's self-reported depression, anxiety, and social functioning. Clinical database 106a may include non-invasive liver fibrosis staging showing the subject's liver stiffness measurements and controlled decay parameters. Clinical database 106a may include liver biopsy data showing the subject's steatosis grade score, inflammation score, ballooning score, and fibrosis stage. Clinical database 106a may include saliva and stool data showing the subject's taxonomic classification and species richness. Clinical database 106a may include plasma and serum data showing the subject's metabolite and proteomics profiles. Clinical database 106a may also include genetic data showing the subject's exome sequencing profile. A non-limiting example of data in clinical database 106a (or other subject database 106) is shown and described in Table 1. [Table 1]

[0040] The omics database 106b may include genomic data showing disease-causing DNA changes, transcriptome data showing subject RNA regulatory profiles, metabolome and lipidome data showing subject metabolite levels, metabolic and inflammatory proteome data showing subject serum inflammatory and metabolic protein levels, and spatial metabolomics showing subject metabolite counts per spot data. Non-limiting examples of the types of data stored in the omics database 106b (or other subject database 106) are shown and described in Table 2. [Table 2]

[0041] The MRI image database 106c may include, for example, radiomic data to show transverse abdominal image data indicating the composition of the subject's liver and body, the subject's liver fat percentage, liver stiffness score, the presence of portal hypertension, or the presence of cirrhosis.

[0042] A mobile health database 106d may include patient-reported medical data from portable devices such as mobile phones, tablets, or laptop computers. Non-limiting examples include subjects' vital signs, weight loss, sleep patterns, and physical activity.

[0043] In some embodiments, as in the exemplary system 100, the analysis system 101 includes an analysis database 104 that functions as a federated database storing multimodal subject data from multiple subject databases 106. The analysis server 102 can receive subject data obtained from various subject databases 106 and store it in the analysis database 104. For the sake of clarity, the system 100 includes the analysis database 104 from which the analysis server 102 retrieves subject data to perform the various processes described herein. However, embodiments are not required to include such a central or federated database.

[0044] Optionally, system 100 includes one or more conditional databases 108. The conditional databases 108 include conditional data, including pre-processed or "cleaned" instances of subject data from subject database 106. The analysis server 102 or other computing device may perform any number of data adjustment methods, including pre-processing functions to clean, normalize, and format subject data from subject database 106 for downstream operation. Non-limiting examples include data completion functions, data noise reduction functions, data transformation functions, and data normalization functions. The output of such data adjustment methods applied to subject data may be stored in one or more adjustment databases 108 and / or analysis database 104. While the exemplary system 100 includes a conditional database 108 corresponding to subject database 106, embodiments are not required to include such correspondences.

[0045] In some embodiments, the analysis server 102 or other computing device can perform a data profiling function in the preprocessing function. The data profiling function generates subject data profiles of subjects. The analysis server 102 can store the data profiles in, for example, the analysis database 104, the conditional database 108, the subject database 106, the non-temporary memory of the user device 114, or other non-temporary storage media of the system 100.

[0046] In the exemplary system 100, the analysis database 104 may function as a federated database storing multi-model subject data from the subject database 106 and / or conditional database 108. The analysis server 102 can receive subject data retrieved from various conditional databases 108 and store it in the analysis database 104. For ease of explanation, system 100 includes the analysis database 104 from which the analysis server 102 retrieves subject data from the subject database 106 and conditional database 108 to perform the various processes described herein.

[0047] As further described below, the analysis server 102 can collect subject data from the subject database 106 and / or conditional database 108 and store the subject data in the analysis database 104. The analysis database 104 can use various types of subject data to train one or more PDTs for subjects (and any number of arbitrary DTs). In some cases, the analysis server 102 can automatically retrieve or receive subject data from the subject database 106 and / or conditional database 108 via the network 112 at pre-configured intervals or when one or more subject databases 106 or conditional databases 108 are updated. In some implementations, a user, subject, or another actor can manually upload or submit subject data from the subject database 106 or conditional database 108 to the analysis system 101 via the network 112. A computing device hosting the analysis database 104 may store the received subject data in the analysis database 104. The analysis server 102 can, for example, host an online precision medicine portal that allows users to upload or submit subject data from the subject database 106 or the conditional database 108 to the analysis system 101.

[0048] The analysis server 102 of the analysis system 101 is any computing device comprising one or more processors and software, capable of performing the various processes and tasks described herein. The analysis server 102 may host or communicate with the analysis database 104, and among other potential components of the system 100, the analysis server 102 receives and processes subject data, experimental results data, experimental configurations, and other types of inputs received from the subject database 106, the conditional database 108, and the user device 114. Although Figure 1 shows only a single analysis server 102, the analysis server 102 may include any number of computing devices. In some cases, the computing devices of the analysis server 102 may perform all or some of the methods and benefits of the analysis server 102. The analysis server 102 may comprise computing devices operating in a distributed or cloud computing configuration and / or virtual machine configuration.

[0049] The analysis server 102 can perform software programming for computational software modules (sometimes referred to as “computational modules”) that generate, develop, and / or run predictive models. As an example, predictive models generated, trained, or otherwise developed by a computational module may include PDTs for predicting diagnoses, prognoses, treatment strategies, drug responses, and clinical biomarkers. The computational module can generate, develop, and / or run PDTs using subject data in the analysis database 104. As another example, predictive models generated, trained, or otherwise developed by a computational module may include a cohort prediction engine for predicting selected cohorts for clinical trials.

[0050] The calculation module (or other software programming of the analysis server 102) can generate or develop PBTs using PBT-related data in the subject data within the analysis database 104.

[0051] The Computation Module can execute various layers of machine learning architectures, among other functions. It can implement a variety of machine learning and artificial intelligence algorithms, taking in preprocessed data and integrating it into various predictive models (e.g., PBT). The Computation Module can also take in PBT-related data, such as experimental results, to develop or update predictive models. Furthermore, the Computation Module can use hand-drawn models to generate or further develop predictive models. In some implementations, the Computation Module uses and fuses both machine learning models and hand-drawn models to generate and develop specific predictive models. Non-limiting examples of models or techniques may include machine models, stochastic models, and Bayesian networks, among other possible types of machine learning or hand-drawn models.

[0052] The analysis server 102 can perform software programming of a layer of machine learning architecture that provides the functionality of a biomarker discovery module (sometimes referred to as the “biomarker module”). The analysis server 102 can train the biomarker module to identify potential clinical biomarkers from PBT-related data, such as subject data and / or experimental outcome data, from any number of subjects. In this way, the biomarker module can identify the biological attributes of subjects in the analysis database 104 for creating PBTs. Additionally or alternatively, the biomarker module (or other machine learning models of the analysis server 102) may identify distinguishing features for training and applying the various types of machine learning models described herein.

[0053] As described above, the analysis server 102 can perform software programming of layers of machine learning architecture that provide the functionality and features of PDT. The computation module can, for example, train one or more PDTs for a given subject to generate outputs using a specific portion of the subject's subject data. Each PDT is a predictive model trained to generate outputs that predict the subject's response to a given stimulus or test scenario. As an example, the analysis server 102 can train a PDT to predict possible outcomes of administering a proposed drug or other therapeutic agent to a subject. During training or retraining, the analysis server 102 can apply layers of machine learning architecture that define the PDT on drug-related data and subject data to output predicted therapeutic outcomes. In some cases, the PDT applies a loss function to predicted therapeutic outcomes and labeled data indicating predicted therapeutic outcomes to identify the level of error. The loss function can adjust or tune the hyperparameters or weights of the PDT's predictive model to reduce the level of error. In some cases, the PDT applies a loss function to experimental outcome data received from a user device 114, where the experimental outcome data is based on a clinician user performing experimental analysis using PBT. The analysis server 102 can receive experimental result data when it is input to the user device 114 by a clinician, when it is generated by a peripheral device connected to the user device 114, or when it is communicating with the analysis server 102.

[0054] As an addition or alternative, in some implementations, the analysis server 102 may perform software programming of a hand-drawn model that provides the functionality and features of the PDT. Using the user device 114, the clinician can input configuration instructions that define the hand-drawn model executed by the computation module. Over time, the clinician may continuously calibrate the parameters or weights of the hand-drawn model to reduce the level of error or to improve the predictive accuracy and / or consistency of the PDT hand-drawn model. The clinician may, for example, continue to calibrate the PDT for a given subject to generate predictive outputs using all or specific portions of the subject's subject data. The PDT is a predictive model calibrated to produce outputs that predict a subject's response to a given stimulus or test scenario.

[0055] In some implementations, the analysis server 102 can generate and develop PDTs according to handcrafted models and machine learning models. As an example, a clinician can initialize the programming of a PDT by instructing the PDT to apply a handcrafted model to subject data and treatment data (e.g., drug-related data) for subjects, according to a trial configuration that shows a specific trial scenario for a given treatment. The handcrafted model may also be applied to PBT-related data (e.g., experimental results data). The user can continuously refine and calibrate the handcrafted model for a predetermined number of runs of experimental analysis and / or for a predetermined period of time. After certain conditions, the clinician or the analysis server 102 can activate a layer of machine learning architecture that implements a machine learning model to apply a loss function and adjust the parameters, hyperparameters, or weights of the PDT.

[0056] The analysis server 102 can run a softer programming layer of a machine learning architecture, including a machine learning model that functions as a cohort prediction engine or cohort agency. The cohort agency layer may be trained, for example, to identify or predict cohorts of one or more subjects based on subject data from the subject database 106, treatment trial configurations, and / or PBT-related data. In this way, the analysis system 102 can improve the potential success for accurate medical outcomes, drug discovery and development outcomes, and clinical treatment outcomes for individual subjects. In some implementations, for example, the analysis server 102 may apply subject PDT to trial configurations indicated by a user request for a cohort or a request to test a proposed therapeutic agent (e.g., a proposed drug). Using the PDT predictive model, the analysis server 102 can predict cohorts of subjects with subject data that show subjects having common key genetic, lifestyle, and environmental attributes. Predictions are performed for specific trial configurations such as safety requirements, drug efficacy requirements, and drug candidate efficacy requirements.

[0057] In some implementations, the analysis server 102 can iteratively train layers of machine learning architecture about the machine learning architecture of the experiment selector. By iteratively training the experiment selector, the analysis server 102 can iteratively select experimental analysis constraints (e.g., PBT experiments) to efficiently build a comprehensive map of drug efficacy within the user's cohort. Rather than testing all possible combinations of drug and PDT models, the analysis server 102 (e.g., the computation module) can iteratively refine the PBT and PDT models. The analysis server 102 can be trained to optimally prioritize PBT experiments that make the greatest contribution to understanding the safety and efficacy of the drug. The analysis server 102 and / or the clinician can perform experimental analysis (or experiments) and select or present design information (as a trial configuration reflecting the experimental design) for creating PBTs corresponding to PDTs.

[0058] Following an experiment, or at predetermined intervals, the analysis server 102 retrieves experimental result data from the PBT and stores it as PBT-related data in the analysis database 104 or other databases 106, 108. The computing module can retrain or calibrate one or more predictive models of the analysis system 101. For example, the analysis server 102 and the computing module can retrieve experimental result data, which can be fed back into the PDT model for the computing module to retrain the PDT model, which may include applying one or more types of computing models, including, among others, machine learning, mechanical, stochastic, and Bayesian networks. The improved PDT is then applied to subsequent rounds of experimental design. This active learning approach can beneficially generate improved PDT models and PBT biomedical models. For example, the loss layer of the PDT machine learning architecture (or other types of computing models performed by the analysis server 102) includes a loss function. The loss function determines the error level of the computational model by determining the distance between the predicted output (predicted experimental results, predicted cohort, etc.) and the corresponding observations generated by the PDT model (machine learning, mechanistic, stochastic, Bayesian network, etc.), as indicated by the experimental results data. The loss function can adjust the hyperparameters or weights of the PDT model based on the error level, and the analysis server 102 can retrain a particular PDT model by applying the PDT model to subject data and level error, among other potential inputs.

[0059] Example of a PBT experimental platform

[0060] Figure 2 shows the data flow between devices of a system 200 for PBT development and experimentation using predictive PDT according to one embodiment. System 200 includes an analysis system 201 having an analysis server 202 and an analysis database 204 for developing PBT, among other predictive models. System 200 further includes a perfusion module 208 that houses any number of microfluidic chips 210 (sometimes called “PBT chips” or “tissue chips”). The microfluidic chips 210 include any number of chambers 206a–206c (commonly referred to as chambers 206). System 200 further includes a cell incubator 212 that houses the perfusion module 208 and an inline imaging reader 214 coupled to the cell incubator 212 via a perfusion controller. Furthermore, Figure 2 illustrates, for example, the data flow for constructing an experimental analysis (or experiment) using predictive models (e.g., PDT, cohort predictor, experiment selector) and other types of data, the execution of the experiment using the PBT and the PBT created according to the experimental configuration, and the capture of experimental result data from executing the experiment. The analysis system 201 is used on the front end of the data flow to design experiments, such as developing and running predictive models for selecting PBT and other relevant information such as compound and subject data.

[0061] (A) The analysis server 202 retrieves and runs a predictive model from the analysis database 204 that includes the subjects' PDT and subject data. In some implementations, the analysis server 202 runs a cohort prediction model to predict and select a cohort of subjects for an experiment. In some implementations, the analysis server 202 runs an experiment selector model to identify the type of experiment to run, given the clinician's trial configuration and the subjects' subject data. Optionally, the experiment selector may be trained to identify and output predicted or suggested compounds or other types of therapeutic parameters. In some cases, this information may be included in the trial configuration used by the analysis server 202 to apply the predictive model for PDT and predict the outcome.

[0062] The analysis server 202 transmits or otherwise outputs predicted results and related information to a user interface operated by the clinician. The output of the predictive model helps the clinician design and set up experimental analyses. The analysis server 202 can run a PDT predictive model trained on subjects by applying PDT to subject data according to various trial configurations. Trial configurations may include PBT-related data showing real-world PBT configurations and real-world compound configurations, which the clinician refers to to set up a real instance of the experiment (corresponding to a digital instance of the experiment modeled by PDT). After running PDT to generate predicted results, the analysis server 202 provides the output of the predicted results and PBT configuration information to a user interface presented on the clinician's user device.

[0063] (B)(C) The clinician sets up and runs the experiment using the output of the analysis system 201, which contains information related to PBT design. The clinician can take samples from a cohort of subjects and prepare the PBT according to the information received from the analysis system 201. As an example, the clinician can take samples from a subject to create a PBT as the subject's physical cells or organoids on one or more microfluidic chips 210. The microfluidic chip 210 includes one or more culture chambers 206, each chamber 206 including perfusion channels and injection ports. When setting up the PBT, the subject's cells are introduced through the injection port or added before assembly of the microfluidic chip. Bubble traps are placed on the fluid pathways to prevent air from entering the culture chambers 206. The chambers 206 provide luminal perfusion, basal perfusion, and cell tubules.

[0064] (D) The clinician introduces the microfluidic chip into the microfluidic chip 210 perfusion module 208. The perfusion module 208 houses one or more microfluidic chips 210. In addition, the perfusion module 208 includes a fluid reservoir, which includes a bio-layer, a reservoir layer, and a pneumatic or hydraulic layer.

[0065] (E) The perfusion module 208 is housed in a CO2 cell incubator 212. The incubator 212 includes a storage rack for the perfusion module 208 (e.g., an 8x4 storage rack) and a pneumatic or hydraulic gas pump located outside the incubator 212. A perfusion controller coupled to the cell incubator 212 manages the function of the cell incubator 212.

[0066] As an example, a single microplate footprint module 208 is divided into four sections for three experiments (e.g., four microfluidic chips 210, each having three chambers 206 for cells), thereby supporting high throughput of 12 experiments per well plate. Furthermore, an integrated perfusion controller allows control of the fluid pathways in each of the four microfluidic chips 210. The pneumatic or hydraulic control system of the cell incubator 212 may include a high-resolution digital pressure controller, along with an on-chip flow restrictor and passive mixing channels, which generates user-programmable drug concentrations for up to 768 PBTs in a single incubator 212. The microfluidic chips 210 are configured to minimize drug absorption loss. The microfluidic chips 210 may be made of injection-molded rigid plastic and, in some embodiments, not made of any silicone. In addition, in some embodiments, the microfluidic chips 210 and / or the cell incubator 212 include an interface for relatively easy coupling to a confocal high-content imaging system 214.

[0067] (F) The experimental apparatus can capture or generate experimental result data based on the output of the cell incubator 212. For example, the inline imaging system imaging reader 214 is associated with the incubator 212 to generate result data, such as live cell readouts. Alternatively, a clinician may input the experimental result data into a user device.

[0068] In some cases, the PBT (Patient-Based Testing) platform components can rapidly and beneficially evaluate each PBT in multidose, multi-drug therapy to assess the proposed therapeutic efficacy and safety. PBT data (e.g., PBT preparation data, result data) includes, for example, secretome, live cells, metabolomics or RNA sequences, and / or endpoint IF images. This PBT-related data can be uploaded to the analysis system 201. In some cases, the PBT-related data is combined with clinical measurements and omics data for upload to the analysis system 201.

[0069] (G) User devices or laboratory equipment can transmit, upload, or otherwise provide experimental result data to the analysis system 201. The analysis server 202 can receive the result data and store it in the analysis database 204. The predictive models used to design and prepare the experiment include a loss function that determines the level of error between the predictive output produced by a particular predictive model and a particular observation in the result data when run by the analysis server 202. In some cases, after the analysis server 202 receives the result data, the analysis server 202 may further apply one or more loss functions of the relevant predictive model to determine the corresponding error level. For a particular predictive model, the loss function may include programming to retrain or adjust the predictive model by adjusting or tuning the hyperparameters or weights of the predictive model based on other potential types of data, such as the level of error, experimental result data, and subject data. During training, the analysis server 202 may determine that any of the predictive models described herein is satisfactorily trained or producing sufficient results if the level of error satisfies a training threshold, a similarity threshold, or other threshold.

[0070] As described above, PBT data (e.g., PBT preparation data, result data) includes, for example, secretome, live cells, metabolomics or RNA sequences, and / or endpoint IF images. This PBT-related data can be uploaded to the analysis system 201. In some cases, PBT-related data may be combined with clinical measurements and omics data for upload to the analysis system 201. The analysis server 202 can incorporate the PBT-related data into predictive models run by the analysis server 202 when designing experiments to be performed in system 200.

[0071] The number of experiments per PBT can proportionally improve the accuracy of the various types of data used in the predictive models of PDTs. This shifts the current whole-population approach to disease treatment to an approach of therapeutic interventions customized for specific subject subgroups, which is algorithmically refined, validated, and optimized for efficacy and / or drug safety. Similarly, the potential benefits to human health and economic aspects would be significant improvements in cohort selection and cost reduction of clinical trials.

[0072] Process example

[0073] Figure 3 illustrates the operation of Method 300 for developing a PDT for predicting and using PBTs in experimental analysis, according to one embodiment. A computer (e.g., analysis server 102) can develop a predictive model of the PDT and, accordingly, develop a biomodel PBT using the PDT. The computer trains one or more PDTs for a given subject using subject data of a given subject stored in a database (e.g., analysis database 104). The predictive model of the PDT is trained according to both a hand-drawn model and a machine learning model to predict experimental outcomes when the experimental analysis is performed according to a test configuration of the experimental analysis, as if the experimental analysis were applied to a subject in the real world, such as measuring disease progression or predicting treatment effects. The trained PDT is applied to an experimental test configuration representing, for example, a disease attribute or a proposed treatment (e.g., drug, procedure) to produce the experimental outcome to be predicted. Method 300 includes both a hand-drawn model and a machine learning model for the PDT, but embodiments may include only a hand-drawn model or only a machine learning model.

[0074] In operation 301, the computer retrieves subject data for any number of subjects from one or more subject databases, the subject data including multiple types of subject data. The databases may include one or more types of databases, such as the analysis database 104, the subject database 106, and / or the conditional database 108. The computer may retrieve subject data by retrieving the subject data at a given interval, or in response to a user command to retrieve subject data. Alternatively, the computer may retrieve subject data by receiving push, replication, or other forms of updates from the databases.

[0075] In some cases, the computer may acquire subject data from only a single subject. Alternatively, the computer may acquire subject data from multiple subjects, such as a group of subjects from a broader population or a subset cohort of subjects.

[0076] In operation 303, the computer generates predictive outputs by applying a hand-drawn model of PDT to the subject data of the subjects. The predictive model of PDT includes programming configured to generate predictive results for the trial configuration according to the hand-drawn model.

[0077] In operation 305, the computer updates one or more parameters of the predictive model and hand-drawn model of the PDT according to user input to calibrate one or more parameters. For example, a clinician may perform multiple iterations of an experiment. Before each iteration, the clinician performs hand-drawn model programming of the predictive model to generate the predicted experimental results (or other types of predicted values).

[0078] In some implementations, clinicians can use both hand-drawn and machine learning models to train and tune predictive models. Hand-drawn models can be used as the initial foundation for predictive models, and clinicians can tune the parameters to calibrate the predictive models. After several iterations or other triggering conditions (e.g., amount of subject data, amount of experimental outcome data), a computer can apply a machine learning model to the same or a new predictive model, including pattern recognition capabilities (e.g., clustering, outlier detection, probabilistic modeling, Bayesian networks) and loss functions to identify features (or subject attributes) and tune the parameters or hyperparameters of the predictive model.

[0079] In some embodiments, handcrafted modeling techniques may be applied only during the initial development of the PDT predictive model (e.g., tuning of weights or parameters). The handcrafted model can be applied to multiple subjects to generate a broader generalized predictive model for a wider subject population. In such embodiments, a computer can store this predictive model as a specific generalized predictive model and generalized PDT for a given set of subjects.

[0080] Correspondingly, in some embodiments, a computer can apply a machine learning model to subject data from only one subject or a small cohort of subjects to generate a customized predictive model for the adjusted PDT. In such embodiments, the computer can store this predictive model as a specific customized predictive model for the adjusted PDT.

[0081] In operation 307, the computer generates a second predicted outcome by applying a predictive model of PDT to subject data and test configuration. The computer may switch to applying a machine learning approach to refine and / or generate the predicted outcome. In operation 309, the computer updates one or more parameters of PDT by applying a machine learning model of PDT to subject data and experimental outcome data of subjects associated with PBT. The loss function measures, for example, the difference between the predicted experimental outcome for a given iteration and the actual experimental outcome data generated from testing the PBT corresponding to PDT. The loss function can then refine this level of error and apply the predictive model when the experiment is performed in the next iteration.

[0082] Implementation example

[0083] Referring again to Figure 1, the components of system 100 can be adopted in various exemplary implementation forms described later for various purposes.

[0084] In some embodiments, the analysis server 102 can use PBT to train and tune predictive models and design experimental analyses (e.g., experiments). The recent widespread adoption of high-throughput and high-content screening has dramatically increased the availability of information about compounds that affect specific molecular targets under different conditions. However, it is not feasible for clinicians to experiment with all possible combinations of drugs, targets, and states. To address this challenge, embodiments can leverage PBT modeling to guide efficient PBT experiments. Embodiments include a server (or other computing device) configured to perform programming for computational modeling techniques for various predictive models, which may include applying (or reapplying in training) one or more types of computational models, such as machine learning, mechanical, stochastic, and Bayesian networks. In some implementation forms, the server may develop an actively learning predictive model for a cohort selection engine ("cohort selector") for the optimized selection of subjects to accept a particular drug, which ingests and uses the PBT model. In some implementations, the server can develop predictive models for experiment selection engines ("experiment selectors") to iteratively predict and select PBT experiments with relatively high predictive information content from a large number of potential experiments. In some implementations, the server can develop predictive models and information capture routines for compound selection that predict or suggest drug compounds and effectively generate compound libraries and small molecule screenings. This beneficially supports higher-throughput PBT experimental research and reduces the need for trial and error to develop therapeutic drugs or for immediate therapeutic needs.

[0085] In some embodiments, by developing an analysis database 104 and subject profiles, detailed phenotypic and genotypic cohorts of NAFLD patients, as well as collections of multimodal datasets and cells, are created. The analysis server 102 can ingest subject data, including deep clinical and multimodal datasets (e.g., Table 1 and / or Table 2), for subjects enrolled in various clinics or treatment centers that have a subject database 106 containing subject data. Clinicians can focus on diverse subjects with a BMI less than 35 who exhibit NAFLD (>5% fat) with or without type 2 diabetes, and the analysis server 102 can store this in a conditional database 108 as subject profiles for the subject cohort. Clinicians create two subcohorts comparing subjects with advanced fibrosis and subjects with no or minimal fibrosis. These database sets in subject database 106 or conditional database 108 include, for example, clinomics, radiomics, subject-reported measurements, and multimodal omics data. Clinicians can use PDT in parallel projects to develop a precision medicine platform for NAFLD.

[0086] In some embodiments, the analysis server 102 develops and implements a preprocessed multimodal data warehouse as the analysis database 104 from the subject database 106 and the conditional database 108. The analysis server 102 can refer to the analysis database 104 to support the creation of PDTs.

[0087] To develop the analysis database 104, the analysis system 101 may collect subject data from subjects using the research data warehouse of subject database 106 or conditional database 108, which includes subject database sets such as those in Table 1 and / or Table 2 (e.g., subcohorts of NAFLD subjects). The conditional database 108 integrates and harmonizes EHRs from various data sources (e.g., different subject databases 106).

[0088] The analysis server 102 has and can run an API connection for user devices 114 to access PBT-related data maintained by the analysis system 101. The analysis system 101 can develop an analysis database 104 as a federated data warehouse accessible via a web portal from user devices 114 or other data sources. The analysis server 102 can run web server software that hosts a cloud computing web application portal for user devices 114 or other data sources to interact with the analysis database 104, which can, for example, host multimodal omics data derived from subjects and PBTs that are not maintained by the analysis system 101. In this way, the analysis system 101 establishes networked or compute-based links between the analysis system 101, data sources (e.g., subject database 106, conditional database 108), subjects, and other actors (e.g., honest broker services). Various entities and computing devices can interact with the analysis system 101 using the API as a centralized interface.

[0089] In some embodiments, the analysis server 102 can use the subject database 106 to develop and host (via API) a clinical natural language processing (NLP) pipeline for clinical phenotyping, which can extract clinical information relevant to NAFLD patients from unstructured clinical notes (or other unstructured data) in the subject database 106. In training and developing the NLP, the clinician and the analysis server 102 map the extracted information to a biomedical ontology (controlled vocabulary) such as UMLS and LOINC for a consistent representation of the clinical information. The analysis server 102 can apply NLP to ingest subject data from the subject database 106, develop a conditional database 108, and / or develop an analysis database 104.

[0090] The analysis server 102 can collect and process data derived from NAFLD subjects, including quantification of genomic and transcriptome data. The clinician or the analysis server 102 can deep mine omics data to extract / construct identifying features for predictive models, such as those reflecting the signaling state of cells for computational discovery and PDT modeling. The analysis system 101 can develop and deploy pipelines and provide an application programming interface (API) to output various types of data for downstream computational discovery or modeling, such as subject data for training predictive models of PDT.

[0091] The analysis server 102 can train predictive models for PDT using knowledge-based handcrafted models, data-driven machine learning models, or both. In the case of the data-driven predictive model approach, the analysis server 102 can apply PDT models to subject data and PBT data from the analysis database 104 (or other data sources), and various types of data are collected from subject cohorts and publicly available clinical trial data. The analysis server 102 outputs PDTs trained on historical knowledge, but customized for specific subject data.

[0092] In some embodiments, the analysis server 102 can initiate the development of PDT by implementing a hand-drawn model of disease progression (historical knowledge-based model) within the analysis system 101. The analysis system 101 can develop the hand-drawn model of PDT to establish an initial mechanism model for predicting clinician-defined disease progression stages, linked by a limited set of experimentally measurable outputs and a set of initially clinician-defined knowledge-derived motion velocity constants fitted from literature-derived longitudinal data (e.g., target data from multiple subjects). The analysis server 102 can apply Bayesian optimization functions to allow the clinician to reconfigure the parameters of the hand-drawn model and identify mappings of model states to additional multimodal outputs (e.g., genomics, metabolomics) using primary subject data from the analysis database 104. The result is, for example, a PDT with the hand-drawn model developed as an algorithmic system of differential equations describing disease progression in a typical subject.

[0093] Next, the analysis server 102 can execute layers of computational model architecture to develop predictive PDT models for patient-specific customization through computational modeling techniques that, among other things, draw strategies for causal inference, such as machine learning, mechanical, stochastic, and Bayesian networks. In some embodiments, for example, the analysis server 102 can apply predictive models configured for aggressive feature selection, pre-screening of significantly correlated individual features, and laser regularization of the remaining feature set. In addition to configuring PDT feature selection, the analysis server 102 can execute deep learning methods to construct / identify optimal PDT discriminant features for predicting patient drug response and disease progression.

[0094] In some implementations, the analysis server 102 can tune, retrain, and validate PDT's predictive models by comparing the predictive outputs of baseline, non-personalized, knowledge-based hand-drawn models with the outputs of other hand-drawn models. The analysis server 102 can assess the level of error by cross-validating the accuracy of PDT's hand-drawn models when stratifying subjects from a primary subject cohort. Using the comparative predictive outputs, the analysis server 102 can identify differences in predicted disease progression outcomes. Subsequently, or simultaneously, the analysis server 102 can similarly validate the predictive outputs of PDT's patient-specific machine learning models therapeutically against experimental results generated using PBT for patient-specific disease progression outcomes.

[0095] As described above, subject experimental representations are biomedical models (also called “subject PBT” or “MPS”) that can be used to understand normal and abnormal function and to provide a basis for preventive or therapeutic interventions in human diseases. In some embodiments, the biomedical model is a patient biomimetic twin (PBT). PBTs are also called MPS (microphysiology system) or OoC (organ on chip). For a given subject, both PBTs and PDTs are generated using data from the same subject, and as a result, the PBT is created as a biomedical (or other physical) model of the subject, while the PDT can be generated to computationally model the same subject. In some embodiments, the experimental representation may include 3D layered cells. 3D layered cells are generated by a combination of continuous cell lamination and intercellular self-organization of specific cell types. Cell types migrate and self-assemble into separate layers that form tissue structures. In some embodiments, one or more cell types are introduced through an injection port. In some embodiments, one or more cell types are organoids, induced pluripotent stem cell (iPSC)-derived cells, or primary cells.

[0096] For example, the 3D layer of cells includes a functional human liver derived from iPSC-derived cells. To produce iPSC-derived cells, a blood sample is taken from a subject. iPSC cells are produced from monocytes and generate iPSC-derived cells. The iPSC cells are programmed into liver-specific cells, which are iPS-hepatocytes, iPS-cholangiocytes, iPS-endothelial cells, iPS-astrocytic cells, and / or iPS-macrophages (also known as iPS-Kupffer cells). It should be understood that the PBT may include or involve one or more connected organ MPS models (e.g., hepatic MPS, islet MPS, etc.) and / or may include one or more types of biomedical models that can function as the PBT described herein.

[0097] In some embodiments, a 3D layer of cells is housed within a microfluidic chip (e.g., microfluidic chip 210). The microfluidic chip comprises at least one chamber having perfusion channels and injection ports. In some embodiments, the microfluidic chip has one or more perfusion channels and one or more injection ports. In some embodiments, the microfluidic chip has a culture medium flowing through it. In some embodiments, one or more microfluidic chips are placed within a perfusion module having a bio-layer, a reservoir layer, and a pneumatic or hydrostatic layer. In some embodiments, one or more of the perfusion modules are placed within a CO2 cell incubator. In some embodiments, the cell incubator includes an inline imaging system.

[0098] In some embodiments, PBTs are designed to replicate key structural, functional, and clinical features of a disease, enabling subject-specific studies of disease progression and drug action mechanisms. PBTs can be used for any experimental analysis. Raw experimental data are generated from the PBT using a user device 114 or other instrument to produce machine-readable data for the analysis server 102. Non-limiting examples include DNA, RNA, protein, and cell function analyses. In some cases, PBTs may be used to analyze toxicity and proliferation. Experimental data generated from the PBT and provided to the analysis server 102 may include secretome data, live cell data, metabolomics data, RNA sequence data, or endpoint immunofluorescence imaging data.

[0099] Among other beneficial applications, PDTs and PBTs are used in combination to predict and / or validate drug efficacy, ADME-Tox, clinical trials, and therapeutic strategies. In some embodiments, a PDT includes a predictive model as a digital representation of how the biology of a particular subject or subject cohort will respond. The analysis server 102 can dynamically integrate multimodal clinical data acquired over time from the subject database 106 and / or conditional database 108 with individual subjects to generate or update the PDT. Examples of subject data datasets are shown in Tables 1 and 2. The analysis server 102 can generate or update the PDT using machine learning models (e.g., statistical, data-driven models) and / or handcrafted models (e.g., mechanical, knowledge-based models). The PDT can predict, for example, the outcome of disease progression in subject biology and / or the effect of a particular therapeutic agent on subject biology.

[0100] Figure 4 illustrates the operation and data flow between the components of System 400 for generating and managing patient data using a machine learning architecture, according to an exemplary embodiment. System 400 includes an analysis server 402 (e.g., analysis servers 102, 202), a user device 414, a PBT database 407 (e.g., analysis databases 104, 204), and a patient database 408 (e.g., analysis databases 104, 204; subject database 106; conditional subject database 108). The analysis server 402 runs a software program or routine that defines or performs the operation of one or more functional engines, such as a patient data engine 409, a PBT data engine 410, and a PDT 411 having one or more predictive models. For ease of explanation and understanding, System 400 is illustrated and described with one example of specific components, such as the analysis server 402, patient data database 408, and PBT data database 407, but embodiments may include any number of components. PBT406 is created from cells or other physical samples from a specific corresponding patient, and PDT411 is generated using patient-related data from the same patient. As shown in Figure 4, PBT406 is used to test predictions generated from the predictive model of PDT411, and the results are then updated in PDT411 or used to update it; PBT406 is used, for example, to test the predicted treatment output from the PDT411 and / or PBT406 models in order to test a drug in PBT406 before treating a patient or cohort.

[0101] The embodiments may include additional or alternative components, or certain components may be omitted from the components shown in Figure 4, and these will still be included within the scope of the disclosure. For example, it is common to include multiple analysis servers 402. The embodiments may include or implement any number of devices capable of performing the various features and tasks described herein. For example, Figure 4 shows the analysis server 402 as a computing device separate from the patient data database 408. In some embodiments, the analysis server 402 includes the integrated patient data database 408.

[0102] The patient data database 408 contains various types of patient data (sometimes referred to as “subject data” or “conditional data”). The patient data database 408 may be hosted on one or more computing devices equipped with hardware components (e.g., non-temporary machine-readable storage media, processors) and software components (e.g., database management systems (DBMS)) capable of performing the various tasks and processes described herein. Similarly, a computing device may host the patient data database 408 or other types of databases, such as the PBT data database 407. In some implementations, the patient data database 408 includes conditional data or a conditional database containing conditional data (e.g., conditional database 108), where conditional data includes pre-processed or “cleaned” instances of patient data. The patient data database 408 receives, stores, manages, and queries various types of patient data. As a non-limiting example, patient data includes, among other types of patient data, clinomics, genomics, metabolomics, proteomics, transcriptomics, epigenomics, and microbiomics. The user device 414 or the analysis server 402 (or other device in system 400) references patient data in the patient data database 408 to generate or update the PDT 411.

[0103] The PBT data database 407 includes various types of PBT data, such as data describing the characteristics of PBT 406, experimental result data showing various information about experiments using PBT, and results of experiments conducted using PBT. The PBT data database 407 may be hosted on one or more computing devices equipped with hardware components (e.g., non-temporary machine-readable storage media, processors) and software components (e.g., database management systems (DBMS)) capable of performing the various tasks and processes described herein. Similarly, a computing device may host the PBT database 407 or other types of databases, such as the patient data database 408. The PBT data database 407 may be a separate database from the patient data database 408, or it may be integrated into the same database as the patient data database 408. The patient data database 408 receives, stores, manages, and queries various types of PBT data. As a non-limiting example, PBT data includes, among other types of data, clinically relevant data, genomics, metabolomics, proteomics, transcriptomics, epigenomics, microbiomics, and experimental results data. User device 414 or analysis server 402 (or other devices in system 400) refer to PBT data in PBT data database 407 to design or update PBT 406 or to generate or update PDT 411.

[0104] User device 414 (e.g., user device 114) includes a computing device that enables an administrator user (e.g., a clinician, healthcare provider, researcher) to interact with the therapeutic modeling services of System 400. Administrators of System 400 include any user responsible for capturing various types of information samples from patients and / or managing the various machine execution processes and tasks described herein. Non-limiting examples of administrators in exemplary embodiments include clinicians, researchers, and healthcare providers (e.g., physicians, nurses).

[0105] The user device 414 may include any computing device comprising hardware (e.g., non-temporary machine-readable storage media, processor) and machine hardware execution software components capable of performing the processes and tasks described herein. Non-limiting examples of the user device 414 may include personal computers (PCs) (workstation computers, laptops), tablets, and smartphones, among other types of electronic devices capable of performing the functions of the user device 414 described herein. The user device 414 may include or be coupled to peripheral devices for receiving user input, such as user I / O devices (e.g., keyboards, mice, monitors), enabling an administrator to interact with the user device 414 and various functions of the system 400. In some cases, the peripheral devices may include electronic measuring devices or instruments that acquire (e.g., receive, generate) patient data (e.g., patient biological measurements) or PBT data (e.g., PBT 406 characteristics; experimental result data when performing a PBT-based experiment on a particular PBT 406). Non-exclusive examples of such electronic devices include image readers, microscopes, live cell image readers, sequencing systems, thermal cyclers, and gel and blot readers. The administrator can input patient-related information as patient data into the user interface of the user device 414, and / or PBT-related information as PBT data into the user interface of the user device 414. The user device 414 transmits or stores the patient data in the patient data database 408 and stores the PBT data in the PBT data database 407.

[0106] In an exemplary embodiment, one or more administrators of System 400 capture various types of patient information and / or patient samples, which are used to generate, update, or use PDT 411 for digitally modeling patients and PBT 406 for physically modeling patients. Patients can register with System 400 (e.g., register with a medical clinic; register or authorize an analysis service). For example, an administrator of System 400 operates a user device 414 to capture patient-related information about a new patient, and the administrator enters the patient information as patient data (sometimes referred to as “subject data”) into the user interface of the user device 414. Additionally or alternatively, electronic measuring devices (e.g., thermometers, electronic heating cuffs) coupled to the user device 414 or the patient data database 408 generate certain types of patient data. The user device 414 or the electronic measuring device transmits the patient data to the patient data database 408 or otherwise stores it.

[0107] The administrator captures or creates patient samples (e.g., blood samples, saliva swabs, skin samples) and / or intermediate physical items (e.g., induced pluripotent stem cells (iPSCs), differentiated organ cells, organoids), and uses them to create one or more PBT406 corresponding to the patient. Using these patient samples (or other physical items), the administrator physically models the patient's biology and prepares one or more PBT406 that are expected to respond to experimental stimuli introduced into the PBT406 by the administrator during a PBT-based experiment according to experimental parameters. As an example, the administrator collects the patient's blood to gather the patient's blood cells, which the administrator then uses to create iPSCs, from which the administrator generates differentiated organ cells, and / or creates organoids from the patient's tissue sample or iPSCs. Using these samples or created physical items, the administrator can then generate PBT406 in a non-animal model (NAMS), such as a microphysiological system (MPS).

[0108] The administrator of system 400 or an electronic measuring device can identify, measure, determine, or otherwise capture various types of information related to PBT406, such as the characteristics of each PBT406 and any experimental results generated using a particular PBT406. The administrator can input PBT-related information as PBT data into the user interface of user device 414, and / or the electronic measuring device can output experimental result data for the PBT data. The user device 414 or the electronic measuring device then stores the PBT data for one or more PBT406 in the PBT data database 407.

[0109] The analysis server 402 of the analysis system 101 is, for example, any computing device comprising one or more processors and software, capable of performing the various processes and tasks described herein. The analysis server 402 uses PBT data, PDT data, or other types of data to perform software programming of functional engines (e.g., patient data engine 409, PBT data engine 410, PDT prediction model 411). The software routines of the functional engines may include the operation of one or more computational models. The executable code that defines the functional engines instructs the analysis server 402 to execute computational models, which may include handcrafted models or machine learning models (e.g., predictive machine learning models) as layers of a machine learning architecture. The analysis server 402 can include, generate, train, or develop various types of computing models by taking various multimodal datasets stored in one or more databases of the system 400 as input.

[0110] The analysis server 402 generates or updates patient-related data or functions of the PDT 411 by running the patient data engine 409, and the analysis server 402 uses the output of the patient data engine 409 to update the PDT 411. In some cases, the analysis server 402 updates the PDT 411 by running the patient data engine 409 and the PBT data engine 410, and the analysis server 402 algorithmically combines or integrates the outputs of the patient data engine 409 and the PBT data engine 410.

[0111] The patient data engine 409 of the analysis server 402 performs various functions using the patient data database 408. In some implementations, the PBT data engine 410 retrieves a specific type of PBT data for one or more sets of patients from the PBT data database 407 and performs specific operations to process or convert the retrieved PBT data into a format compatible with the output of the patient data engine 409 or compatible with the data type of the PDT 411.

[0112] In some implementations, the PBT data engine 410 can train, tune, or develop one or more machine learning models for PBT data. As an example, the PBT data engine 410 can take in PBT-related data, such as experimental results data, to develop or update a PBT data selection model (e.g., a biomarker discovery model) that identifies distinguishive features in PBT data and / or PDT data, and the PBT data selection model may receive feedback from predictive outputs 421 or experimental results data from the PBT data database 407. As another example, the PBT data engine 410 can take in PBT-related data to develop or update a predictive model for PDT 411. The PBT data engine 410 trains, tunes, or develops one or more machine learning models by running specific PBT-related models that take in various types of data, such as PBT data (e.g., experimental results data), PDT data, and predictive outputs 421, among other types of data for training PBT-related machine learning models. The analysis server 402 algorithmically combines or integrates the outputs of the patient data engine 409 and the PBT data engine 410 to generate or update the PDT 411.

[0113] The analysis server 402 runs the predictive model of the PDT 411 to generate a predictive output 421 containing predicted experimental result data. An administrator can input various configuration inputs into the user device 414 to configure a digital experiment with specific experimental parameters. The predictive model of the PDT 411 then runs a digital instance of the experiment using the PDT data and experimental parameters that define the experiment. The PDT 411 then generates a predictive output 421 showing the predicted experimental results, in the form of experimental result data.

[0114] Next, the administrator can test the predictive output 421 from the PDT 411 by conducting a PBT-based experiment using the PBT 406 corresponding to the PDT 411 involved in the digital instance of the experiment. The administrator creates and configures the PBT 406 according to the experiment configuration and parameters. The administrator conducts a PBT-based experiment on one or more PBT 406s and stores the experimental results as PBT-related information in the PBT data database 407. In some cases, the administrator inputs specific experimental results into the user device 414 in the form of experimental result data for experiments conducted on one or more PBT 406s. The user device 414 transmits or stores the experimental result data for one or more PBT 406s in the PBT data database 407. Additionally or alternatively, an electronic measuring device monitors, generates, or otherwise captures experimental result data for experiments conducted on one or more PBT 406s. The electronic measuring device can transmit or store the experimental result data for one or more PBT 406s in the user device 414 or the PBT data database 407.

[0115] The administrator can instruct the analysis server 402 to run the patient data engine 409 and / or the PBT data engine 410 to update the predictive model of PDT 411 (or other machine learning model). In some cases, the analysis server 402 updates PDT 411 based on the difference between the observed experimental outcome data generated from the PBT-based experiment and the predictive output 421, which includes the predicted experimental outcome generated by PDT 411 for a digital instance of the experiment. In this way, the administrator can adjust or confirm the experimental parameters and / or assumed predictions for a proposed real-world clinical experiment before conducting and applying the real-world experiment using actual patients or patient cohorts.

[0116] The administrator, analysis server 402, or user device 414 may generate or determine specific analysis outputs 423 that can be referenced, output, or stored as actionable knowledge. The analysis outputs 423 may be generated or determined based, for example, on experimental result data generated from PBT-based experiments, predictive outputs 421 generated from PDT-based digital experiments, or real-world experimental data generated from conducting real-world experiments on actual patients. The analysis outputs 423 may include, for example, actionable knowledge such as toxicology, disease progression, drug testing, and clinical trials, among other forms of information generated or extrapolated from experimental results generated from PBT-based experiments, predictive outputs 421 generated from PDT-based digital experiments, or real-world experimental data generated from conducting real-world experiments on actual patients.

[0117] The embodiment includes a system having one or more computers having at least one processor, and a non-temporary machine-readable medium for storing processor execution instructions or computer implementation methods for performing a specific operation.

[0118] In the embodiment, the computer comprises at least a processor and is configured to retrieve subject data of a subject from one or more subject databases. The subject data includes multiple types of subject data. The computer can generate predictive outputs by applying a patient digital twin model to the subject data of a subject. The patient digital twin is configured to generate a first predictive result for a trial configuration. The computer can update the model parameters according to user input. The computer can generate a second predictive result by applying the patient digital twin to the subject data and trial configuration. The patient digital twin for a subject includes biomimetic twin configuration data related to the patient biomimetic twin for the subject. The patient biomimetic twin includes an experimental representation of the subject according to the biomimetic twin configuration data. The computer can retrieve the subject's experimental results data and trial configuration related to the subject biomimetic twin for the subject. The computer can update the parameters of the patient digital twin to train the patient digital twin by applying the patient digital twin machine learning model to the subject data and the subject's experimental results data.

[0119] Subject data may indicate that the subject has non-alcoholic fatty liver disease (NAFLD). A predictive model of the patient digital twin is trained on subject data indicating NAFLD. Biomimetic twin configuration data may indicate that the subject biomimetic twin of the subject includes NAFLD in its experimental representation of the subject biomimetic twin based on the subject data.

[0120] The study configuration can indicate subject NAFLD. The biomimetic twin configuration data shows that the subject's patient biomimetic twin includes NAFLD in the experimental representation of the patient biomimetic twin according to the study configuration.

[0121] The computer can apply a model to subject data from multiple subjects. The computer outputs the patient digital twin as a generalized patient digital twin, configured to generate predictive results for a generalized trial configuration across multiple subjects.

[0122] The computer can iteratively train a patient digital twin using subject data and experimental outcome data for specific subjects, thereby outputting the patient digital twin as a customized patient digital twin that is tuned to generate predictive results for a trial configuration for a specific subject.

[0123] The computer can identify cohorts of subjects, including subjects, according to the trial configuration by applying a machine learning architecture cohort prediction engine to subject data. The cohort prediction engine is trained to predict multiple subjects having a set of subject attributes for the trial configuration.

[0124] The computer can retrieve experimental result data related to the patient biomimetic twin of the subject from the laboratory testing equipment.

[0125] A computer can generate a user interface that presents a patient biomimetic twin configuration. The subject biomimetic twin configuration includes a set of subject attributes associated with the subject digital twin and the study configuration.

[0126] A subject's biomimetic twin experimental representation may be placed on a microfluidic chip. The subject's experimental representation may include 3D layered cells housed within the microfluidic chip. The microfluidic chip may comprise one or more chambers equipped with perfusion channels and injection ports. One or more cell types are introduced through the injection ports or incorporated during the assembly of the device. One or more cell types may be selected from organoids, iPSC-derived cells, or primary cells. One or more microfluidic chips may be placed within a perfusion module comprising a biolayer, a reservoir layer, and a gas pressure layer, supplying at least one of pneumatic or hydrostatic pressure. One or more perfusion modules may be placed within a CO2 cell incubator containing an inline imaging system.

[0127] In embodiments, the system includes one or more computers comprising at least a processor and configured to perform specific operations, or a computer implementation method for developing and administering precision therapy. The computer can acquire subject data of a subject and a patient digital twin according to one or more therapy configurations representing the therapy. The computer can acquire predicted therapy outcomes for the therapy according to a predictive model of the patient digital twin. The computer can construct a patient biomimetic twin for a subject based on biomimetic twin configuration data indicated by subject data for the subject. The computer can obtain experimental outcome data by subjecting the patient biomimetic twin to therapy. The computer can input the experimental outcome data into the computer, and the computer updates the predictive model of the patient digital twin based on the level of error between the experimental outcome data and the predicted therapy outcome. The computer can administer the therapy to the subject.

[0128] A computer can iteratively update a patient digital twin using subject data and multiple experimental outcome data iterations, thereby training the patient digital twin as a customized patient digital twin tuned to generate predictive treatment outcomes. The computer can select one or more treatment configurations to develop a precision medicine protocol for treatment. Iterations of the patient biomimetic twin are iteratively applied to treatments according to one or more treatment configurations. The computer iteratively generates iteratives of predicted treatment outcomes by iteratively applying a predictive model of the patient digital twin to subject data according to one or more treatment configurations. Treatments may include a precision medicine protocol for treating non-alcoholic fatty liver disease (NAFLD). Subject data may indicate that the subject has NAFLD. The predictive model of the patient digital twin is trained on subject data indicating NAFLD. The trial configuration may indicate that the subject has NAFLD. The biomimetic twin configuration data indicates that the subject's patient biomimetic twin includes NAFLD in the experimental representation of the patient biomimetic twin according to the trial configuration.

[0129] The computer can run a cohort prediction engine on subject data from multiple subjects to predict a cohort of one or more subjects for treatment. The computer can update the cohort prediction engine based on experimental outcome data obtained for each subject in the cohort. The patient digital twin prediction model can meet an error threshold level before treatment is administered.

[0130] In embodiments, the system includes one or more computers, each having at least a processor and configured to perform specific operations, or a computer implementation method for identifying a cohort of subjects. The computer can acquire subject data for multiple subjects, each subject data including one or more attributes of the subject. The computer can acquire a cohort of subjects for a treatment trial according to one or more treatment configurations, and the computer applies a cohort prediction engine to the subject data of multiple subjects using one or more treatment configurations. For each subject in the cohort, the computer constructs a patient biomimetic twin for the subject based on biomimetic twin configuration data indicated by the subject data for the subject. The computer can acquire experimental outcome data related to the subject's patient biomimetic twin from a laboratory test apparatus by having the patient biomimetic twin receive treatment for the treatment trial. The computer can administer treatment to the cohort of subjects.

[0131] The computer can iteratively update the cohort prediction engine for repeated treatment trials. The computer iteratively updates the subjects for the cohort of subjects obtained for the treatment trial.

[0132] Various exemplary logic blocks, modules, circuits, and algorithmic steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of both. To make this hardware-and-software compatibility evident, various exemplary components, blocks, modules, circuits, and steps have been generally described above in terms of their function. Whether such functions are implemented in hardware or software depends on the specific application and the design constraints imposed on the system. Those skilled in the art may implement the described functions in various ways for specific applications, but such decisions should not be construed as causing a departure from the scope of the invention.

[0133] Embodiments implemented in computer software may be implemented in software, firmware, middleware, microcode, hardware description language, or any combination thereof. A code segment or machine-executable instruction may represent a procedure, function, subprogram, program, routine, subroutine, module, software package, class, or any combination of instructions, data structures, or program statements. A code segment may be coupled to another code segment or hardware circuit by passing and / or receiving information, data, claims, attributes, or memory contents. Information, claims, attributes, data, etc., may be passed, transferred, or transmitted via any preferred means, including memory sharing, message passing, token passing, network transmission, etc.

[0134] The actual software code or dedicated control hardware used to implement these systems and methods does not limit the invention. Therefore, the operation and behavior of the systems and methods have been described without reference to specific software code, it is understood that software and control hardware may be designed to implement the systems and methods based on the description herein.

[0135] When implemented in software, the functionality may be stored as one or more instructions or codes on a non-temporary computer-readable or processor-readable storage medium. The steps of the methods or algorithms disclosed herein may be embodied in a processor-executable software module that may reside on a computer-readable or processor-readable storage medium. Non-temporary computer-readable or processor-readable media include both computer storage media and tangible storage media that facilitate the transfer of computer programs from one location to another. Non-temporary processor-readable storage media may be any available medium that can be accessed by a computer. Such non-temporary processor-readable media may include, but are not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage devices, magnetic disk storage devices or other magnetic storage devices, or any other tangible storage media that can be used to store desired program code in the form of instructions or data structures and can be accessed by a computer or processor. When used herein, discs and discs include compact discs (CDs), laser discs, optical discs, digital multipurpose discs (DVDs), floppy discs, and Blu-ray discs, where discs typically reproduce data magnetically, and discs reproduce data optically using a laser. Combinations of the above should also be included within the scope of computer-readable media. Furthermore, the operation of a method or algorithm may exist as one or any combination or set of code and / or instructions on non-temporary processor-readable media and / or computer-readable media that can be incorporated into a computer program product.

[0136] As used herein and in the claims, the singular forms "a," "an," and "the" include singular and plural references unless the context clearly indicates otherwise.

[0137] As used herein, the term “comprising” is intended to mean that a composition and method includes the enumerated elements but does not exclude others. “Essentially consisting of” means, when used to define a composition and method, to exclude other elements that are essentially important to the composition or method. “Consists of” means to exclude trace amounts or more of other elements for the claimed composition and substantial method steps. Embodiments defined by each of these terms are within the scope of this disclosure. Thus, methods and compositions may include (comprising) additional steps and components, or, alternatively, non-significantly (essentially consisting of) or alternatively, steps and compositions intended to consist only of the described method steps or compositions.

[0138] As used herein, “optional” or “optionally” means that the events or circumstances described later may or may not occur, and that the descriptions include both cases in which such events or circumstances occur and cases in which they do not occur.

[0139] The foregoing description of the disclosed embodiments is provided to enable those skilled in the art to construct or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Accordingly, the invention is not intended to be limited to the embodiments shown herein, but should be given the broadest scope consistent with the following claims and the principles and novel features disclosed herein.

[0140] In addition, if any feature or aspect of the present disclosure is described in relation to the Markush group, a person skilled in the art will recognize that the present disclosure also describes any individual member or subgroup of members of the Markush group.

[0141] All publications, patent applications, issued patents, and other documents referenced herein are incorporated by reference in such a manner as if each individual publication, patent application, issued patent, or other document were specifically and individually indicated to be incorporated by reference in whole. Definitions contained in the texts incorporated by reference are excluded to the extent that they conflict with the definitions in this disclosure.

[0142] While various aspects and embodiments have been disclosed, other aspects and embodiments are contemplated. The various aspects and embodiments disclosed are for illustrative purposes only and are not intended to limit, and the true scope and spirit are indicated by the following claims.

Claims

1. A computer implementation method, A computer retrieves subject data of a subject from one or more subject databases, wherein the subject data includes multiple types of subject data; A step of generating a predictive output by applying a predictive model of a subject digital twin to the subject data relating to the subject using the computer, wherein the subject digital twin includes a predictive model configured to generate a first predictive result of the trial configuration; The computer updates the parameters of the prediction model in response to user input; A step of generating a second prediction result by applying the subject digital twin to the subject data and the test configuration using the computer, wherein the subject digital twin for the subject includes biomimetic twin configuration data related to the subject biomimetic twin for the subject, and the subject biomimetic twin includes an experimental representation of the subject according to the biomimetic twin configuration data; The steps include using the computer to obtain experimental result data of the subject associated with the patient biomimetic twin and the test configuration; The steps include updating the parameters of the subject digital twin by applying a machine learning model of the subject digital twin to the subject data and experimental results data of the subject using the computer; A method of having.

2. The method according to claim 1, wherein the subject data indicates that the subject has non-alcoholic fatty liver disease (NAFLD), the predictive model of the patient digital twin is trained on the subject data indicating the NAFLD, and the biomimetic twin configuration data indicates that the patient biomimetic twin of the subject includes the NAFLD in the experimental representation of the patient biomimetic twin according to the subject data.

3. The method according to claim 1, wherein the test configuration indicates that the subject has NAFLD, and the biomimetic twin configuration data indicates that the patient biomimetic twin of the subject includes NAFLD in the experimental representation of the patient biomimetic twin according to the test configuration.

4. The method according to claim 1, wherein the computer applies the model to the subject data of a plurality of subjects, and the computer outputs the patient digital twin as a generalized patient digital twin configured to generate predictive results of the generalized test configuration for the plurality of subjects.

5. The method according to claim 1, wherein the computer iteratively trains the patient digital twin with subject data and experimental results data for a specific subject, thereby outputting the patient digital twin as a customized patient digital twin that is tuned to generate predictive results for the test configuration for the specific subject.

6. The method according to claim 1, further comprising the step of using the computer to identify a cohort of a plurality of subjects, including the subject, according to the trial configuration, and applying a cohort prediction engine of a machine learning architecture to the subject data, thereby training the cohort prediction engine to predict the plurality of subjects having a set of subject attributes for the trial configuration.

7. The method according to claim 1, wherein the step of acquiring the experimental result data includes receiving the experimental result data relating to the patient biomimetic twin of the subject from the laboratory testing apparatus using the computer.

8. The method according to claim 1, further comprising the step of using the computer to generate a user interface that presents the patient biomimetic twin configuration of the biomimetic twin, the patient biomimetic twin configuration including a set of subject attributes related to the patient digital twin and the test configuration.

9. The computer implementation method according to claim 1, wherein the experimental representation of the patient biomimetic twin is located on a microfluidic chip.

10. The computer implementation method according to claim 9, wherein the subject experimental representation includes 3D layered cells contained within the microfluidic chip.

11. The computer mounting method according to claim 9, wherein the microfluidic chip comprises one or more chambers having an irrigation channel and an injection port.

12. The computer implementation method according to claim 11, wherein one or more cell types are introduced through the injection port or incorporated during the assembly of the device, and the one or more cell types are selected from organoids, iPSC-derived cells, or primary cells.

13. The computer implementation method according to claim 9, wherein one or more of the microfluidic chips are placed in a perfusion module comprising a bio-layer, a reservoir layer, and a gas pressure layer that supply at least one of pneumatic or hydrostatic pressure.

14. One or more cell perfusion modules include an inline imaging system. ２ The method according to claim 13, wherein the cell is placed inside a cell incubator.

15. A system comprising a computer having a processor, wherein the processor is A step of obtaining subject data of a subject from one or more subject databases, wherein the subject data includes multiple types of subject data; A step of generating a predictive output by applying a model of a subject digital twin to subject data for the subject, wherein the subject digital twin is configured to generate a first predictive result of the trial configuration; A step of updating the parameters of the model according to user input; A step of generating a second prediction result by applying the patient digital twin to the subject data and the test configuration, wherein the patient digital twin for the subject includes biomimetic twin configuration data related to the patient biomimetic twin for the subject, and the patient biomimetic twin includes an experimental representation of the subject according to the biomimetic twin configuration data; A step of obtaining experimental result data for the subject related to the subject biomimetic twin and the test configuration for the subject; A step of training the subject digital twin by updating the parameters of the subject digital twin by applying a machine learning model of the subject digital twin to the subject data and experimental result data of the subject; A system configured to perform the following actions.

16. The system according to claim 15, wherein the computer applies the model to the subject data of a plurality of subjects, and the computer outputs the patient digital twin as a generalized patient digital twin configured to generate predictive results of the generalized test configuration for the plurality of subjects.

17. The system according to claim 15, wherein the computer iteratively trains the patient digital twin using the subject data and experimental results data for a specific subject, thereby outputting the patient digital twin as a customized patient digital twin that is tuned to generate predictive results for the test configuration for the specific subject.

18. The system according to claim 15, wherein the computer is further configured to identify a cohort of subjects, including the subject, according to the trial configuration, and the cohort prediction engine of a machine learning architecture is trained to predict the subject having a set of subject attributes for the trial configuration by applying the subject data to the cohort prediction engine.

19. The system according to claim 15, wherein the computer is further configured to acquire the experimental results data relating to the patient biomimetic twin of the subject from a laboratory testing apparatus.

20. The system according to claim 15, wherein the computer is further configured to generate a user interface that presents the subject biomimetic twin configuration of the biomimetic twin, the subject biomimetic twin configuration comprising a set of subject attributes related to the subject digital twin and the test configuration.

21. The system according to claim 20, wherein the experimental representation of the patient biomimetic twin is located on a microfluidic chip.

22. The system according to claim 21, wherein the subject experimental representation includes 3D layered cells contained within the microfluidic chip.

23. The system according to claim 21, wherein the microfluidic chip comprises one or more chambers having an irrigation channel and an injection port.

24. The system according to claim 23, wherein one or more cell types are introduced through the injection port or incorporated during the assembly of the apparatus, and the one or more cell types are selected from organoids, iPSC-derived cells, or primary cells.

25. The system according to claim 21, wherein one or more of the microfluidic chips are placed in a perfusion module comprising a bio-layer, a reservoir layer, and a gas pressure layer that supply at least one of pneumatic or hydrostatic pressure.

26. One or more of the perfusion modules are equipped with an inline imaging system. ２ The system according to claim 15, which is placed inside a cell incubator.

27. A method for developing and administering precision treatment, A step of obtaining subject data and patient digital twins for subjects according to one or more treatment configurations that represent the treatment; A step of obtaining a predicted treatment result for the treatment according to the predictive model of the patient digital twin; A step of constructing a patient biomimetic twin for the subject based on biomimetic twin configuration data shown by the subject data for the subject; A step of obtaining experimental result data by subjecting the patient biomimetic twin to treatment; A step of inputting the experimental results data into a computer, wherein the computer updates the predictive model of the patient digital twin based on the level of error between the experimental results data and the predicted treatment outcome; A step of administering the treatment to the subject; A method of having.

28. The method according to claim 27, wherein the computer iteratively updates the patient digital twin using the subject data and multiple experimental results data iterations, thereby training the patient digital twin as a customized patient digital twin tuned to generate the predictive results for the treatment.

29. The method further comprises the step of selecting one or more treatment configurations for developing a precision medical protocol for the treatment, The patient biomimetic twin repeat is repeatedly applied to the treatment according to the one or more treatment configurations. The computer iteratively generates the predicted treatment results by iteratively applying the predictive model of the patient digital twin to the subject data according to one or more treatment configurations. The method according to claim 28.

30. The method according to claim 29, wherein the treatment comprises a precision medical protocol for treating non-alcoholic fatty liver disease (NAFLD).

31. The method according to claim 29, wherein the subject data indicates that the subject has non-NAFLD, and the predictive model of the patient digital twin is trained on the subject data indicating NAFLD.

32. The method according to claim 29, wherein the test configuration represents the subject NAFLD, and the biomimetic twin configuration data represents that the subject's patient biomimetic twin includes NAFLD in the experimental representation of the patient biomimetic twin according to the test configuration.

33. The method according to claim 27, wherein the computer runs a cohort prediction engine on the subject data of a plurality of subjects to predict a cohort of one or more subjects for the treatment.

34. The method according to claim 33, wherein the computer updates the cohort prediction engine based on the experimental results data obtained for each subject of the cohort.

35. The method according to claim 27, wherein the predictive model of the patient digital twin satisfies an error threshold level before the treatment is administered.

36. A method for identifying subject cohorts, A step of obtaining subject data of multiple subjects, wherein the subject data of each subject includes one or more attributes of the subject; A step of obtaining a cohort of subjects for a treatment trial according to one or more treatment configurations, wherein a computer applies a cohort prediction engine to the subject data of the multiple subjects using the one or more treatment configurations; For each subject in the cohort, the step of constructing a patient biomimetic twin of the subject based on the biomimetic twin configuration data shown by the subject data of the subject; The step of obtaining experimental result data related to the patient biomimetic twin of the subject from a laboratory testing apparatus by subjecting the patient biomimetic twin to the treatment for the treatment trial; A step of administering the treatment to the cohort of subjects; A method of having.

37. The method according to claim 36, wherein the computer iteratively updates the cohort prediction engine over repeated treatment trials, and the computer iteratively updates the subjects for the cohort of subjects acquired for the treatment trials.

Images