Mapping Method for Microfluidic Biochips under Distributed Channel Storage

Abstract

The present invention relates to a mapping method for a microfluidic biochip under distributed channel storage. The mapping method is based on the following two effective strategies: 1) Application mapping: Map biochemical reactions onto a given chip architecture to minimize the total biochemical reaction time and satisfy dependency, resource, and wiring constraints; 2) Consider distributed channel storage: Comprehensively consider the priority value of operations and the transportation time, and analyze the impact of the storage scheduling of intermediate liquids on the total biochemical reaction time. The present invention can effectively reduce the total biochemical reaction time of the microfluidic biochip, thereby improving the execution efficiency of the microfluidic biochip.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer-aided design of flow-based microfluidic biochips, and particularly relates to a mapping method for microfluidic biochips with distributed channel storage. Background Art

[0002] Continuous-flow microfluidic biochips have become potential low-cost and fast-response lab-on-a-chip platforms. They have attracted much attention for their ability to automatically perform various biochemical applications simultaneously within a chip area the size of a coin. To further improve the execution efficiency and reduce the manufacturing cost, a distributed channel storage architecture can be considered, enabling the same channels to switch between transport and storage roles. Thus, fluid transport, caching, and extraction can be performed when passing through different flow paths. Such planning needs to be carefully considered during the mapping process from biochemical applications to a given biochip architecture. On the other hand, a scheduling scheme for biochemical operations is also needed to systematically consider the transport paths, caching locations, and corresponding execution durations to reduce the execution time of bioassays and efficiently execute all fluid handling tasks, thereby accelerating the implementation of bioassays. Summary of the Invention

[0003] The purpose of the present invention is to provide a mapping method for microfluidic biochips with distributed channel storage, which can effectively reduce the total biochemical reaction time of microfluidic biochips, thereby improving the execution efficiency of microfluidic biochips.

[0004] To achieve the above purpose, the technical solution of the present invention is: a mapping method for microfluidic biochips with distributed channel storage, including the following two strategies:

[0005] A. Distributed channel storage:

[0006] Replace the memory with channels, enabling the channels to have the dual functions of storage and transmission; comprehensively consider the priority value of operations and the transport time, and analyze the impact of the storage scheduling of intermediate fluids on the total biochemical reaction time.

[0007] B. Application mapping:

[0008] Map the biochemical reactions onto a given chip architecture, with the objective function of minimizing the total biochemical reaction time and satisfying dependency, resource, and routing constraints.

[0009] Compared with the prior art, the present invention has the following beneficial effects: By selecting a better routing scheme and a better mapping scheme, the present invention effectively reduces the total biochemical reaction time and the total length of the channels. Brief Description of the Drawings

[0010] Figure 1 It is a schematic diagram of a microfluidic mixer.

[0011] Figure 2 It is the LoC architecture model.

[0012] Figure 3 It is a timing diagram.

[0013] Figure 4 It is an operation binding diagram.

[0014] Figure 5 It is operation scheduling. Detailed implementation manners

[0015] The following combines the accompanying drawings to specifically illustrate the technical solution of the present invention.

[0016] A method for mapping a microfluidic biochip under distributed channel storage according to the present invention includes the following two strategies:

[0017] A. Distributed channel storage:

[0018] Replace the dedicated memory with a channel so that the channel has the dual functions of storage and transmission; comprehensively consider the priority value of the operation and the transportation time, and analyze the influence of the storage scheduling of the intermediate fluid on the total biochemical reaction time;

[0019] B. Application mapping:

[0020] According to the timing diagram and the existing architecture of the biochip, determine the specific execution time of each biochemical reaction operation, the time interval for transporting the fluid for each fluid transportation task, as well as the channels, components, and ports used by the task, the channels used by each distributed storage task, and the time interval for storing the fluid using the channel. Map the biochemical reactions onto the given chip architecture, with the objective function of minimizing the total biochemical reaction time and satisfying the dependencies, resources, and wiring constraints.

[0021] The following is the specific implementation process of the method of the present invention.

[0022] 1. Component model:

[0023] First, each component can be characterized using a two-layer modeling framework consisting of a flow layer model and a control layer model.

[0024] Flow layer model: The flow layer model L=(P, C, H) of each component is characterized by a set of operation phases P, execution time C, and geometric dimensions H. The following are the flow layer model components of eight common microfluids. The geometric dimensions H are also given in terms of length and width and scaled proportionally, with the unit length being 150 μm (for example, length 10 corresponds to 1500 μm), and the different operation phases of each component are listed: For some components, these phases must be serialized, as in the case of the mixer.

[0025] (1) Mixer:

[0026] Phase (P): Input 1 / Input 2 / Mixing / Output 1 / Output 2

[0027] Operation execution time: 0.5s

[0028] H: 30×30

[0029] (2) Sieve:

[0030] Phase (P): Input / Filter / Output 1 / Output 2

[0031] Operation execution time: 20s

[0032] H: 120×30

[0033] (3) Detector:

[0034] Phase (P): Input / Detection / Operation

[0035] Operation execution time: 5s

[0036] H: 20×20

[0037] (4) Separator:

[0038] Phase (P): Input 1 / Input 2 / Separation / Output 1 / Output 2

[0039] Operation execution time: 140s

[0040] H: 70×20

[0041] (5) Heater:

[0042] Phase (P): Input / Heating / Operation

[0043] Operation execution time: 20℃ / s

[0044] H: 40×15

[0045] (6) Meter:

[0046] Phase (P): Input / Metering / Output 1 / Output 2

[0047] H: 30×15

[0048] (7) Multiplexer:

[0049] Phase (P): Input / Output

[0050] H: 30×10

[0051] (8) Memory:

[0052] Phase (P): Input / Output

[0053] H: 90×30

[0054] Control layer model: When the flow layer model captures the high-level behavior of each component, the control layer model captures the micro-valve drive details required to operate it. As an example, the following is the control layer model of a mixer, showing the state of each micro-valve in each operating state. Here, 0 represents the valve being closed, 1 represents the valve being off, and 0 represents the valve being open.

[0055] Input 1 stage: (v1, v2, v3, v4, v5, v6, v7, v8, v9) = (0, 0, 1, 0, 0, 0, 0, 0, 1)

[0056] Input 2 stage: (v1, v2, v3, v4, v5, v6, v7, v8, v9) = (0, 1, 0, 0, 0, 0, 1, 0, 0)

[0057] Mixing stage: (v1, v2, v3, v4, v5, v6, v7, v8, v9) = (1, 0, 0, mixing, mixing, mixing, 0, 1, 0)

[0058] Output 1: (v1, v2, v3, v4, v5, v6, v7, v8, v9) = (0, 0, 1, 0, 0, 0, 0, 0, 1)

[0059] Output 2: (v1, v2, v3, v4, v5, v6, v7, v8, v9) = (0, 1, 0, 0, 0, 0, 1, 0, 0)

[0060] Figure 1 It is a microfluidic mixer. During the "mixing" state, the peristaltic drive sequence of the valve group v4, v5, v6 is dynamic. The microfluidic very large scale integration chip laboratory is usually controlled by a host PC, which is used to issue control signals with the granularity of the control layer model. The host can also perform data acquisition and signal processing operations as needed.

[0061] 2. Architecture model:

[0062] A microfluidic very large scale integration chip laboratory architecture is modeled as a topological graph (or netlist) A = (N, D). For example, as Figure 2 shown, where N = M ∪ S is a set of vertices (M is the set of components, S is the set of switches), and D is a finite set of edges representing fluid channel segments. In principle, a channel fluid channel segment can support fluid transportation in either direction. However, in some cases, due to the imperfection of the objective condition route, the flow direction may indirectly affect the correctness of the biometric measurement. Therefore, we represent each directional channel segment with a directed edge, and use a pair of directed edges d i,j and d j,iRepresents each bidirectional channel segment. The bidirectional fluid channel segment must satisfy the constraint that fluid cannot flow through d i,j and d j,i simultaneously.

[0063] The flow path P i,j represents a subset of one or more directed edges of the edge set D, indicating a directed channel between any two vertices n i , n j ∈ N using a series of directed edges in D (e.g., in Figure 3 , represents a flow path from vertex input 1 to vertex mixer 1). Fluid transmission using the flow path is similar to a circuit-switched network: the entire flow path is reserved until the fluid transmission is completed (e.g., until the fluid reaches mixer 1).

[0064] 3. Biochemical application model:

[0065] We usually use the timing diagram G = (O, E) to model biochemical applications. It is directed, acyclic, and polar, that is, there is a source vertex without predecessors (a vertex with in-degree 0, i.e., the input liquid) and a sink vertex without successors (a vertex with out-degree 0, i.e., the generated experimental results). Figure 3 shows an example. Each vertex o i ∈ O represents an operation that will be scheduled and bound to an architecture component in M, represented by the binding function F Op : O → M. The latency of o i is a known constant when bound to m j . The edge set E models the dependency constraints in the bioassay, that is, the edge e i from o j to o i,j ∈ E indicates that the fluid output of o i is then input to o j . If the fluid output from o i cannot be immediately used by o j (e.g., it has to wait for another operation to complete on the component F Op (o j ), then it must be stored in a "storage unit".

[0066] Each input phase of an operation corresponding component has an input edge, and each output phase has at most one edge; if the number of its output edges is less than the number of output phases, the remaining fluid is transported to the waste liquid port. Before an operation can be executed, all inputs of a given operation must reach its component. All output edges of the source vertices in the scheduling graph must also be bound to input containers (to ensure the correct fluid is input into the device), and all input edges to the sink vertices must also be bound to output or waste liquid ports. And we assume in advance that each input operation dispenses the correct volume of fluid through metering.

[0067] 4. Problem statement:

[0068] Given a library L of characterized microfluidic very large scale integration components, a microfluidic very large scale integration chip laboratory architecture A, and a bioassay G modeled as a timing graph, we wish to synthesize an implementation Ψ = <T, F>, where T is the schedule and F is the binding. Specifically, T = <T Op , T Path >, where T Op is the operation plan (the set of operation start times), and T Path is the plan for fluid transportation. Similarly, F = <F Op , F Path >, where F Op is the binding of operations to components, and F Path is the binding path for fluid transportation. The goal of the scheduler is to minimize the bioassay completion time, denoted as δ G , and each operation o i has a corresponding preparation time t ready (o i ), start time t start (o i ), and end time t finish (o i ).

[0069] The preparation time is determined based on the earliest time when o i can be executed and the completion time of the predecessors of o i . The start and end times of o i are calculated from the schedule. It follows that t ready (o i ) ≤ t start (o i ) < t finish (o i ). Each dependency edge e i,j can be associated with a fluid transportation operation. If F Op (o i ) ≠ F Op (o j ), then oi and o j are bound to different components and a fluid transportation operation is required to transfer fluid from o i to o j , whose start and end times are t start (e i,j ) and t finish (e i,j ); The fluid transportation operation is not yet ready for sequencing. If F Op (o i ) = F Op (o j ), then o i and o j are bound to the same component and no fluid needs to be transported.

[0070] The schedule needs to meet the following constraints.

[0071] a. Operability: Each operation o i is bound to a component m j = F Op (o i ), and m j is able to execute o i .

[0072] b. Dependency: For each dependency edge e i,j , the operation o i must be completed first, followed by a complete fluid transfer (if required), and then the operation o i is started, i.e.: t finish (o i ) ≤ t start (e i,j ) ≤ t finish (e i,j ) ≤ t start (o j ).

[0073] c. Resources: No two operations can be bound to the same component at the same time, i.e., if F Op (o i ) = F Op (o j ) then t finish (o i ) ≤ t start (o j ) or vice versa.

[0074] d. Wiring: There cannot be two flow paths bound to the same component at the same time.

[0075] 5. List Scheduling Algorithm Process

[0076] The biochemical reaction application mapping based on list scheduling designed is represented by LSAM, and its process is simplified into three stages.

[0077] In the first stage, first traverse all eligible components m ∈ M and calculate the earliest time t when m can be used to perform the next operation. o j Bind to m', whose available time t' is the smallest among all components. In Figure 4 it can be obtained that o3 is bound to component Heater 1.

[0078] The first stage also binds the fluid transfer operation to the wiring path, including: (1) removing the unused fluid from m' j and (2) transferring the used fluid (from other places on the chip to m' (discussed in more detail in the section on flow path scheduling and binding). j

[0079] In the second stage, then calculate the latest time when all the input fluids of o j arrive at component m', denoted as t". Obviously t" ≥ t' because the fluids cannot arrive at m' before m' is ready to receive them.

[0080] The second stage first arranges each fluid dependency e j of o i,j e i,j that was bound to the flow path P i,j during the previous iteration process, which schedules and binds o j . Then it is to schedule the fluid transportation operation, generating the start and end times t start (e i,j ) and t finish (e i,j ). Then if t finish (e i,j ) > t", then update t". After arranging each necessary fluid transfer to m', the second stage sets t start (o j ) to t" and arranges o j to start executing o start (o j ) on m' at t j . In Figure 5 the fluid arrives at Heater 1 from Component Input 5 through Filter 1. Once it arrives (at 7.21 seconds), the heating operation o3 is scheduled on Heater 1.

[0081] If o k is ready to be scheduled, the third stage adds the successor o j of o k to the queue Q; the ready time t of o k ​​ready (o k ) is initialized to the latest completion time among its predecessors.

[0082] (1) Operation binding

[0083] In the first stage, when determining the time t at which the eligible component m is available to perform the operation o j , three cases need to be considered:

[0084] Case 1: The function op(m) refers to the operation that was most recently scheduled and bound to the component m. This can be regarded as a reverse binding function, where the time interval for scheduling the operation is implicit. It produces an output fluid that is input to o j . o j must wait for the operation op(m) of the bound component m to complete, and the required fluid already exists, so m is ready.

[0085] Case 2: op(m) generates at least one fluid that is not input to o j . This fluid must be transferred to the memory, releasing m to perform o j .

[0086] The fluid transfer delay is not until later when the iteration binds the storage operation to the eligible storage component. This constrained decision is made only when o j is bound to m. The future binding decision is not made a priori, but the wiring delay is estimated. In this case, t depends on t finish (op(m)) plus the estimated wiring overhead.

[0087] Case 3: m has not performed any operation and contains no fluid; t is estimated as t ready (o j ) because the component does not restrict scheduling.

[0088] (2) Path scheduling and binding

[0089] Consider the fluid dependency e i,j ∈ E, and let F Op (o i ) = m i and F Op (o j ) = m j . t finish (o j ) is known from the previous iteration of the basic list scheduling algorithm: t start (o j ) cannot be known until the delay of the fluid transportation operation for delivering the input is determined.

[0090] The first step is to take e i,jBound to the flow path. Simply finding the path from component m i to m j in the architecture is not enough: an input port m 输入 needs to be selected to provide the buffer and an output port m 输出 to output, and connected to three sub-paths: P i,j = p 输入,i ·p i,j ·p j,输出 . The present invention uses the Dijkstra algorithm to calculate the required flow path. First, the architecture diagram is augmented with a super source connected to each buffer input and a super sink connected to each output. The Dijkstra algorithm is called three times to calculate the three sub-paths:

[0091] a. From the super source to m i , which implicitly selects an input port m 输入 in (removing the super source from this sub-path results in p 输入,i )

[0092] b. The sub-path p i from m j to m i,j ;

[0093] c. From m i to the super sink, which implicitly selects the output port m 输出 to output (removing the super sink from this sub-path results in p j,输出 ). Then the three flow paths are connected to form P i,j . The Dijkstra algorithm is limited to avoiding components or channels that are currently contaminated or otherwise in use. Passing through components, certain components should be avoided if they affect the bioassay function. For example, passing through a heater may affect the temperature of the fluid, or passing through a filter may affect its composition;

[0094] On the other hand, passing through a clean and inactive mixer does not change the fluid. Once the flow path is calculated, the next step is to determine the start and end times of fluid transfer, t start (e i,j ) and t finish (e i,j ). t start (e i,j ) is the first time when the following three conditions are met:

[0095] a. m i no longer performs operations. b. m j has not performed operations and has discharged the fluid. c. All channel segments d k,l ∈P i,jAll are available. The formula for calculating this time point is as follows:

[0096] t start (e i，j ) = max(t finish (o i ), t ready (o j ), {t finish (d k，l ) | d k，l ∈P i，j} ) (1)

[0097] Where t finish (d k,l ) is the completion time of any previous plan, that is, t start (e i,j ) plus the routing delay t path (P i,j ). The completion time of the flow path is given by the following formula:

[0098] t finish (e i，j ) = t start (e i，j ) + t path (P i，j ) (2)

[0099] 6. Distributed Storage-Driven Algorithm

[0100] Previous studies on LSAM were based on biochips with dedicated storage units, and the great flexibility of the flow channels in the fluid cache was still ignored. Therefore, further consider distributed storage-driven, utilize the flow channels in the fluid cache, and represent the biochemical reaction application mapping algorithm after considering distributed storage-driven as SDAM. The specific algorithm process is as follows.

[0101] First, each operation o i ∈O is associated with a ready time t ready (o i ), representing the time point when the last parent operation has been completed. Each component m ∈ M is also associated with a ready time t ready (m k ), indicating the time point. Each edge e i,j ∈E may be associated with a transportation task and is bound to the flow path p in the biochip.

[0102] Next, the Dijkstra algorithm is adopted to pre-compute the required flow paths starting from the input ports and ending at the output / waste ports at each edge. The number of flow channels d used to perform the transportation tasks is initialized to 0. The priority value for each operation is calculated, which is specified as the path with the largest weight from the source to the sink in the sorting graph. A higher priority value indicates that the transportation task associated with edge e i,j is more likely to be executed in advance.

[0103] After that, the operations that are ready to be executed are inserted into the queue and processed in the order of their priority values. For the operation o i that has just dequeued from the queue, it is bound to a component m. If the operation that was recently scheduled and bound to component m generates at least one fluid that is not input to o i , the fluid must be cached in the channel segment for release. Two strategies are considered to cache the intermediate fluid:

[0104] (1) Assume that the intermediate fluid is the input of operation m a and has the highest priority value among all the operations to be performed in the bound component. In this case, the intermediate fluid can be directly transported to component m a .

[0105] (2) Assume that component m b is the target component for the intermediate fluid that needs to be cached. We select an idle channel segment that does not conflict with other transportation tasks and is close to component m b .

[0106] Next, for the bound component m, we calculate the ready time t ready (m), which represents the time point when the component is available. To start executing the operation when the corresponding component is ready, all the input fluids should be transported to the corresponding component.

[0107] Finally, once all the input fluids of the operation are transported to the corresponding component, the execution can start, and the exact start time and end time can be calculated, thus obtaining the execution time of the biochemical reaction.

[0108] The above are the preferred embodiments of the present invention. All changes made according to the technical solutions of the present invention, when the functions and effects produced do not exceed the scope of the technical solutions of the present invention, shall fall within the protection scope of the present invention.

Claims

1. A mapping method for a microfluidic biochip under distributed channel storage, characterized in that, It includes the following two strategies: A. Distributed channel storage: Replace the memory with a channel so that the channel has the dual functions of storage and transmission; comprehensively consider the priority value of the operation and the transportation time, and analyze the impact of the storage scheduling of the intermediate fluid on the total biochemical reaction time; B. Application mapping: Map the biochemical reaction to a given chip architecture, with the minimization of the total biochemical reaction time as the objective function and satisfying the dependence, resource, and wiring constraints; The method is implemented as follows: S1. Build a component model; S2. Build an architecture model; S3. Build a biochemical reaction application model; S4. Problem formulation; S5. List scheduling; Among them, S1. Building a component model is as follows: Use a two-layer modeling framework composed of a flow layer model and a control layer model to characterize each component; S2. Building an architecture model is as follows: The microfluidic very large scale integrated chip laboratory architecture is modeled as a topological graph or netlist A = (N, D), where N = M ∪ S is a set of vertices, M is a set of components, S is a set of switches, and D is a finite set of edges representing fluid channel segments; each directed channel segment is represented by a directed edge, and each bidirectional channel segment is represented by a pair of directed edges d i,j and d j,i ; the bidirectional channel segment must satisfy the constraint that fluid cannot flow through d i,j and d j,i simultaneously; the flow path P i,j represents a subset of one or more directed edges of the edge set D, indicating a directed channel between any two vertices n i , n j ∈ N using a series of directed edges in D; S3. Building a biochemical reaction application model is as follows: Model the biochemical reaction application using the timing diagram G = (O, E), which is directed, acyclic, and polar, i.e., it has a source vertex with no predecessors and a sink vertex with no successors; each vertex o i ∈ O represents an operation that will be scheduled and bound to a component m j in M, represented by the binding function F Op : O → M; the latency of o i when bound to m j is a known constant; the edge set E models the dependency constraints in the bioassay, i.e., an edge e i from vertex o j to vertex o i,j ∈ E represents the fluid output of o i and then input to o j ; if the fluid output from o i cannot be used immediately by o j , it must be stored in a "storage unit", i.e., a channel that caches the fluid; each input phase of an operation corresponding to a component has an input edge, and each output phase has at most one edge; if its output edges are fewer than the number of output phases, the remaining fluid is transported to the waste port; all inputs of a given operation must arrive at its component before the operation can execute; all output edges of the source vertex in the timing diagram must also be bound to the input container, and all input edges to the sink vertex must also be bound to the output or waste port; and it is assumed in advance that each input operation correctly dispenses the correct volume of fluid by metering.

2. The mapping method of the microfluidic biochip under distributed channel storage according to claim 1, characterized in that, The method also includes: S4. Problem formulation is as follows: Given a library L of characterized microfluidic very large scale integration components, a microfluidic very large scale integration chip laboratory architecture A, and a biometric assay modeled as a timing diagram G, synthesize an implementation Ψ = <T, F>, where T is the schedule and F is the binding, T = <T Op , T Path >, where T Op is the set of operation plans, i.e., the operation start times, and T Path is the plan for fluid transfer, F = <F Op , F Path >, where F Op is the binding of operations to components and F Path is the binding path for fluid transfer; and the goal of the scheduler is to minimize the biometric assay completion time, denoted as δ G , and each operation o i has a corresponding preparation time t ready (o i ), start time t start (o i ) and end time t finish (o i ); The preparation time is based on the earliest time that can be executed and the completion time of the predecessors. i The start and end times of are calculated by the schedule, which follows i ( i ) ≤ t ready ( i ) < t start ( i ) < t finish ( i ); Each dependency edge e i,j can be associated with a fluid transfer operation; If F Op ( i ) ≠ F Op ( j ), then i and j are bound to different components, and a fluid transfer operation is required to transfer the fluid from i to j , and its start and end times are t start (e i,j ) and t finish (e i,j ); If F Op ( i ) = F Op ( j ), then i and j are bound to the same component and no fluid transfer is required; S5. List scheduling is as follows: Represent the designed biochemical reaction application mapping based on list scheduling with LSAM, and simplify its process into three stages: In the first stage, traverse all eligible components m ∈ M and calculate the earliest time t at which m can be used to perform the next operation, o j Bind to m', whose available time t' is the smallest among all components; The first stage also binds fluid transfer operations to the wiring path, including: (1) removing o from m'; j unused fluid, and (2) the transfer of j used fluid; In the second stage, calculate the latest arrival time of all the input fluids o j at the component m', denoted as t", where t" ≥ t' because the fluids cannot arrive at m' before m' is ready to receive them; In the second stage, first arrange o j for each fluid dependency e i,j , e i,j that was bound to flow path P during previous iterations i,j , which schedules and binds o j ; followed by scheduling the fluid transportation operation, generating start and end times t start (e i,j ) and t finish (e i,j ); then, if t finish (e i,j ) > t", update t"; after arranging each necessary fluid transfer to m', the second stage sets t start (o j ) to t" and schedules o j to execute o start (o j ) on m' starting from t j ; If o k is ready to be scheduled, the third stage will add o j 's successor o k to the queue Q; o k 's ready time t ready (o k ) is initialized to the latest completion time among its predecessors; S6. Distributed storage drive: The biochemical reaction application mapping after considering the distributed storage drive is represented as SDAM, and the specific process is as follows: First, each operation o i ∈ O is associated with a ready time t ready (o i ), representing the time point when the last parent operation has been completed. Each component m ∈ M is also associated with a ready time t ready (m k ), and each edge e i,j ∈ E is associated with a transportation task and is bound to the flow path p in the biochip; Next, the Dijkstra algorithm is adopted to pre-compute the required flow paths starting from the input ports and ending at the output / waste ports at each edge; the number of flow channels d for performing transportation tasks is initialized to 0; calculate the priority value for each operation, which is specified as the path with the largest weight from the source to the sink in the timing diagram; the higher the priority value, the earlier the transportation task related to the edge e i,j is executed. After that, the operations ready to execute are inserted into a queue and processed in the order of their priority values; for the operation o that has just dequeued from the queue i , which is bound to a component m, if the operation that was recently scheduled and bound to component m produces at least one fluid that has not been input to o i , the fluid must be cached in the channel segment for release; Next, for the bound component m, calculate the ready time t ready (m), which represents the time point when the component is available; to start the operation when the corresponding component is ready, all input fluids should be delivered to the corresponding component; Finally, once all the input fluids of the operation are transmitted to the corresponding components, the execution starts, and the exact start time and end time are calculated, so as to obtain the execution time of the biochemical reaction.

3. The mapping method of the microfluidic biochip under distributed channel storage according to claim 2, wherein Flow layer model: The flow layer model L=(P, C, H) of each component is characterized by a set of operation stages P, execution time C, and geometric dimensions H; Control layer model: When the flow layer model captures the high-level behavior of each component, the control layer model captures the micro-valve drive details required to operate it.

4. The mapping method of the microfluidic biochip under distributed channel storage according to claim 2, characterized in that The schedule needs to meet the following restrictions: a. Operability: Each operation o i is bound to a component m j = F Op (o i ), and m j is able to execute o i ; b. Dependency: For each dependency edge e i,j , operation o i must be completed first, followed by a complete fluid transfer, and then operation o j begins, i.e.: t finish (o i ) ≤ t start (e i,j ) ≤ t finish (e i,j ) ≤ t start (o j ); c. Resource: No two operations can be bound to the same component simultaneously, i.e., if F Op (o i ) = F Op (o j ) then t finish (o i ) ≤ t start (o j ); d. Wiring: There cannot be two flow paths bound to the same component simultaneously.

5. The mapping method of the microfluidic biochip under distributed channel storage according to claim 2, wherein In step S5, in the first stage, when determining the time t at which the qualified component m can be used to perform the operation o j three cases need to be considered: Case 1: The function op(m) refers to the operation that was most recently scheduled and bound to component m, which is considered a reverse-binding function, where the time interval for scheduling the operation is implicit; it produces an output liquid that is input to o j ; o j must wait for the operation op(m) of the bound component m to complete, and the required fluid already exists, so m is ready; Case 2: op(m) generates at least one liquid that is not input into o j ; This liquid must be transferred to the memory, releasing m to execute o j ; The fluid transfer delay does not bind the iteration storage operation to the qualified storage component until later; only when o j is bound to m will this constrained decision be made; the future binding decision is not made a priori, but the wiring delay is estimated; in this case, t depends on t finish (op(m)) plus the estimated wiring overhead; Case 3: m does not perform any operation and contains no liquid; t is estimated to be t ready (o j ) because the component does not restrict scheduling.

6. The mapping method of the microfluidic biochip under distributed channel storage according to claim 3, wherein In step S5, Consider fluid dependence e i，j ∈ E, and let F Op (o i ) = m i and F Op (o j ) = m j ; t start (o j ) cannot be known until the delay of the fluid transport operation that conveys the input is determined; The first step is to bind e i，j to the flow path; select an input port m 输入 to provide buffer solution and an output port m 输出 to output and connect it to three sub-paths: P i，j = p 输入，i ·p i，j ·p j，输出 ; First, expand the architecture diagram using a super-source connected to each buffer input and a super-sink connected to each output; call the Dijkstra algorithm three times to calculate the three sub-paths: a. From the super source to m i , which implicitly selects an input port m 输入 At, removing the super source from this subpath produces p 输入，i ; b. from m i to m j sub-path p i，j ; c. From m i to the super sink, it implicitly selects the output port m 输出 to go out, and deletes the super sink from this sub-path to generate p j，输出 ; then connects the three flow paths to form P i，j ; On the other hand, the fluid will not be changed by a clean and inactive mixer; Once the flow path is calculated, the next step is to determine the start and end times of fluid transfer, t start (e i,j ) and t finish (e i,j ); t start (e i,j ) is the first time the following three conditions are met: a.m i No further operation is performed; b.m j The operation is not performed and the fluid has been discharged; c. All channel segments d k，l ∈P i，j are all available; t start (e i，j ) The calculation formula is as follows: t start (e i，j ) = max(t finish (o i ), t ready (o j ), {t finish (d k，l ) | d k，l ∈ P i，j )} (1) where t finish (d k，l ) is the completion time of any previous schedule, i.e., t start (e i,j ) plus the routing delay t path (P i，j ) for the constrained fluid transport operation using the channel segment; the completion time of the flow path is given by: t finish (e i，j ) = t start (e i，j ) + t path (P i，j ) (2).

7. The mapping method of the microfluidic biochip under distributed channel storage according to claim 2, wherein In step S6, consider two strategies to cache the intermediate fluid: (1) Assume that the intermediate fluid is the input of operation m a and has the highest priority value among all the operations to be performed in the bound components; in this case, the intermediate fluid is directly conveyed to component m a ; (2) Assume component m b is the target component of the intermediate fluid that needs to be cached, and select an idle channel segment that does not conflict with other transportation tasks and is close to component m b .

Images