High-level synthesis method for continuous microfluidic biochip considering volume management

Abstract

The application provides a continuous micro-fluidic biochip high-level synthesis method considering volume management, is based on a discrete particle swarm algorithm with high solving efficiency, and comprises the following key technologies: 1) a particle coding scheme suitable for the high-level synthesis problem considering volume management is provided; 2) an automatic scheduling scheme generation method is provided, which can accurately generate a corresponding operation scheduling scheme according to the particle coding; and 3) according to the characteristics of the problem, a corresponding particle updating strategy is designed, and the convergence of the algorithm is accelerated. The application can obtain a high-level synthesis scheme satisfying volume constraints and having high quality in a shorter time.

Description

TECHNICAL FIELD

[0001] The present application relates to the continuous microfluidic biochip computer-aided design field, and particularly relates to a high-level synthesis method of a continuous microfluidic biochip considering volume management. BACKGROUND

[0002] The design of a continuous microfluidic biochip is divided into two stages: a high-level synthesis stage and a physical design stage. In the high-level synthesis stage, component allocation and operation scheduling are mainly performed, and in the physical design stage, component layout and routing are mainly performed. The work of the high-level synthesis stage affects the subsequent physical design stage and is related to the execution efficiency of biochemical reactions. In addition, with the advancement of microfluidic chip production technology, the integration of microfluidic chips is becoming higher and higher. Such progress makes the continuous microfluidic biochip can execute more and more biochemical operations at the same time. How to reasonably schedule the execution of biochemical operations becomes a key factor affecting the performance of the biochip. Therefore, high-level synthesis is a crucial step in the design of a continuous microfluidic biochip.

[0003] Previous work did not consider the volume factor of the transported fluid in the high-level synthesis stage, making it difficult to meet the actual biochemical operation requirements of the generated high-level synthesis scheme, thereby affecting the experimental results. In order to solve this problem, the existing work considers volume management in the high-level synthesis stage, but this work is based on integer linear programming, and the algorithm complexity is high, which is not sufficient to meet the large-scale integration requirements of the continuous microfluidic biochip. Therefore, there is an urgent need for an algorithm that can quickly generate a high-quality continuous microfluidic high-level synthesis scheme considering volume constraints. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a high-level synthesis method of a continuous microfluidic biochip considering volume management, which can obtain a high-quality high-level synthesis scheme meeting volume constraints in a shorter time.

[0005] To achieve the above-mentioned purpose, the present application adopts the following technical solution: a high-level synthesis method of a continuous microfluidic biochip considering volume management, analyzing the relationship between the variables generated in the design process, using the relationship between the variables, reducing the number of variables, and proposing a particle coding scheme suitable for the high-level synthesis problem considering volume management; an automatic scheduling scheme generation method is proposed, and the corresponding high-level synthesis scheduling scheme meeting the volume management is accurately generated according to the particle coding; according to the characteristics of the problem, a corresponding particle update strategy is designed, so that the particles learn the experience of individuals and populations in the iteration process, and accelerate the convergence of the algorithm.

[0006] In a preferred embodiment, the high-level synthesis stage determines the device binding scheme of the continuous microfluidic biochip according to the operation information of the biochemical reactions and the given device library, and then determines the time of each transport task, and determines the operation start execution time according to the reagent loading completion time; the operation end time is obtained according to the known operation start time and the operation time required; the biochemical reactions in the continuous microfluidic biochip use a timing diagram G(O, E) to describe the relationship between each operation; the timing diagram G(O, E) is a directed acyclic graph, wherein a node o i ∈O represents an operation in the biochemical reaction; an edge e i，j ∈E in the timing diagram G(O, E) represents the dependency relationship between two biochemical reaction operations o i and o j ; e i，j represents that the operation o i is the parent operation of the operation o j ; the operation o j must be executed after the operation o i is completed, and after the operation o i reaction is completed, the reagent of the reaction needs to be transported to the device bound by the operation o j as the input of the operation o j .

[0007] In a preferred embodiment, the minimum volume limit of single input of each device is set, and the volume of the reagent loaded into the device each time cannot be less than the minimum volume limit of single input of the device;

[0008] If there is a mixing operation o1 waiting to be executed, the operation will mix two reagents, each of 2 microliters; if there is a mixer d1 available at this time, the minimum volume limit of single input of the mixer d1 is 4 microliters; the operation o1 needs to continue to wait for a mixer with a minimum volume limit of single input less than or equal to 2 microliters.

[0009] In a preferred embodiment, three types of transport tasks need to be considered in the design process of the continuous microfluidic biochip CFMBs; the three types of transport tasks are as follows:

[0010] The first type of transport task: the transport task of reagents between devices;

[0011] The second type of transport task: the transport task of removing excess liquid;

[0012] The third type of transport task: the transport task of removing waste liquid;

[0013] Before loading the reagent into the mixer, it is necessary to determine whether there is any waste liquid left over from the previous device in the mixer. If so, the waste liquid is removed through the third type of transport task. Transporting the reagent into the mixer is the first type of transport task. When loading the reagent into the mixer, the amount loaded is not less than the amount that the mixer channel can hold. At this time, there will be some excess liquid at both ends of the mixer. If there is, the excess liquid is removed through the second type of transport task.

[0014] In the design of continuous microfluidic biochips (CFMBs), a separate flow path is planned for each transport task.

[0015] In a preferred embodiment, discrete particle swarm optimization is used to solve discrete problems. The input to the algorithm is a time sequence diagram of the current biochemical reaction, a provided device library, and an operation information table. The time sequence diagram provides the dependencies between various operations in the current reaction; the device library provides the basis for subsequent analysis of the operation scheduling scheme; and the operation information table provides the types of equipment required for the operation and the execution time.

[0016] During the iterative update phase, after updating each particle, it is necessary to re-analyze the operation scheduling scheme and fitness value corresponding to each particle; then update the individual historical best solution and the population historical best solution for each particle; finally, output the found optimal device binding and operation scheduling scheme.

[0017] The binding scheme is as follows: operation o1 is bound to device d1, and operation o2 is bound to device d2. Operation o1 has two parent operations and one child operation. The minimum input volume limit for device d1 is 2 μL. The two parent operations each load 2 μL of reagent into device d1. After the reaction of operation o1, a total of 4 μL of reagent can be obtained. Operation o1 has only one child operation, so the actual volume transported from operation o1 to operation o2 is 4 μL. If the minimum input volume limit for device d2 is 2 μL, then 2 μL of excess liquid also needs to be removed.

[0018] In a preferred embodiment, a timing diagram of the mixing phase in PCR is designed. The timing diagram includes nodes 1 to 7, representing operations o1 to o7, and nodes 8 to 15, representing the input of liquid samples s1 to s8 from the outside; it also includes node 16, representing the output of reagents s9 from the outside; and the position vector X of the particles in the particle swarm is... i The encoding is (Y, B); where the priority vector Y = (y1, y2, y3, ..., y4) is... n ), where n represents the total number of nodes in the sequence diagram, y i This indicates the priority of the node labeled i; when the number of operands is n and the number of devices is m, the binding vector B = (b 1，1 b 1，2 , ..., b 1，m b 2，1 , ..., bn，m ) ; b i，j is a 0-1 variable, indicating whether the operation o i is bound to the device d j ; if the operation o i is executed in the device d j , then b i，j = 1; if the operation o i is not executed in the device d j , then b i，j = 0; using this coding mode can realize fast updating of particles, and the operation scheduling scheme can be analyzed according to the priority vector and the binding vector.

[0019] In a preferred embodiment, the operation can start to be executed when the parent operations of the operation are all executed and the bound device is in an idle state; whether there is an operation that can be executed is constantly detected, and if there is, the operation is immediately executed; if there are operations bound to the same device and satisfying the execution condition at the same time, the operation with a larger priority is executed first;

[0020] To execute the operation o i , the device d i bound by the operation o j is found; from the time when the device is in an idle state, the operation o i can use the device d j to perform reagent removal work, and waiting for subsequent reagent loading can start to be executed; the end time of the execution of the operation o i is the time when the device d j can be used by the next operation; after updating the time, the device d j is handed over to the next operation for use; after analyzing all operations, the operation scheduling scheme corresponding to the particle is obtained;

[0021] According to the scheduling scheme, the time T a required to complete the biochemical reaction and the total number N p of flow paths that need to be planned are determined; the fitness value F i of the particle X i is calculated through formula (1):

[0022] F i = α × T a + β × N p (1)

[0023] Wherein, α and β are two weighted factors.

[0024] In a preferred embodiment, the operation scheduling scheme of high-level synthesis is solved by using discrete particle swarm optimization DPSO; for the particle coding mode of this time, the updating strategy is as follows:

[0025] In the iteration process, the particles generate new particles by using the following formula:

[0026]

[0027] wherein, r1, r2 are random numbers generated in each update process, the value of r1, r2 is in the interval [0, 1); r is a random number generated in each update process, the value of r is in the interval [0, 1); F1 represents the velocity update of the particle, F2 represents the perception and learning of the particle to the individual experience, and F3 represents the perception and learning of the particle to the population experience; represents the i-th particle at the t+1-th iteration; represents the i-th particle at the t-th iteration;

[0028] The mutation strategy of the particle is as follows:

[0029]

[0030] wherein, w is the probability threshold of mutation; when the generated random number r is less than the probability threshold, the particle is mutated to obtain a new particle; first, two random integers i and j in the interval [1, n] are randomly generated, n represents the number of nodes in the time sequence diagram of biochemical analysis; the values of y i and y j are exchanged to obtain the new particle after mutation;

[0031] The perception and learning of the particle to the individual experience are as follows:

[0032]

[0033] wherein, c1 is the probability threshold of the perception and learning of the individual experience; when the generated random number r1 is less than the probability threshold in each update process, the particle retains the priority vector Y of itself and learns the binding vector B of the individual historical optimal solution; the binding vector B of the particle X i indicates the binding relationship of three operations and three devices, and the individual optimal solution is taken as the learning target; if a randomly generated integer is 2, the binding vector B of the particle X i learns the binding relationship of the 2nd operation and device from the learning target, i.e., learns b 2，1 , b 2，2 , and b 2，3 ; after learning, it is needed to judge whether the current binding relationship can satisfy the constraint of volume management, if not, the binding relationship of the 3rd operation is continuously learned; the binding vector B' satisfying the constraint of volume management is obtained through continuous learning; the binding vector B' and the priority vector Y of the particle Xi are combined to obtain the new particle X' i =(Y, B') after learning;

[0034] The perception and learning of the particle to the population experience are as follows:

[0035]

[0036] Wherein, c2 is the probability threshold of perception and learning of population experience; when the random number r2 produced in each updating process is less than the probability threshold, the particle retains its priority vector Y, learns the binding vector B of the historical optimal solution of the population, and the learning method is consistent with the experience of the above-mentioned learning individual.

[0037] Compared with the prior art, the present application has the following beneficial effects: an algorithm for quickly generating a high-quality continuous microfluidic high-level synthesis scheme considering volume constraints. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 The flow chart of the algorithm of the preferred embodiment of the present application;

[0039] Figure 2 The timing chart of the redesigned mixing stage in PCR of the preferred embodiment of the present application;

[0040] Figure 3 The flow chart of the analysis function of the preferred embodiment of the present application;

[0041] Figure 4 The flow chart of the analysis of the operation to be performed of the preferred embodiment of the present application;

[0042] Figure 5 The initial value of the binding vector of the particle and the learning target of the preferred embodiment of the present application;

[0043] Figure 6 The first learning condition of the particle of the preferred embodiment of the present application;

[0044] Figure 7 The second learning condition of the particle of the preferred embodiment of the present application. DETAILED DESCRIPTION

[0045] The present application will be further described below in conjunction with the drawings and embodiments.

[0046] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as generally understood by those skilled in the art to which the present application belongs.

[0047] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; as used herein, the singular form "a", "an" and "the" include plural references unless the context clearly dictates otherwise. Furthermore, it is to be understood that the term "comprising" and / or "including" used herein is meant to be inclusive and not exclusive.

[0048] The present application specifically comprises the following design process:

[0049] (1) Analyze the relationship between the variables generated by the design process, use the relationship between the variables to reduce the number of variables, and propose a particle coding scheme suitable for high-level synthesis problems considering volume management.

[0050] (2) An automatic scheduling scheme generation method is proposed, which can accurately generate the corresponding scheduling scheme of high-level synthesis that meets the volume management according to the particle coding.

[0051] (3) According to the characteristics of the problem, the corresponding particle update strategy is designed, which enables the particle to effectively learn the experience of individuals and populations in the iteration process, accelerating the convergence of the algorithm.

[0052] 1. High-level synthesis:

[0053] The design of the continuous microfluidic biochip is divided into two stages: the high-level synthesis stage and the physical design stage. The first stage, the high-level synthesis stage, determines the device binding scheme of the continuous microfluidic biochip according to the operation information of biochemical reactions and the given device library, and then determines the time of each transportation task. According to the time of reagent loading completion, the operation execution time is determined. Knowing the start time of the operation and the time required for the operation, the end time of the operation can be obtained. The second stage, the physical design stage, first determines the placement position of the devices in the device library, and then performs wiring according to the flow path information in the high-level synthesis to build flow channels for the flow path.

[0054] The present application mainly studies the high-level synthesis stage of the continuous microfluidic biochip design. The problem of high-level synthesis is described as follows:

[0055] (1) Input: timing diagram, device library.

[0056] (2) Output: Device binding scheme and operation scheduling scheme of the continuous microfluidic biochip.

[0057] (3) Optimization goal: 1) Minimize the completion time of biochemical analysis. 2) Minimize the total number of flow paths

[0058] Biochemical reactions in continuous microfluidic biochips are usually described by a time graph G(O, E) to represent the relationship between each operation. The time graph G is a directed acyclic graph, where the nodes o i ∈ O represent an operation in the biochemical reaction. An edge e i，j ∈ E in the time graph G represents the dependency between two biochemical reaction operations o i and o j . e i，j indicates that the operation o i is the parent operation of the operation o j . The operation o j must be executed after the operation o i is completed, and the reaction is finished after the operation o i . The reagent of the reaction needs to be transported to the device bound to the operation o j as the input of the operation o j .

[0059] 2. Volume management

[0060] Since the channels in the device are not in a vacuum state, if there is still air in the channel of the device after the reagent is loaded, it may cause the operation reaction to be incomplete, affecting the accuracy of the experimental results. To solve this problem, the minimum volume limit of each device is set, and the volume of the reagent loaded into the device each time cannot be less than the minimum volume limit of the single input of the device, so as to ensure that the air in the channel is completely pushed out each time, and the operation can fully react.

[0061] If there is a mixing operation o1 waiting to be executed, the operation will mix two reagents, each 2 microliters. If there is a mixer d1 available at this time, the minimum volume limit of the single input of the mixer d1 is 4 microliters. Without considering volume management, in order to ensure the efficiency of the biochemical reaction, the mixer d1 will generally be assigned to the operation o1 for use. But once the operation o1 is executed in the mixer d1, there will be a lot of air in the mixer during the execution process, which cannot guarantee that the reagents can be fully mixed, which may affect the experimental results. If volume management is considered, the operation o1 needs to continue to wait for a mixer with a single input minimum volume limit less than or equal to 2 microliters.

[0062] 3. Transport tasks

[0063] In the design process of the continuous microfluidic biochip CFMBs, three types of transport tasks need to be considered in the case of considering volume management. The three types of transport tasks are as follows:

[0064] The first type of transport task: the transport task of reagents between devices.

[0065] The second type of transport task is a transport task of removing extra liquid.

[0066] The third type of transport task is a transport task of removing waste liquid.

[0067] For example, before loading reagents into a mixer, it is necessary to determine whether there is waste liquid left by the previous device in the mixer, and if there is, the waste liquid is removed by the third type of transport task. The transport of reagents to the mixer is the first type of transport task. Since the channel in the mixer is not in a vacuum state, if there is still air in the channel during the mixing operation, it may cause the reagents to be mixed insufficiently, affecting the accuracy of the experimental results. Therefore, in order to remove the air in the channel of the mixer, the amount of reagents loaded into the mixer is not less than the capacity of the channel of the mixer at this time, and there may be some extra liquid at both ends of the mixer, which is removed by the second type of transport task if it exists.

[0068] If different reagents appear on the same flow channel at the same time, it may cause contamination of the reagents. Therefore, in the design of the continuous microfluidic biochip CFMBs, a separate flow path needs to be planned for each transport task. In a biochemical reaction process, reagents need to be transported to the target device multiple times, and because the number of devices in the chip is limited, multiple operations often bind to the same device, so in the design process of the continuous microfluidic biochip CFMBs, a large number of flow paths required by transport tasks need to be planned.

[0069] 4. Algorithm flow

[0070] The device binding and operation scheduling scheme solved by the application belongs to a discrete problem, so a discrete particle swarm is selected to solve the problem. The algorithm flowchart of the application is shown in Figure 1 The optimization goal of the algorithm of the application is to minimize the biochemical analysis completion time and minimize the total number of flow paths.

[0071] The input of the algorithm is the timing diagram of the biochemical reaction performed this time, the provided device library and the operation information table. The timing diagram can provide the dependency relationship between the operations for this reaction. The device library provides the basis for subsequent analysis of the operation scheduling scheme, and the operation information table provides the type of device required by the operation and the execution time.

[0072] In the iteration update phase, after updating each particle, the operation scheduling scheme and the fitness value corresponding to each particle need to be analyzed again. Then the individual historical optimal solution and the population historical optimal solution of each particle are updated. Finally, the best device binding and operation scheduling scheme found is output.

[0073] Because the minimum input volume of the devices provided by the device library is known, and the timing diagram also provides the direction of reagent transport, the volume of reagent transported each time can be calculated once the binding scheme is determined. For example, the binding scheme is that operation o1 is bound to device d1, and operation o2 is bound to device d2. Operation o1 has two parent operations and one child operation. The minimum input volume limit for device d1 is 2 μL. The two parent operations each load 2 μL of reagent into device d1, and a total of 4 μL of reagent can be obtained after the reaction of operation o1. Operation o1 has only one child operation, so the actual volume transported from operation o1 to operation o2 is 4 μL. If the minimum input volume limit for device d2 is 2 μL, then 2 μL of excess liquid also needs to be removed.

[0074] If the volume of reagent transported each time can be determined, it can be judged whether the reagent volume of all transport tasks corresponding to the device's binding scheme meets the volume management conditions. To solve the reagent volume problem of continuous microfluidic biochips, it is necessary to assign binding schemes that meet volume management constraints to particles during the initialization phase, and also ensure that the reagent volume of all transport tasks corresponding to the updated binding scheme meets the volume management conditions when learning from individual and population experiences during the update phase.

[0075] 5. Particle encoding

[0076] To determine the scheduling scheme for the biochemical reaction, it is necessary to determine the binding relationship between each operation and equipment, the start and end times of each operation, and the start and end times of the three types of transportation tasks.

[0077] If all these variables are included in the particle's position vector X i In this process, the particle dimensionality becomes larger, resulting in slower update speeds. Furthermore, the variables are interconnected; for example, the start time of an operation must be later than the completion of loading all input reagents for that operation. It's difficult for initialized or updated particles to ensure that all variables satisfy the expected relationships.

[0078] Redesign the timeline of the mixing phase in PCR, such as... Figure 2 As shown, nodes 1 to 7 in the timing diagram represent operations o1 to o7, and nodes 8 to 15 represent the input of liquid samples s1 to s8 from the outside. Node 16 represents the output of reagent s9 from the outside. Based on the redesigned timing diagram, the position vector X of the particles in the particle swarm is... i The encoding is (Y, B). Where the priority vector Y = (y1, y2, y3, ..., y...). n ), where n represents the total number of nodes in the sequence diagram, y i This indicates the priority of the node labeled i. When there are n operands and m devices, the binding vector B = (b... 1，1 b 1，2 , ..., b1，m , b 2，1 , …, b n，m ). b i，j is a 0-1 variable, indicating whether the operation o i is bound to the device d j . If the operation o i is assigned to be executed in the device d j , then b i，j = 1. If the operation o i is not assigned to be executed in the device d j , then b i，j = 0. Using this encoding, the particle can be updated quickly, and the operation scheduling scheme can be quickly analyzed according to the priority vector and the binding vector.

[0079] 6. Analysis method of scheduling scheme

[0080] For the particle encoding method of the present application, an analysis method is designed to analyze the particle into an operation scheduling scheme, to analyze the priority of the operation, the binding device, and then to analyze in combination with the timing diagram, as shown in Figure 3 When the parent operations of the operation are all executed and the binding device is in an idle state, the operation can start to be executed. It is constantly detected whether an operation can be executed, and if so, the operation is immediately executed. If operations bound to the same device simultaneously satisfy the execution condition, the operation with a higher priority is executed first.

[0081] The currently executable operations are analyzed one by one, as shown in Figure 4 To execute the operation o i , the device d i bound to the operation o j is found first. From the time when the device is in an idle state, the operation o i can use the device d j to perform the reagent removal work, and after waiting for the subsequent reagent loading, the operation o i can start to be executed. The execution end time of the operation o j is the time when the device d i is available for use by the next operation. After updating the time, the device d j is handed over to the next operation for use. After analyzing all the operations, the operation scheduling scheme corresponding to the particle is obtained.

[0082] According to the scheduling scheme, the time T a required to complete the biochemical reaction and the total number N p of flow paths to be planned can be determined. The fitness value F i of the particle X i can be calculated by formula (1):

[0083] F i = α × T a + β × Np Equation (1)

[0084] where a, b are two weighted factors.

[0085] 7. Update strategy

[0086] In the literature

[17] , the discrete particle swarm DPSO is used to solve the component layout solution of the flow physical design. The present invention uses the discrete particle swarm DPSO to solve the operation scheduling scheme of the high-level synthesis. For the particle coding mode, the appropriate update strategy is designed as follows.

[0087] In the iteration process, the particles generate new particles using the following formula:

[0088]

[0089] where r1, r2 are random numbers generated in each update process, the values of r1, r2 are in the interval [0, 1); r is a random number generated in each update process, the value of r is in the interval [0, 1). F1 represents the speed update of the particle, F2 represents the perception and learning of the particle to the individual experience, and F3 represents the perception and learning of the particle to the population experience. represents the i-th particle at the t+1-th iteration; represents the i-th particle at the t-th iteration.

[0090] The mutation strategy of the particle is as follows:

[0091]

[0092] where w is the probability threshold of mutation. When the generated random number r is less than the probability threshold, the particle is mutated to obtain a new particle. First, two random integers i and j in the interval [1, n] are randomly generated, and n represents the number of nodes in the time sequence diagram of biochemical analysis. The values of y i and y j are exchanged to obtain the new particle after mutation. For example: when the generated random number r is less than w, if the two randomly generated integers are 2 and 4, the priority vector Y of the particle X i =(1, 2, 3, 4, 5) exchanges the values of y2 and y4 to generate a new priority vector Y'=(1, 4, 3, 2, 5). The new priority vector Y' and the binding vector B of the original particle X i compose the new particle X' i =(Y', B) after mutation.

[0093] The perception and learning of the particle to the individual experience are as follows:

[0094]

[0095] where c1 is the probability threshold of the perception and learning of individual experience. When the generated random number r1 is less than the probability threshold, the particle retains its priority vector Y and learns the binding vector B of the individual historical optimal solution. As shown in Figure 5 i The binding vector B of particle X i represents the binding relationship of three operations and three devices, and the individual optimal solution is taken as the learning target. If a random integer is generated as 2, the binding vector B of particle X 2，1 learns the binding relationship of the 2nd operation and device from the learning target, i.e. learns b 2，2 , b 2，3 . The learning situation is shown in Figure 6 After learning, it is necessary to judge whether the current binding relationship can satisfy the constraint of volume management. If not, the binding relationship of the 3rd operation is continuously learned, and the situation after learning is shown in Figure 7 The binding vector B' satisfying the constraint of volume management is continuously learned. The binding vector B' and the priority vector Y of particle X i form the new particle X' i =(Y, B') after learning.

[0096] The perception and learning of the population experience by the particle are as follows:

[0097]

[0098] where c2 is the probability threshold of the perception and learning of population experience. When the generated random number r2 is less than the probability threshold, the particle retains its priority vector Y and learns the binding vector B of the population historical optimal solution. The learning method is consistent with the learning of individual experience.​

Claims

1. A high-level synthesis method for continuous microfluidic biochips considering volume management, characterized in that, An automated scheduling scheme generation method is proposed, which generates a corresponding scheduling scheme of high-level synthesis satisfying volume management according to particle coding; according to the characteristics of the problem, a corresponding particle update strategy is designed, which enables the particle to learn the experience of individuals and populations in the iteration process, accelerating the convergence of the algorithm; wherein: (1) Particle encoding scheme: the position vector of a particle in the particle swarm is encoded as ; where the priority vector , , represents the priority of the operation with label ; the operand is , and the number of devices is , the binding vector ; is a 0-1 variable, indicating whether the operation is bound to the device , if bound, then =1; otherwise =0; (2) Automated scheduling scheme generation method: operations need to satisfy that the parent operation is executed and the bound device is in an idle state; real-time detection of executable operations is performed and immediately started; if there are operations bound to the same device that satisfy the execution condition at the same time, the operation with higher priority is executed first; To perform an operation , first find the operation bound device ; when the device is idle, the operation can use the device to remove the reagent, and after waiting for the subsequent reagent loading, it can start execution; the operation end time is the time when the device is available for the next operation; after updating, hand over the device ; after analyzing all operations, the operation scheduling scheme corresponding to the particles is obtained; the biochemical reaction completion time , the total number of flow paths , the fitness value of the particle is calculated , , wherein , are two weighted factors; (3) Particle update strategy: DPSO is used to solve the high-level synthesis operation scheduling scheme; the new particle is generated according to formula (2): Set the minimum volume limit of each device that can be input at a time, and the volume of reagent loaded into the device each time cannot be less than the minimum volume limit of the device that can be input at a time; (2) , , r is a random number in interval [0, 1]; represents the velocity update of a particle, represents the perception and learning of individual experience by a particle, represents the perception and learning of population experience by a particle, , respectively represent the i-th particle at the t+1, t-th iteration. Mutation strategy of particles: (3) is a probability threshold; when a generated random number is less than the probability threshold, the particle is subjected to mutation to obtain a new particle; first, two random integers are randomly generated within the interval [1, and ; the values of and in the particle are exchanged to obtain a new particle after mutation; Individual experience learning: when random number Probability threshold , particle retains its priority vector , learn the binding vector of the individual history optimal solution , learn the binding relationship between the corresponding operation and the device once by a random integer, until the binding vector satisfies the volume management constraint, new particle = ( ); Particles learn from the experience of the population: random numbers Probability threshold At this time, the particle retains its priority vector and the binding vector of the best solution in the population history The learning method is consistent with the experience of the individual described above.

2. The volume management aware continuous microfluidic biochip high-level synthesis method of claim 1, wherein, The high-level synthesis stage determines the device binding scheme of the continuous microfluidic biochip according to the operation information of biochemical reactions and a given device library, determines the time of each transport task, and determines the operation execution start time according to the reagent loading completion time; the operation end time is obtained according to the known operation start time and operation time required; the biochemical reactions in the continuous microfluidic biochip use a timing diagram to describe the relationship between each operation ; the timing diagram is a directed acyclic graph, wherein a node represents an operation in the biochemical reaction; the timing diagram has an edge between two biochemical reaction operations and ; the edge represents the dependency relationship between the operations and ; the operation is the parent operation of the operation ; the operation must be executed after the operation is completed; after the reaction ends, the reaction complete reagent needs to be transported to the device bound to the operation as the input of the operation .

3. The volume management aware continuous microfluidic biochip high-level synthesis method of claim 1, wherein, Three types of transportation tasks need to be considered in the design process of continuous microfluidic biochips CFMBs; the three types of transportation tasks are as follows: If there is one mixing operation waiting to be performed, the operation will mix two reagents, 2 microliters each; if there is a mixer available at this time, the mixer has a single input minimum volume limit of 4 microliters; the operation will continue to wait for a mixer with a single input minimum volume limit less than or equal to 2 microliters.

4. The volume management aware continuous microfluidic biochip high level synthesis method of claim 1, wherein, The first type of transportation task: the transportation task of reagents between devices; The second type of transportation task: the transportation task of removing excess liquid; The third type of transportation task: the transportation task of removing waste liquid; Before loading the reagent into the mixer, it needs to be determined whether there is waste liquid left in the mixer by the previous device, if there is, the waste liquid is transported away by the third type of transportation task; transporting the reagent to the mixer is the first type of transportation task; when loading the reagent into the mixer, no less than the amount that the mixer channel can hold will be loaded, at this time there will be some excess liquid at both ends of the mixer, if there is, the excess liquid will be removed by the second type of transportation task; In the design of continuous microfluidic biochips CFMBs, a separate flow path is planned for each transportation task. DPSO is used to solve discrete problems, and the input of the algorithm is the timing diagram of the biochemical reaction performed this time, the provided device library and the operation information table, the timing diagram provides the dependency relationship between each operation for this reaction; 5. The volume consideration continuous microfluidic biochip high-level synthesis method according to claim 1, wherein, The device library provides the basis for subsequent analysis of the operation scheduling scheme, and the operation information table provides the type of device required by the operation and the execution time; In the iteration update phase, after updating each particle, the operation scheduling scheme and the fitness value corresponding to each particle need to be analyzed again; then the individual historical optimal solution and the population historical optimal solution of each particle are updated; finally, the best device binding and operation scheduling scheme found are output; ​ Binding scheme for operation Binding device , operation Binding device ; operation There are two parent operations and one child operation; the minimum input volume of the device is 2 microliters; two parent operations load 2 microliters of reagent into the device ; operation After the reaction is complete, 4 microliters of reagent is available; operation There is one child operation, so operation Transports to operation with an actual volume of 4 microliters; if the minimum input volume of the device is 2 microliters, then 2 microliters of excess liquid must be removed.

6. The volume management aware continuous microfluidic biochip high-level synthesis method of claim 1, wherein, Design a timeline diagram for the mixing phase in PCR, including nodes 1 to 7, representing the operations. It also includes nodes 8 to 15, which represent liquid samples input from the outside. It also includes node 16, which represents the output of reagents to the outside. ; the position vectors of particles in the particle swarm Encoded as ;wherein the priority vector , The timing diagram shows a total of node, Indicates the label is The priority of the node; the number of operands is The number of devices is When binding vector ; It is a 0-1 variable, representing an operation. Is it bound to a device? If operation Distributed in the equipment If executed in the middle, then =1; if operation Not in device If executed in the middle, then =0; This encoding method enables rapid particle updates and allows for the analysis of operation scheduling schemes based on priority vectors and binding vectors.

Images