Patent Application
20180181889
The present disclosure relates generally to systems and methods for design of processes that result in analytical information or products.
Multi-stage processes are relied upon in the research and manufacture of a wide range of products including biologics, pharmaceuticals, mechanical devices, electrical devices, and food, to name a few examples. Unfortunately, such processes typically have many sources of variation. While most of these sources are minor and may be ignored, the dominant sources of variation may adversely affect the efficiency or even viability of such processes. If identified, however, resources to remove these dominant sources of variation can be engaged and, potentially, such dominant sources of variation can be removed, minimized or contained. Once these dominant sources of variation are addressed, a process may be considered stabilized. When a process is stable, its variation should remain within a known set of limits. That is, at least, until another assignable source of variation occurs. For example, a laundry soap packaging line may be designed to fill each laundry soap box with fourteen ounces of laundry soap. Some boxes will have slightly more than fourteen ounces, and some will have slightly less. When the package weights are measured, the data will demonstrate a distribution of net weights. If the production process, its inputs, or its environment (for example, the machines on the line) change, the distribution of the data will change. For example, as the cams and pulleys of the machinery wear, the laundry soap filling machine may put more than the specified amount of soap into each box. Although this might benefit the customer, from the manufacturer's point of view, this is wasteful and increases the cost of production. If the manufacturer finds the change and its source in a timely manner, the change can be corrected (for example, the cams and pulleys replaced),
While identification of variation of processes is nice in theory, in practice there are many barriers to finding such variation. Most processes combine many different functional components each with their own data forms and types of errors. For instance, a process for manufacturing a synthetic compound using a cell culture combines chemical components, biological components, fermentation components, and industrial equipment components. Each of these components involves different units of quantification, measurement, and error. As such, the rate-limiting step for developing and stabilizing processes is not development of the algorithms that are used in such processes; it is the acquisition and contextualizing of the data in such processes. This requires data aggregation and reproducibility assessment across many disparate systems and functionalities so that scientific reasoning is based on reproducible data rather than on artifacts of noise and uncertainty. Conventional systems fail to deliver adequate capabilities for such analysis. They focus on storing files and data without providing the structure, context or flexibility to enable real-time analytics and feedback to the user.
For instance, electronic lab notebooks (ELNs) are basically “paper on glass” and have inadequate ability to streamline longitudinal analytics across studies. Lab information management systems (LIMS) focus on sample data collection, but don't provide the process or study context to facilitate analytics, nor the flexibility to adapt to changing workflows “on-the-fly” and the many disparate functionalities that are often found in processes. Thus the relationship between process and outcome remains unclear or even inaccessible and information systems become “dead” archives of old work mandated by institutional policies rather than assets that drive process stabilization.
As a result, billions of dollars are lost each year on material and life science research that are not stabilized and thus have unsatisfactory reproducibility rates. Moreover, the incidence of multi-million dollar failures during process transfer to manufacturing remains high.
Another obstacle for such processes are providing systematic methods to determine the relationship between factors affecting a process and the output of that process. In other words, finding cause-and-effect relationships. Such information is needed to manage process inputs in order to optimize the output.
Thus, given the above background, what is needed in the art are improved systems and methods for formulation of experiments to identify sources and causes of variation.
The disclosed embodiments address the need in the art for improved systems and methods for the formulation of experiments to identify sources and causes of variation in processes that result in analytical information or products. The disclosed embodiments address this need by building a run hypergraph from a plurality of parameter combinations for a process subject to a plurality of run constraints for the process As used herein the term “product” refers to, for example, tangible products such as materials, compositions, ingredients, medicines, bulk materials, and the like; and the term “analytical information” refers to, for example, categorical or quantitative data describing measurements of materials, equipment, or process settings. The disclosed systems and methods advantageously and uniquely provides systems and methods for designing a set of runs for a process that will adequately determine the relationship between factors affecting the process and the output of the process.
One aspect of the present disclosure provides a non-transitory computer readable storage medium for building a run hypergraph from a plurality of parameter combinations for a process subject to a plurality of run constraints for the process. The run hypergraph comprises (i) a plurality of nodes, (ii) a plurality of runs, each run in the plurality of runs being associated with a node in the plurality of nodes, and (iii) a plurality of run edges. Each run edge joins (a) a run in the plurality of runs associated with a parent node in the plurality of nodes and (b) a run in the plurality of runs associated with a child node in the plurality of nodes. The process results in a product or analytical information. The non-transitory computer readable storage medium stores instructions, which when executed by a first device, cause the first device to perform a method.
In the method, a process hypergraph for the process is obtained. The process hypergraph comprises the plurality of nodes found in the run hypergraph. In the process hypergraph, these nodes are connected by process edges in a plurality of process edges. Each respective node in the plurality of nodes is associated with a set of parameterized resource inputs to the respective node. At least one parameterized resource input in this set of parameterized resource inputs is associated with one or more input properties, the one or more input properties including an input specification limit. Each respective node in the plurality of nodes is also associated with a set of parameterized resource outputs to the respective node. At least one parameterized resource output in the set of parameterized resource outputs is associated with one or more output properties, the one or more output properties including a corresponding output specification limit.
Each respective process edge in the plurality of process edges of the process hypergraph specifies the set of parameterized resource outputs of a node (parent node) in the plurality of nodes that is included in the set of parameterized resource inputs of at least one other node (child node) in the plurality of nodes and identifies the at least one other node.
In the method, a plurality of factors is identified. Each respective factor in the plurality of factors is associated with (i) an input property in the one or more input properties of a resource input in the set of parameterized resource inputs of a corresponding node in the plurality of nodes or (ii) an output property in the one or more output properties of a resource output in the set of parameterized resource outputs of a corresponding node in the plurality of nodes.
In the method, for each respective factor in the plurality of factors, a number of levels for the input property or output property associated with the respective factor is identified. For example, this may by user specified or read from an input source.
In the method, the plurality of parameter combinations is defined. Each parameter combination in the plurality of parameter combinations is (i) assigned a unique parameter combination identifier from a plurality of unique parameter combinations identifiers, and (ii) includes an instance of each factor in the plurality of factors, where each respective factor in the instance of the plurality of factors is set to a level in the number of levels of the property associated with the respective factor.
In the method, the plurality of run constraints is obtained. Each respective run constraint in the plurality of run constraints corresponds to a different parent node/child node pair in the plurality of nodes that are connected by a process edge in the plurality of process edges. Each respective run constraint in the plurality of run constraints specifies a relationship between a number of runs of the parent node to a number of runs for the child node for the corresponding parent node/child node pair.
In the method, the run hypergraph is build. Each respective run in the plurality of runs of the run hypergraph comprises: (i) an index to a corresponding node in the plurality of nodes, (ii) a run identifier, and (iii) a parameter combination identifier of a parameter combination in the plurality of parameter combinations.
In some embodiments, each respective run in the plurality of runs further comprises (iv) a flag that specifies whether the respective run is marked included, and the building (F) comprises for each respective parameter combination in the plurality of parameter combinations, performing a first enumeration process.
In some embodiments, the first enumeration process comprises adding each node in the plurality of nodes to a first data structure. A first node in the first data structure is removed and used to perform a second enumeration process for the first node when the first data structure is not empty. This processing of removing a first node in the first data structure and using it to perform the second enumeration process is repeated until the first data structure is empty.
In some embodiments, the second enumeration process for the first node comprises adding each parent-child or child-parent node relationship with the first node through a process edge in the plurality of process edges to a parent node-child node connection to a connection data structure. Then, for each respective parent node-child node connection from the connection data structure, the respective parent node-child node connection is removed from the connection data structure and used to perform a third enumeration process for the respective parent node-child node.
In some embodiments, the third enumeration process for the respective parent node-child node connection comprises adding a respective run to the plurality of runs, where the respective run is (i) marked with the identifier for the respective parameter combination, (ii) associated with the parent node, and (iii) includes a level for a factor for the parent node that is specified by the respective parameter combination, when no such run is present in the plurality of runs, where, when such a run is added, the parent node is added back to the first data structure.
In some embodiments, the third enumeration process for the respective parent node-child node connection comprises adding a respective run, to the plurality of runs, where the respective run is (i) marked with the identifier for the respective parameter combination, (ii) associated with the child node, and (iii) includes a level for a factor for the child node that is specified by the respective parameter combination, when no such run is present in the plurality of runs, where, when such a respective run is added, the child node is added to the first data structure.
In some embodiments, the third enumeration process comprises obtaining a subset of runs in the plurality of runs from a bipartite subgraph of the run hypergraph including the parent node and the child node are (i) associated with the parent node and include a level for a factor for the parent node that is specified by the respective parameter combination or (ii) associated with the child node and include a level for a factor for the child node that is specified by the respective parameter combination.
In some embodiments, the third enumeration process is aborted when each run in the subset of runs is marked as “included.” For each respective run in the subset of runs that has not been marked as “included,” a fourth enumeration process is performed.
In some embodiments, the fourth enumeration process for the respective run comprises marking the respective run as “active” and then marking as “active” any run in the subset of runs that are (i) connected to the respective run by a run edge in the plurality of run edges or a combination of run edges in the plurality of edges. The fourth enumeration process continues with the identification within the plurality of runs or adding to the plurality of runs, one or more runs indexed to the parent node or the child node that specifies the level for the factor specified by the respective parameter combination for the respective parent node or child node, when the respective run constraint in the plurality of run constraints between the parent node and the child node is not satisfied by the runs in the plurality of runs that are marked as “active,” thereby satisfying the respective run constraint. The fourth enumeration process continues with marking as “active” these runs that have been newly identified or added, wherein, when runs are added to the parent node or the child node, the parent node or the child node is added back to the first data structure. Further all runs newly identified or added are linked to all parent runs or child runs in the subset of runs that are marked as “active” by assigning each newly identified or added run with a run edge to a parent run or a child run in the subset of runs that is marked as “active.”
The fourth enumeration process continues by marking as “included” all runs in the plurality of runs that are marked “active” and clearing the “active” label from all runs in the plurality of runs.
In some embodiments, the plurality of nodes comprises five or more nodes.
In some embodiments, the set of parameterized resource inputs for a node in the plurality of nodes comprises a first and second parameterized resource input, the first parameterized resource input specifies a first resource and is associated with a first input property, the second parameterized resource input specifies a second resource and is associated with a second input property, and the first input property is different than the second input property.
In some embodiments, the set of parameterized resource inputs for a node in the plurality of nodes comprises a first and second parameterized resource output, the first parameterized resource output specifies a first resource and is associated with a first output property, the second parameterized resource input specifies a second resource and is associated with a second output property, and the first output property is different than the second output property.
In some embodiments, the set of parameterized resource inputs for a node in the plurality of nodes comprises a first parameterized resource input, the first parameterized resource input specifies a first resource and is associated with a first input property and a second input property, where the first input property is different than the second input property.
In some embodiments, the set of parameterized resource inputs for a node in the plurality of nodes comprises a first parameterized resource output, the first parameterized resource output specifies a first resource and is associated with a first output property and a second output property, where the first output property is different than the second output property.
In some embodiments a first input property or a first output property is a viscosity value, a purity value, composition value, a temperature value, a weight value, a mass value, a volume value, or a batch identifier of the first resource.
In some embodiments, the set of parameterized resource inputs for a first node in the plurality of nodes comprises a first parameterized resource input, and an input property associated with the first parameterized resource input specifies a process condition associated with the corresponding node. In some such embodiments, the process condition comprises an intensive quantity, an extensive quantity, a temperature, a volume, time, a space, a quality, a type of equipment, an order, a state, or a batch identifier.
In some embodiments, the set of parameterized resource outputs for a first node in the plurality of nodes comprises a first parameterized resource output, and an output property associated with the first parameterized resource output specifies a process condition associated with the corresponding node. In some such embodiments, the process condition comprises an intensive quantity, an extensive quantity, a temperature, a volume, time, a space, a quality, a type of equipment, an order, a state, or a batch identifier.
In some embodiments, the corresponding output specification limit comprises a nominal value, an upper limit or a lower limit for an output property of a corresponding parameterized resource output.
In some embodiments, the corresponding output specification limit comprises an enumerated list of allowable types or states.
In some embodiments, a factor in the plurality of factors is a continuous factor, a discrete numeric factor, or a categorical factor.
In some embodiments, the defining of the plurality of parameter combinations implements a full factorial design of the plurality of factors to define the plurality of parameter combinations, where the plurality of parameter combinations collectively defines, for each respective factor in the plurality of factors, the specified number of levels of the specified property associated with the respective factor.
In some embodiments, the defining of the plurality of parameter combinations implements a fractional factorial design (e.g., Taguchi design or a Latin Squares design) of the plurality of factors to define the plurality of parameter combinations, where the plurality of parameter combinations collectively defines, for each respective factor in at least a subset of the plurality of factors, a subset of the levels of the specified property associated with the respective factor.
In some embodiments, the defining of the plurality of parameter combinations implements a D-optimal or I-optimal design algorithm (e.g., a Fedorov algorithm) to define the plurality of parameter combinations.
In some embodiments, the defining of the plurality of parameter combinations requires repeating the defining until an exit condition is satisfied. Examples of an exit condition include user acceptance of the plurality of parameter combinations, or a power calculation based upon the plurality of parameter combinations satisfying a first threshold level (e.g., at least eighty percent, at least ninety percent, at least 99 percent, at least 99.9 percent).
In some embodiments, a first run constraint in the plurality of run constraints is an equality or inequality property imposed between the output of the parent node and the input of the child node in the parent node/child node pair associated with the first run constraint.
In some embodiments, a first run constraint in the plurality of run constraints is a mass balance inequality constraint between the output of the parent node and the input of the child node in the parent node/child node pair associated with the first run constraint.
In some embodiments, a first run constraint in the plurality of run constraints is a one-to-one, many-to-one, or one-to-many relationship between (i) the number of runs of the parent node and (ii) the number of runs for the child node for the corresponding parent node/child node pair.
In some embodiments, the method further comprises adding runs to the plurality of runs prior to building the run hypergraph, where the adding comprises (i) obtaining a set of runs, each run in the set of runs associated with a respective node in the plurality of nodes, (ii) joining a subset of runs in the set of runs, where each run in the subset of runs is linked to at least one other run in the subset of runs by a run edge included in or added to the plurality of run edges, (iii) assigning each run in the subset of runs with the parameter combination identifier of a parameter combination in the plurality of parameter combinations when the subset of runs includes a respective run for each respective factor in the plurality of factors at the respective level specified in the parameter combination for the respective factor, (iv) removing the subset of runs from the set of runs, (v) repeating the obtaining (i), joining (ii), assigning (iii) and removing (iv) until an exit condition is achieved, and (vi) adding each run that has been assigned a parameter combination identifier in the assigning (iii) to the plurality of runs. In some such embodiments, each run in the set of runs specifies a level for a factor for the respective node corresponding to the factor. In some such embodiments, the set of runs are created by a user. In some embodiments, the exit condition is depletion of the set of runs.
In some embodiments, the method further comprises adding runs to the plurality of runs prior to the building the run hypergraph, wherein the adding comprises: (i) obtaining a set of runs, where each run in the set of runs is associated with a respective node in the plurality of nodes, (ii) joining a subset of runs in the set of runs, where each run in the subset of runs is linked to at least one other run in the subset of runs by a run edge included in or added to the plurality of run edges, (iii) removing the subset of runs from the set of runs, (iv) repeating the obtaining (i), joining (ii), and assigning (iii) until an exit condition is achieved, thereby achieving a plurality of subsets of runs, (v) co-clustering each subset of runs in the plurality of subsets of runs that includes a run for each factor in the plurality of factors with the plurality of parameter combinations, where the co-clustering produces a plurality of clusters, and where each cluster in the plurality of clusters includes at most one parameter combinations in the plurality of parameter combinations; (vi) assigning each run in the plurality of subsets of runs that co-clusters with a respective parameter combination the parameter combination identifier assigned to the respective co-clustered parameter combination; and (vii) adding each run that has been assigned a parameter combination identifier in the assigning (vi) to the plurality of runs. In some such embodiments, each run in the set of runs specifies a level for a factor for the respective node corresponding to the factor. In some such embodiments, the set of runs are created by a user. In some such embodiments, the co-clustering is performed by k-means clustering or hierarchical clustering based on a distance metric. In some such embodiments, the distance metric is a Euclidian distance metric, a Hamming distance metric, or a correlation
In some embodiments, the method further comprises pruning the plurality of runs by counting a number of runs at each node in the plurality of nodes that have the same assigned parameter combination identifier.
Like reference numerals refer to corresponding parts throughout the several views of the drawings.
Reference will now be made in detail to embodiments, examples of which are illustrated in the accompanying drawings. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. However, it will be apparent to one of ordinary skill in the art that the present disclosure may be practiced without these specific details. In other instances, well-known methods, procedures, components, circuits, and networks have not been described in detail so as not to unnecessarily obscure aspects of the embodiments.
It will also be understood that, although the terms first, second, etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first subject could be termed a second subject, and, similarly, a second subject could be termed a first subject, without departing from the scope of the present disclosure. The first subject and the second subject are both subjects, but they are not the same subject.
The terminology used in the present disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the description of the invention and the appended claims, the singular forms “a”, “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that the term “and/or” as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items. It will be further understood that the terms “comprises” and/or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
As used herein, the term “if” may be construed to mean “when” or “upon” or “in response to determining” or “in response to detecting,” depending on the context. Similarly, the phrase “if it is determined” or “if [a stated condition or event] is detected” may be construed to mean “upon determining” or “in response to determining” or “upon detecting [the stated condition or event]” or “in response to detecting [the stated condition or event],” depending on the context.
A detailed description of a system 48 for building a run hypergraph from a plurality of parameter combinations for a process subject to a plurality of run constraints for the process in accordance with the present disclosure is described in conjunction with
Of course, other topologies of system 48 are possible, for instance, computer system 200 can in fact constitute several computers that are linked together in a network or be a virtual machine in a cloud computing context. As such, the exemplary topology shown in
Referring to
Turning to
The memory 192 of computer system 200 stores:
In some implementations, one or more of the above identified data elements or modules of the computer system 200 are stored in one or more of the previously mentioned memory devices, and correspond to a set of instructions for performing a function described above. The above identified data, modules or programs (e.g., sets of instructions) need not be implemented as separate software programs, procedures or modules, and thus various subsets of these modules may be combined or otherwise re-arranged in various implementations. In some implementations, the memory 192 and/or 290 optionally stores a subset of the modules and data structures identified above. Furthermore, in some embodiments the memory 192 and/or 290 stores additional modules and data structures not described above.
Turning to
In some embodiments, a node 304 is a complete and self-contained description of a transformative event that can be used to build larger processes. A node 304 is sufficiently general to serve in a wide array of processes, such as chemical processes, life science processes, and food preparation processes. Advantageously, nodes 304 do not lose their meaning or utility when copied into other processes. As such, the definition of a node 304 does not depend on the definition of other nodes in a process hypergraph 302 in preferred embodiments.
Each respective node 304 in the plurality of nodes of a process hypergraph 304 is associated with a set of parameterized resource inputs 308 to the respective stage in the corresponding process. At least one parameterized resource input 310 in the set of parameterized resource inputs 308 is associated with one or more input properties 312, the one or more input properties including an input specification limit 314. Examples of input properties 312 are the attributes (e.g., measurements, quantities, etc.) of things such as people, equipment, materials, and data. There can be multiple input properties for a single parameterized resource input (e.g., temperature, flow rate, viscosity, pH, purity, etc.). In some embodiments, there is a single input property for a particular parameterized resource input.
Each respective node 304 in the plurality of nodes is also associated with a set of parameterized resource outputs 315 to the respective stage in the corresponding process. At least one parameterized resource output 316 in the set of parameterized resource outputs 315 is associated with one or more output properties 318, the one or more output properties including a corresponding output specification limit 320. Examples of output properties 318 include attributes (e.g., measurements, quantities, etc.) of things such as people, equipment, materials, and data. There can be multiple output properties for a single parameterized resource output. In some embodiments, there is a single output property for a particular parameterized resource output. Further discussion of such parameterized resource inputs and parameterized resource outputs is disclosed in PCT publication WO 2016/019188 A1 entitled “Systems and Methods for Process Design and Analysis,” in particular the text describing FIGS. 17 and 18 of WO 2016/019188 A1, which is hereby incorporated by reference.
Returning to
As
In some instances, a destination node 304 of a process hypergraph 302 includes only a single process edge 322 from one source node 324. In such instances, the set of parameterized resource outputs 315 for the source node 324 constitutes the entire set of parameterized resource inputs 308 for the destination node 326.
To illustrate the concept of a node in a process represented by a process hypergraph 302, consider a node that is designed to measure the temperature of fermenter broth. The set of parameterized inputs 308 to this node include a description of the fermenter broth and the thermocouple that makes the temperature measurement. The thermocouple will include input properties that include its cleanliness state, calibration state and other properties of the thermocouple. The set of parameterized outputs 315 to this node 304 include the temperature of the fermenter broth, and output specification limits for this temperature (e.g., an acceptable range for the temperature). Another possible parameterized resource output 316 of the node 304 is the thermocouple itself along with properties 318 of the thermocouple after the temperature has been taken, such as its cleanliness state and calibration state. For each of these output properties 318 there is again a corresponding output specification limit 320.
In some instances, a destination node of a process hypergraph 302 includes multiple process edges 322, each such edge from a different source node. In such instances, the set of parameterized resource outputs 315 for each such source node collectively constitute the set of parameterized resource inputs 308 for the destination node.
As an example, consider the case where a plurality of factors 226 consists of 10 factors, with each of the 10 factors having one of two possible levels. A first parameter combination 406-1 in the plurality of parameter combinations 228 will contain a first instance of the plurality of factors 226-1 (10 factors), with each respective factor 402 in the first instance of the plurality of factors 226-1 independently assigned to one of the two possible levels 404 for the respective factor, a second parameter combination 406-2 in the plurality of parameter combinations 228 will contain a second instance of the plurality of factors 226-2 (10 factors), with each respective factor 402 in the second instance of the plurality of factors 226-1 independently assigned to one of the two possible levels 404 for the respective factor, and so forth.
As another example, consider the case where a plurality of factors 226 consists of 5 factors, with each of the 5 factors having one of a plurality of possible levels. A first parameter combination 406-1 in the plurality of parameter combinations 228 will contain a first instance of the plurality of factors 226-1 (5 factors), with each respective factor 402 in the first instance of the plurality of factors 226-1 independently assigned to one of the plurality of possible levels 404 for the respective factor, a second parameter combination 406-2 in the plurality of parameter combinations 228 will contain a second instance of the plurality of factors 226-2 (5 factors), with each respective factor 402 in the second instance of the plurality of factors 226-1 independently assigned to one of the plurality of possible levels 404 for the respective factor, and so forth.
Now that details of a system 48 for building a run hypergraph 204 from a plurality of parameter combinations 228 for a process subject to a plurality of run constraints 230 for the process have been disclosed, details regarding how a run hypergraph build module 103 of the system 48 builds a run hypergraph 204 in accordance with an embodiment of the present disclosure are disclosed with reference to
Referring to block 502, A run hypergraph build module 103 of the system 48 builds a run hypergraph 204 from a plurality of parameter combinations 228 for a process subject to a plurality of run constraints 230 for the process is provided. The run hypergraph 204 comprises (i) a plurality of nodes 304, (ii) a plurality of runs 208, each run 208 in the plurality of runs being associated with a node 304 in the plurality of nodes, and (iii) a plurality of run edges 218. Each run edge 218 joins (a) a run in the plurality of runs associated with a parent node in the plurality of nodes and (b) a run in the plurality of runs associated with a child node in the plurality of nodes. The process results in a product or analytical information. The non-transitory computer readable storage medium stores instructions, which when executed by a first device, cause the first device to perform a method.
Referring to block 504, in some embodiments, the plurality of nodes comprises five or more nodes, 10 or more nodes, 15 or more nodes, or 100, or more nodes.
Referring to block 506, a process hypergraph 302 is obtained for the process. The process hypergraph 302 comprise a plurality of nodes 304 connected by process edges 322 in a plurality of process edges. Each respective node 304 in the plurality of nodes comprises a process stage label representing a respective stage in the corresponding process.
Each node 304 is associated with a set of parameterized resource inputs 308 to the respective stage in the corresponding process. At least one parameterized resource input 310 in the set of parameterized resource inputs 308 is associated with one or more input properties 312. The one or more input properties include an input specification limit 314. Each node 304 is also associated with a set of parameterized resource outputs 315 to the respective stage in the corresponding process. At least one parameterized resource output 316 in the set of parameterized resource outputs is associated with one or more output properties. The one or more output properties include a corresponding output specification limit.
Each respective process edge 322 in the plurality of process edges specifies that the set of parameterized resource outputs of a node in the plurality of nodes is included in the set of parameterized resource inputs of at least one other node in the plurality of nodes. Thus, turning to
To illustrate a set of parameterized resource inputs 308, in some embodiments, the set of parameterized resource inputs 308 for a node 304 in the plurality of nodes of a process hypergraph 302 comprises a first 310-1 and second parameterized resource input 310-2. The first parameterized resource input specifies a first resource and is associated with a first input property 312-1 (508). The second parameterized resource input 310-2 specifies a second resource and is associated with a second input property 312-2. In some embodiments, the first input property is different than the second input property.
In some embodiments, the set of parameterized resource inputs 308 for a node 304 in the plurality of nodes comprises a first parameterized resource input 310. The first parameterized resource input 310 specifies a first resource and is associated with a first input property 312 and a second input property 312. The first input property is different than the second input property (510).
In some embodiments, the first input property is a viscosity value, a purity value, composition value, a temperature value, a weight value, a mass value, a volume value, or a batch identifier of the first resource (512).
In some embodiments a resource input 310 is a single resource. For instance, in
In some embodiments, the set of parameterized resource inputs 308 for a first node 304 in the plurality of nodes of a process hypergraph 302 comprises a first parameterized resource input 310 and this first parameterized resource input specifies a process condition associated with the corresponding stage of the process associated with the first node 304 (514). For example, in some embodiments, this process condition specifies an intensive quantity, an extensive quantity, a temperature, a volume, time, a space, a quality, a type of equipment, an order, a state, or a batch identifier (516).
As noted above in some embodiments, for a given node, at least one of the parameterized resource outputs in the set of parameterized resource outputs for the node is associated with one or more output properties, and in some such embodiments the one or more output properties includes a corresponding output specification limit. In some embodiments, this corresponding output specification limit comprises a nominal value, an upper limit, and/or a lower limit for the corresponding parameterized resource output (518). To illustrate, an example of an output property is pH of a composition. In such an example, the output specification limit specifies the allowed upper limit for the pH of the composition and the allowed lower limit for the pH of the composition. In alternative embodiments, this corresponding output specification limit comprises an enumerated list of allowable types (520). To illustrate, an example of an output property is a crystallographic orientation of a material. In such an example, the output specification limit specifies an enumerated list of allowed crystallographic orientations for material.
Referring to block 522 of
Next, referring to block 544, there is identified, for each respective factor 402 in the plurality of factors, a number of levels 404 for the input property 312 or output property 318 associated with the respective factor.
Referring to block 526, in some embodiments, a factor 402 in the plurality of factors is a continuous factor, a discrete numeric factor, or a categorical factor. For instance, referring to
Referring to block 528 of
Referring to block 530 of
Referring to block 532 of
Referring to block 534 of
Referring to block 538 of
In some embodiments the creation of the plurality of parameter combinations is repeated and/or continued until the plurality of parameter combinations achieves a certain power calculation. Power calculations are used to determine the sample size required in to detect a meaningful scientific effect with sufficient power. For instance, in R (R Core Team, 2012, “R: A language and environment for statistical computing. R Foundation for Statistical Computing,” Vienna, Austria. ISBN 3-900051-07-0, URL R-project.org/, which is hereby incorporated by reference), there are functions to calculate either a minimum number of parameter combinations 506 needed for a specific power for a given plurality of factors.
Referring to block 540 of
Thus, referring to
Run constraint 232-2 corresponds to the parent node/child node pair 304-10 “Prepare glucose media”/304-4 “Fill 100 ml flask to 40 ml with growth media & glucose solution” that are connected by the process edge 322-15. The run constraint 232-2 specifies a relationship between a number of runs 208 of the parent node 304-10 to a number of runs 208 for the child node 304-4 for the corresponding parent node/child node pair. Specifically run constraint 232-2 specifies that the sum of the volume of input glucose solution over the child runs (for the node 304-4) must be less than the sum of volume of output glucose solution of parent runs 208 (for node 304-10). Thus, run constraint 232-2 sets a limit between the number of runs of the parent node 304-10 and the number of runs of the child node 304-4.
Run constraint 232-3 corresponds to the parent node/child node pair 304-11 “Treatment solution prep”/304-5 “Add 9 ml treatment solution to growth media” that are connected by the process edge 322-16. The run constraint 232-3 specifies a relationship between a number of runs 208 of the parent node 304-11 to a number of runs 208 for the child node 304-5 for the corresponding parent node/child node pair. Specifically run constraint 232-3 specifies that the sum of the input chemical over the child runs (for the node 304-5) must be less than the sum of volume of output chemical of parent runs 208 (for node 304-11). Thus, run constraint 232-3 sets a limit between the number of runs of the parent node 304-11 and the number of runs of the child node 304-5.
Run constraint 232-4 corresponds to the parent node/child node pair 304-2 “Fill microtiter plate with 270 μL growth media”/304-9 “Transfer 30 μl cell culture to microtiter plate” that are connected by the process edge 322-4. The run constraint 232-4 specifies a relationship between a number of runs 208 of the parent node 304-2 to a number of runs 208 for the child node 304-9 for the corresponding parent node/child node pair. Specifically, run constraint 232-4 specifies that the count of the number of child runs (for the node 304-9) must be less than the sum of the number of wells outputted by microtiter plates in the parent runs 208 (for node 304-2). Thus, run constraint 232-4 sets a limit between the number of runs of the parent node 304-2 and the number of runs of the child node 304-9.
Run constraint 232-5 corresponds to the parent node/child node pair 304-8 “Dilute culture to 10× target OD”/304-9 “Transfer 30 μl cell culture to microtiter plate” that are connected by the process edge 322-5. The run constraint 232-5 specifies a relationship between a number of runs 208 of the parent node 304-8 to a number of runs 208 for the child node 304-9 for the corresponding parent node/child node pair. Specifically, run constraint 232-5 specifies that the count of the number of child runs (for the node 304-9) must be less than four times the count of the parent runs 208 (for node 304-8). Thus, run constraint 232-5 sets a limit between the number of runs of the parent node 304-9 and the number of runs of the child node 304-8.
Referring to block 542 of
Referring to block 544 of
Referring to block 546 of
Referring to block 548 of
Referring to block 550 of
Referring to block 552 of
Referring to block 554 of
In some embodiments the third enumeration process for the respective parent node-child node connection 2006 also comprises checking to see if there is a qualifying run for the child node in the parent node-child node connection 2006. Specifically, (b) adding a respective run 208 to the plurality of runs of the run hypergraph 204, where the respective run is (i) marked with the identifier 408 for the respective parameter combination 406, (ii) associated with the child node 304, and (iii) includes a level 404 (e.g., for an input or output property 312/318 specified by the factor) for a factor 402 for the child node 304 that is specified by the respective parameter combination 406, when no such run 208 is present in the plurality of runs, where, when a respective run is added in (b), the child node 304 is added to the first data structure 2002. Thus, referring to
Thus, at least upon completion of steps (a) and (b) of the third enumeration process, there exists (i) a run 208 associated with the parent node 304 of the respective parent node-child node connection 2006 that specifies a level 404 for a property of a factor 402 associated with the parent node 304 in the parameter combination 406 specified in block 550 and (ii) a run 208 associated with the child node 304 of the respective parent node-child node connection 2006 that specifies a level 404 for a property of a factor 402 associated with the child node 304 in the parameter combination 406 specified in block 550. In step (c) of the third enumeration process, a subset of runs in the plurality of runs from a bipartite subgraph of the run hypergraph 204 including the parent node 304 and the child node that are (i) associated with the parent node and include a level for (a property of) a factor for the parent node that is specified by the respective parameter combination 406 or (ii) associated with the child node 304 and include a level for (a property of) a factor for the child node that is specified by the respective parameter combination are obtained. Such a selection will at least include (i) runs 208 associated with the parent node 304 of the respective parent node-child node connection 2006 that specify a level 404 for a property of a factor 402 associated with the parent node 304 in the parameter combination 406 specified in block 550 and (ii) runs 208 associated with the child node 304 of the respective parent node-child node connection 2006 that specifies a level 404 for a property of a factor 402 associated with the child node 304 in the parameter combination 406 specified in block 550. Step (d) of the third enumeration process calls for aborting the third enumeration process when each run 208 in the subset of runs of step (c) is marked as “included.” In some embodiments, the include flag 214 is used to denote whether a run is include or not, although any means for tracking whether a run is included may be used and is encompassed within the scope of the present disclosure. In step (e) of the third enumeration process, there is run for each respective run in the subset of runs that has not been marked as “included,” a fourth enumeration process.
Referring to block 556 of
A general algorithm for computing the runs required in order to satisfy a set of equality and inequality constraints such as the plurality of run constraints 230 on runs 208 of nodes in a process follows. This is a general algorithm class that answers the question “how many runs of each step are needed to satisfy a run ratio or run inequality specified by a run constraint 232?” and can be applied to any set or subset of nodes in a run hypergraph 204 (e.g., all the nodes of the process embodied by the run hypergraph or somewhere in between). The problem of generating runs subject to constraints can be posed as an integer linear program (ILP). The canonical form of the ILP is:
where,
x is vector, of which each element in x corresponds to the number of runs of that node, thus each element must be an integer greater than zero,
A is a matrix having integer values,
b is a vector holding integer values,
s is a vector holding integer values,
A, b, s, x, together define the equalities and inequalities imposed on the system by the plurality of run constraints 230.
This canonical problem can be solved using methods such as cutting-place, branch and bound, branch and cut, branch and price.
One algorithm for analytically computing runs required in order to satisfy run ratios is as follows. Consider nodes A and B, with a run ratio of rA:rB imposed by a run constrain 232, where rA and rB are the required ratios of numbers of runs 208 for each node 304. Define nA and nB as the required number of runs, to be calculated. Assume there are existing runs for each node, denoted nA0, nB0. If nA0/nB0≠rA/rB, then more runs of A and/or B are required. First, compute a new lower limit on the required number of runs of A and B, denoted as nA1 and nB1, respectively, using the ceiling operator to round up values to the nearest integer. Compute nA1=rA*ceiling(nA0/rA). Compute nB1=rB*ceiling(nB0/rB). Then select the appropriate case to evaluate:
1) If nA1/nB1=rA/rB, then nA=nA1 and nB=nB1
2) If nA1/nB1>rA/rB, then nA=nA1 and nB=nA1*rB/rA
3) If nA1/nB1<rA/rB, then nA=nB1*rA/rB and nB=nB1
Finally, add (nA−nA0) runs of step A and (nB−nB0) runs of step B
As an example, consider the scenario:
rA=3
rB=2
nA0=7
nB0=3
Compute nA1 and nB1:
nA1=3*ceiling(7/3)=9
nB1=2*ceiling(3/2)=4
Here, case two from the algorithm above applies, because nA1/nB1=9/4>3/2=rA/rB. So: nA=9, nB=6. So, add two runs to node A and 3 runs to node b.
In step (d) of the fourth enumeration process, there is marked as “active” all runs identified or added in step (c) of the fourth enumeration process, where, when runs are added to the parent node or the child node in step (c) of the fourth enumeration process, the parent node or the child node is added back to the first data structure 2002. In step (e) of the fourth enumeration process all runs identified or added in step (c) of the fourth enumeration process are linked to all parent runs or child runs in the subset of runs that are marked as “active” by assigning each run, identified or added in step (c) with a run edge 218, to a parent run (i.e., associated with the parent node) or a child run (i.e., associated with the child node) in the subset of runs that is marked as “active.” In step (f) of the fourth enumeration process all runs in the plurality of runs that are marked “active,” are marked as “included” and the “active” label is cleared from all runs in the plurality of runs in the run hypergraph 204.
Thus, by the recursive application of the first through fourth enumeration processes a run hypergraph is built from a plurality of parameter combinations for a process subject to a plurality of run constraints for the process.
In some embodiments there are additions and/or variants to the process described in conjunction with
In alternative embodiments, one or more runs 208 are added to the plurality of runs of the run hypergraph 204 prior to performing building block 548 or contemporaneously with performing building block 548. In these alternative embodiments, such runs 208 may represent preexisting runs or runs added by a user. In such embodiments, a set of runs is obtained, where each run in the set of runs is associated with a respective node in the plurality of nodes of the run hypergraph 204. A subset of runs in the set of runs is joined, where each run in the subset of runs is linked to at least one other run in the subset of runs by a run edge 218 included in or added to the plurality of run edges. The subset of runs is removed from the set of runs and this process is repeated until an exit condition is achieved, thereby achieving a plurality of subsets of runs. Each subset of runs includes a level of a property 312/318 in the plurality of properties and thus potentially may be grouped with one of the plurality of parameter combinations 228 defined in block 528 of
In some embodiments, the method further comprises pruning the plurality of runs by counting a number of runs at each node in the plurality of nodes that have the same assigned parameter combination identifier. For instance, in some embodiments, working down the process flow of the run hypergraph 204 from node 304 to node 304 by traversing run edges 218, all runs 208 with a common parent run that share the same property values are collapsed into a minimum set of runs that still satisfy all run constraints 232.
1) A user selects (a) properties of resources on nodes 304 (a.k.a., factors, variables) to be varied, (b) the desired test levels 404 for each factor 402, and (c) design of experiment algorithm parameters.
2) A user executes a design of experiment (DOE) algorithm. The DOE algorithm generates a design matrix of parameter combinations across nodes (plurality of parameter combinations 228) using, for example a D-optimal or I-optimal design algorithm such as the Fedorov algorithm implemented, for example, in R. The DOE algorithm also generates design evaluation information such as the Power of the design and the number of runs. The user iterates on DOE design matrix using design evaluation info until satisfied with the plurality of parameter combinations 228.
3) The user defines run constraints 232 on the run edges 218 between nodes 204 in the run hypergraph 204 that relate the number of runs 208 on an upstream node 304 to the number of runs on a downstream node. In some embodiments, the run constrain 232 may or may not be specified explicitly in terms of run counts.
4) The user executes a run enumeration procedure thus building out the run hypergraph 204. In phase 1, which is optional, a first analysis algorithm checks existing runs for matching parameter combinations in the DOE design matrix (e.g., in the plurality of parameter combinations 228). The algorithm associates the runs to matching parameter combinations (if found). In phase 2, a run enumeration algorithm generates a set of runs on the entire process that satisfy (A) the DOE design matrix parameter combinations (the plurality of parameter combinations 228 and (B) the plurality of run constraints 230. In phase 3, a second analysis algorithm checks final set of runs for balance (specifically, it checks if each DOE parameter combination in the plurality of parameter combinations 228 is represented the same number of times in the final run set). The algorithm tags a balanced subset of runs. In phase 4, a run pruning algorithm attempts to find the most parsimonious run set by combining runs with shared parameter values while still satisfying the DOE design matrix (the plurality of parameter combinations 228) and the plurality of run constraints 230. A user can modify runs. In some embodiments, any change in a number of runs, run constraint 232, or node addition/removal will cause re-run of the run enumeration algorithm to ensure the DOE (plurality of parameter combinations 228) and run constraints 232 are satisfied. In some embodiments, a user can turn-off automatic re-run, or set the mode to ‘check only’ which will run only phase 1 of the enumeration algorithm only.
All references cited herein are incorporated herein by reference in their entirety and for all purposes to the same extent as if each individual publication or patent or patent application was specifically and individually indicated to be incorporated by reference in its entirety for all purposes.
The present invention can be implemented as a computer program product that comprises a computer program mechanism embedded in a nontransitory computer readable storage medium. For instance, the computer program product could contain the program modules shown in any combination of
Many modifications and variations of this invention can be made without departing from its spirit and scope, as will be apparent to those skilled in the art. The specific embodiments described herein are offered by way of example only. The embodiments were chosen and described in order to best explain the principles of the invention and its practical applications, to thereby enable others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. The invention is to be limited only by the terms of the appended claims, along with the full scope of equivalents to which such claims are entitled.
This application claims priority to U.S. Provisional Application No. 62/184,556, filed Jun. 25, 2015, entitled “Computer-Implemented Method for Recording and Analyzing Scientific Test Procedures and Data,” which is hereby incorporated by reference.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/US2016/039227 | 6/24/2016 | WO | 00 |
| Number | Date | Country | |
|---|---|---|---|
| 62184556 | Jun 2015 | US |
