Gene sequencing temperature control method and system and gene sequencing platform

Abstract

The invention provides a gene sequencing temperature control method, a gene sequencing temperature control system and a gene sequencing platform. The method comprises the following steps: acquiring a working surface temperature of a semiconductor cooler for performing temperature regulation on a gene sequencing chip, a cooling liquid temperature for performing temperature regulation on the semiconductor cooler and an environment temperature which are acquired in real time; the working face temperature, the cooling liquid temperature and the environment temperature corresponding to the current time step serve as control input parameters, the control input parameters are input into the trained temperature control model, and control current corresponding to the next time step is output through the temperature control model; wherein the temperature control model is obtained by training a strategy network by adopting a training data set based on a reinforcement learning algorithm, and the reinforcement learning algorithm adopts a reinforcement learning algorithm based on state value function iteration; and controlling the working state of the semiconductor cooler based on the control current.

Description

Technical Field

[0001] The present application relates to the field of gene sequencing technology, and in particular to a gene sequencing temperature control method, a gene sequencing temperature control system and a gene sequencing platform. Background Art

[0002] In the next generation gene sequencer (Next Generation Sequencing, NGS), the gene samples loaded in the sequencing chip need to undergo biochemical reactions before they can be sequenced. And the biochemical reaction requires a suitable temperature. Therefore, a temperature control platform is set at the bottom of the sequencing chip. Usually, the temperature control platform of the second generation gene sequencer is achieved by TEC (also known as semiconductor refrigeration or Peltier cooler). At present, the control of TEC is mainly through PID (proportional-integral-differential) control of TEC input current to perform temperature control. PID is a feedback control algorithm, that is, the control output is adjusted according to the feedback signal to achieve the desired output effect.

[0003] However, during the gene sequencing process, the temperature control accuracy of the gene sequencing chip is required to be high. When the PID control is used to adjust the temperature of the gene sequencing chip, an overshoot time point is likely to occur. Figure 1 , which is a schematic diagram of the temperature control curve of the gene sequencing chip under PID control. During the temperature adjustment process of the gene sequencing chip, before the temperature of the gene sequencing chip stabilizes, the temperature will overshoot and callback many times, and it will be stabilized to the target temperature after repeated many times. The occurrence of this overshoot time point may affect the normal operation of TEC, and it is also easy to affect the normal sequencing reaction of the gene sequencing chip, thereby affecting the sequencing accuracy, and even causing sequencing failure, and greatly delaying the time for the gene sequencing chip to reach a stable target temperature. Summary of the invention

[0004] In order to solve the existing technical problems, the present application provides a gene sequencing temperature control method, a gene sequencing temperature control system and a gene sequencing platform which have more precise temperature control and effectively shorten the temperature rise control time.

[0005] In a first aspect, a method for controlling temperature of gene sequencing is provided, comprising:

[0006] Acquire the real-time collected working surface temperature of the semiconductor refrigerator for temperature regulation of the gene sequencing chip, the coolant temperature for temperature regulation of the semiconductor refrigerator, and the ambient temperature;

[0007] The working surface temperature, the coolant temperature and the ambient temperature corresponding to the current time step are used as control input parameters, the control input parameters are input into the trained temperature control model, and the control current corresponding to the next time step is output through the temperature control model; wherein the temperature control model is obtained by training the policy network based on the reinforcement learning algorithm using the training data set, and the reinforcement learning algorithm adopts the reinforcement learning algorithm based on the iteration of the state value function;

[0008] The working state of the semiconductor refrigerator is controlled based on the control current.

[0009] In a second aspect, a gene sequencing temperature control system is provided, comprising:

[0010] A semiconductor refrigerator, including a working surface close to the gene sequencing chip, for adjusting the temperature of the gene sequencing chip;

[0011] A liquid cooling module is arranged on one side of the non-working surface of the semiconductor refrigerator;

[0012] A temperature controller is used to execute the gene sequencing temperature control method described in any embodiment of the present application, and the temperature controller controls the working state of the semiconductor refrigerator by outputting the control current.

[0013] In a third aspect, a gene sequencing platform is provided, comprising:

[0014] A workbench, used to carry gene sequencing chips;

[0015] The gene sequencing temperature control system described in any embodiment of the present application.

[0016] The gene sequencing temperature control method provided in the above embodiment, in the process of temperature adjustment for the gene sequencing chip, collects in real time the working surface temperature of the semiconductor refrigerator for temperature adjustment of the gene sequencing chip, the coolant temperature and the ambient temperature for temperature adjustment of the semiconductor refrigerator as control input parameters, and uses the control input parameters as the input of the temperature control model obtained after training the policy network based on the reinforcement learning algorithm. The temperature control model outputs the control current to control the working state of the semiconductor refrigerator. The reinforcement learning algorithm adopts a reinforcement learning algorithm based on state value function iteration, and constructs a learning algorithm by calculating the state value function to determine the optimal strategy from the data. In this way, the reinforcement learning algorithm can be used to obtain the optimal control scheme for the control current of the semiconductor refrigerator with the goal of minimizing the time taken for the gene sequencing chip to reach and stabilize at the target temperature. The temperature control is more accurate and the temperature rise control time is effectively shortened.

[0017] In the above embodiments, the gene sequencing temperature control system and the gene sequencing platform belong to the same concept as the corresponding gene sequencing temperature control method embodiments, and thus have at least the same technical effects as the corresponding gene sequencing temperature control method embodiments, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a diagram of optional application scenarios of a temperature control method for gene sequencing in one embodiment.

[0019] Figure 2 Flow chart of a temperature control method for gene sequencing in one embodiment.

[0020] Figure 3 Schematic diagram of a policy network in one embodiment.

[0021] Figure 4 The present invention is a flowchart of a temperature control method for gene sequencing in an optional specific example.

[0022] Figure 5 Schematic diagram of the structure of a temperature control system for gene sequencing in one embodiment. DETAILED DESCRIPTION

[0023] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0024] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0025] In the following description, the expression "some embodiments" is involved, which describes a subset of all possible embodiments. It should be noted that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.

[0026] It should also be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may also be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "inside", "outside", "left", "right" and similar expressions used herein are only for the purpose of explanation in conjunction with the embodiments of the accompanying drawings and do not represent the only implementation method.

[0027] See also Figure 1, is an optional application scenario diagram of a gene sequencing temperature control method provided by an embodiment, wherein the temperature control system applying the gene sequencing temperature control method includes a semiconductor refrigerator (TEC) 11, a liquid cooling module 12, and a controller 13 for controlling the operation of the TEC. TEC11 uses the heat or cold energy generated by the working surface to accurately adjust the temperature of the gene sequencing chip during the sequencing reaction. The liquid cooling module 12 is arranged on the non-working surface side of the TEC11 to promptly discharge the waste heat generated by the operation of the TEC11 to avoid partial PN junction failure of the TEC11 during use, which affects the normal operation of the TEC11, and the temperature control curve will also be offset, making the temperature control inaccurate. The controller 13 loads and executes a computer program for the gene sequencing temperature control method provided in the embodiment of the present application, and uses a temperature control model obtained by training a policy network based on a reinforcement learning algorithm using a training data set. The working surface temperature of TEC11 for temperature regulation of the gene sequencing chip, the coolant temperature for temperature regulation of TEC11, and the ambient temperature collected in real time during the sequencing reaction are used as control input parameters, and outputs a control current for controlling the working state of TEC11. By optimizing the control scheme for controlling the working state of TEC11, it is ensured that the gene sequencing chip can reach and stabilize at the target temperature in the shortest time through the temperature regulation of TEC11, and avoids repeated overshoot or callback of the temperature before the gene sequencing chip reaches temperature stability, thereby ensuring that the gene sequencing chip completes the normal sequencing reaction at the precise target temperature and guarantees the sequencing accuracy.

[0028] It should be noted that the controller 13 may refer to a controller integrated in the gene sequencing platform for executing sequencing reaction control; it may also refer to other controllers integrated in the gene sequencing platform but independent of the controller for executing sequencing reaction control; it may also refer to a controller independent of the gene sequencing platform that can control TEC based on a communication connection, and this application does not impose any restrictions on this.

[0029] See also Figure 2 , a gene sequencing temperature control method provided in an embodiment of the present application, comprising the following steps:

[0030] S101, obtaining the real-time collected working surface temperature of a semiconductor refrigerator for temperature regulation of a gene sequencing chip, the cooling liquid temperature for temperature regulation of the semiconductor refrigerator, and the ambient temperature.

[0031] Thermal Electronic Cooler (TEC) is a thermoelectric cooler based on the Peltier effect. The smallest unit of TEC is composed of a pair (group) of N-type and P-type semiconductors plus connecting electrodes (sintering points), and the connecting electrodes form the cold end and the hot end. Under the action of an external electric field, the current can carry the heat generated in the semiconductor from one end of the TEC to the other end, generating a hot side and a cold side on the TEC. When the current direction is reversed, the hot and cold ends of the TEC will switch.

[0032] In the gene sequencing reaction, the controller outputs a control current to the TEC to control its working state, and the working surface of the TEC generates heat energy or cold energy to provide to the gene sequencing chip, so as to adjust the temperature of the gene sequencing chip in real time during the sequencing reaction. Real-time acquisition of the working surface temperature of the semiconductor refrigerator for temperature regulation of the gene sequencing chip, the coolant temperature and the ambient temperature for temperature regulation of the semiconductor refrigerator means that in the gene sequencing reaction, the working surface temperature of the TEC, the coolant temperature in the liquid cooling module and the current sequencing ambient temperature are collected in real time at a certain frequency, and the working surface temperature, coolant temperature and ambient temperature of the TEC collected at each moment are used as a set of input control parameters of the temperature control model.

[0033] S103, taking the working surface temperature, the coolant temperature and the ambient temperature corresponding to the current time step as control input parameters, inputting the control input parameters into the trained temperature control model, and outputting the control current corresponding to the next time step through the temperature control model; wherein, the temperature control model is obtained by training the policy network based on the reinforcement learning algorithm using the training data set, and the reinforcement learning algorithm uses a reinforcement learning algorithm based on state value function iteration.

[0034] Reinforcement Learning (RL), also known as reinforcement learning, evaluation learning or enhanced learning, is a field of machine learning, which is used to describe and solve the problem of an agent learning a strategy to maximize rewards or achieve a specific goal during the interaction with the environment. The policy network is trained based on a training data set based on a reinforcement learning algorithm. The policy network can obtain an action output based on the environmental input, and the policy network weight parameters are updated through training. In this way, the temperature control model obtained after training can obtain an action output based on the real-time environmental input during the application process. In the embodiment of the present application, the temperature control model uses the working surface temperature, coolant temperature and ambient temperature of the TEC collected in real time in the gene sequencing reaction as control input parameters to adjust the control current for controlling the working state of the TEC. For example, the environmental input refers to the working surface temperature of the TEC collected in real time, the coolant temperature and ambient temperature for temperature regulation of the semiconductor refrigerator as control input parameters, and the action output refers to the control current for regulating the working state of the TEC. One input and one action output correspond to one state t.

[0035] The current time step may refer to a time point at which the working surface temperature, coolant temperature and ambient temperature are collected at a certain frequency, and accordingly, the next time step refers to the next time point at which the working surface temperature, coolant temperature and ambient temperature are collected at the certain frequency; or, the current time step may also refer to the working surface temperature, coolant temperature and ambient temperature collected at n consecutive time points as a time step, and accordingly, the next time step refers to the time period corresponding to the next n consecutive time points. The temperature control model uses the working surface temperature, coolant temperature and ambient temperature at the current time step as control input parameters to adjust the control current corresponding to the next time step. It is based on the control input parameters of the current time step, with the purpose of heating the gene sequencing chip to reach and stabilize at the target temperature in the shortest time, and the predicted output value of the control current of the TEC is used for real-time regulation in the next time step.

[0036] The reinforcement learning algorithm adopts a reinforcement learning algorithm based on the iteration of the state value function. Under any state t, the learning algorithm is constructed by calculating the state value function to obtain the optimal strategy from the data. Each strategy corresponds to a state value function, and the action with the largest value function is the current optimal strategy. In this embodiment, under any state t, a set of control input parameters (the working surface temperature, the coolant temperature and the ambient temperature) corresponds to an action output (the control current for adjusting the working state of the TEC), and the temperature value corresponding to the preset time after the current output control current is calculated to meet the preset conditions, and the minimum time required for the gene sequencing chip to reach the target temperature is calculated to meet the preset conditions, and the state value function corresponding to the current strategy is calculated, and then the state value function is used to improve the current strategy until the optimal strategy is determined. During the training process of the strategy network, the loss value calculated in each iterative training is used for back propagation to update the weight parameters of the strategy network. In this way, after the strategy network converges through training, it can achieve the purpose of heating the gene sequencing chip to reach and stabilize at the target temperature in the shortest time, and the control current of the TEC is controlled in real time.

[0037] S105, controlling the working state of the semiconductor refrigerator based on the control current.

[0038] In the gene sequencing reaction, the temperature control model uses the real-time collected TEC working surface temperature, coolant temperature and ambient temperature as control input parameters, and the control input parameters of the current time step are used to predict and determine the control current for controlling the working state of TEC in the next time step. In this way, the optimal control scheme for the control current input to TEC can be obtained.

[0039] The gene sequencing temperature control method provided in the above embodiment, in the process of temperature adjustment for the gene sequencing chip, collects in real time the working surface temperature of the semiconductor refrigerator for temperature adjustment of the gene sequencing chip, the coolant temperature and the ambient temperature for temperature adjustment of the semiconductor refrigerator as control input parameters, and uses the control input parameters as the input of the temperature control model obtained after training the policy network based on the reinforcement learning algorithm. The control current is output by the temperature control model to control the working state of the semiconductor refrigerator. The reinforcement learning algorithm adopts a reinforcement learning algorithm based on state value function iteration, and constructs a learning algorithm by calculating the state value function to determine the optimal strategy from the data. In this way, the reinforcement learning algorithm can be used to obtain the optimal control scheme for the control current of the semiconductor refrigerator with the goal of minimizing the time taken for the gene sequencing chip to reach and stabilize at the target temperature. The temperature control is more accurate and the temperature rise control time is effectively shortened.

[0040] In some embodiments, the loss function of the evaluation strategy of the strategy network is as shown in the following formula 1:

[0041] δ t =R t+1 +γV(S t+λ )-V(S t ; (Formula 1)

[0042] Among them, δ t is the loss value, R t+1 refers to the reward of state t+1, γ refers to the discount factor, which can be a preset value, V(S t+λ ) refers to the value function of state t+λ, V(S t ) refers to the value function of state t.

[0043] The reinforcement learning algorithm uses a reinforcement learning algorithm based on state value function iteration, parameterizes the evaluation strategy, searches for the optimal strategy in the parameter space based on the state value function calculation, and learns the optimal parameters to achieve the goal. The evaluation strategy is a strategy evaluation algorithm that calculates each state value function under the given strategy based on the current input control parameters, calculates the loss value corresponding to each state value function, and searches for the optimal parameters to achieve the goal in the parameter space.

[0044] In this embodiment, based on the loss function of the evaluation strategy, the output expected value is calculated based on the current control current and the temperature value after the preset time λ, and then the minimum time required to reach the target temperature to meet the preset conditions is calculated as the goal. The state value function corresponding to the current strategy is calculated, and then the state value function is used to improve the current strategy until the optimal strategy is determined. The loss value is formed based on the iteration of the state value function, and the convergence of the calculated loss value is used as the condition for the completion of training, so that the trained temperature control model can achieve the purpose of heating up to the target temperature in the shortest time, and the control current of TEC can be adjusted in real time.

[0045] In some embodiments, the given strategy in the strategy evaluation function may be a known current strategy, or may be a greedy strategy (for selecting the action with the maximum reward value from the action list of the state) to optimize the current strategy. The evaluation strategy of the strategy network also includes updating the strategy based on the reward value;

[0046] The calculation method of the reward value is shown in the following formula 2:

[0047]

[0048] Where T is the time to reach the target temperature, T a That is, the working surface temperature, T min is the minimum time to reach the target temperature, η is the reward conversion coefficient, Tpt T is the target temperature, Tpt T +3 is the upper limit of the set temperature.

[0049] The calculation of the reward value in Formula 2 means that if the surface temperature of the working surface does not exceed the upper temperature limit and is greater than the minimum statistical time, the calculated reward value is given; if the surface temperature of the working surface does not exceed the upper temperature limit and is less than the minimum statistical time, the maximum reward value of 100 is given, and the minimum statistical time is updated; when the surface temperature of the working surface exceeds the upper temperature limit, a direct penalty of -100 is imposed.

[0050] Updates to the strategy include:

[0051] The calculation of the fitness trajectory is shown in Formula 3:

[0052]

[0053] Where γ is the discount factor, σ is the weight, 0≤σ≤1, and ∑σ=1; Based on the loss value, determine whether the appropriate state S is matched. t , if the match is found, the fitness trajectory of the state is marked as 1, that is, the state is strongly correlated with the current state.

[0054] For the matching state S t , the update value function is shown in Formula 4:

[0055] V(S t )=V(S t )+α*δ t *E t (s); (Formula 4)

[0056] Among them, V(S t ) is S t State value function, E t (s) indicates state S t The corresponding fitness trajectory, E t-1 (s) represents the fitness trajectory corresponding to state t-1, δ t refers to the loss value, and α refers to the weight coefficient, which is a preset value.

[0057] According to the value function calculated by matching the state, the final strategy π(s) = argmaxV(s) is output. In this way, by calculating each state value function under each given strategy, the loss value corresponding to each state value function is calculated, and the strategy of the optimal parameters to achieve the goal is searched in the parameter space. Finally, using the updated optimal parameter strategy, the strategy network is saved as the temperature control model.

[0058] In some embodiments, the gene sequencing temperature control method further comprises:

[0059] Acquire a training data set; wherein the training data set includes training samples {I, Ta, Tb, Tc} formed by collecting control current I, working surface temperature Ta, coolant temperature Tb and ambient temperature Tc at a preset sampling time interval, with the optimization goal of minimizing the time (T1-T0) taken for the working surface temperature to rise to a preset temperature during the control of the semiconductor refrigerator;

[0060] The training data set is used to train the policy network based on a reinforcement learning algorithm to obtain the temperature control model.

[0061] Taking the target temperature of 60℃ as an example, the training sample can be the working surface temperature Ta rising from the ambient temperature Tc to the target temperature of 60℃. Starting from time T0, the controller outputs the control current I of TEC, and records the current value I, the working surface temperature Ta, the water cooling liquid temperature Tb and the ambient temperature Tc every 0.1s. During the heating process of the gene sequencing chip, the controller will also adjust the output control current I according to (Ta, Tb, Tc), where I∈[0, Imax], Imax is the maximum allowable current of TEC, such as 10A. At time T1, Ta=60℃ is the stop trigger condition. T1-T0 is the time used, and (T1-T0)min is the optimization target. {(T1-T0)(I,Ta,Tb,Tc)} is a complete training sample data set.

[0062] In the above embodiment, in the process of temperature adjustment for the gene sequencing chip, the TEC working surface temperature, the coolant temperature and the ambient temperature are set as control input parameters, and the time taken for the gene sequencing chip to reach the target temperature from the initial temperature is minimized. A training data set for online training is obtained, and a reinforcement learning algorithm is used to train using the temporal difference (TD(λ)) algorithm to obtain the optimal control method for the TEC input current and the working surface temperature of the gene sequencing chip.

[0063] In some embodiments, see Figure 3 , the policy network is a convolutional neural network, the hidden layer of the convolutional neural network includes an activation layer, and the activation function used is as follows:

[0064]

[0065] Here, e is the base of natural logarithms.

[0066] The strategy network obtains the action output (control current) according to the environmental input (control input parameters consisting of TEC working surface temperature, coolant temperature and ambient temperature). A convolutional neural network can be used, which can be divided into input layer, hidden layer and output layer. The initial number of nodes in the input layer can be adjusted according to the variables of the input environment. This scheme has three input parameters Ta, Tb, and Tc.

[0067] The initial number of hidden layers of the neural network is 3+3+1 layers, and the number of nodes ranges from 64 to 512, including BN (normalization) layers, convolutional layers, pooling layers, and activation layers.

[0068] The neuron formula can be expressed as: y=ax+b; x is the input, y is the output, a and b are weights, which are the parameters to be trained and updated.

[0069] Among them, the pooling layer of the convolutional neural network adopts the maximum pooling function (MaxPooling).

[0070] The activation function used in the activation layer is the tanh function.

[0071] After the final output layer is normalized, the value with the highest probability is selected for output.

[0072] The number of nodes in the output layer is adjusted according to actual needs. In this scheme, the output of the output layer is a continuous current value I.

[0073] In the above embodiment, the policy network adopts a convolutional neural network, which is beneficial to reducing the number of network parameters and the amount of calculation. The activation function adopts the tanh function, which can generate nonlinear output within the input range, and helps the convolutional neural network model to better fit nonlinear data and provide powerful nonlinear expression capabilities.

[0074] In order to have a more comprehensive understanding of the gene sequencing temperature control method provided in the embodiments of the present application, please refer to Figure 4 , taking an optional specific example, the temperature control method for gene sequencing is described:

[0075] S11, load and initialize the policy network. The policy network can use a convolutional neural network such as Figure 3 shown.

[0076] S12, obtaining a training data set. For example, {(T1-T0)(I, Ta, Tb, Tc)} is a complete training sample, forming a training data set.

[0077] S13, using the training data set to perform online training on the policy network using a reinforcement learning algorithm to obtain a trained temperature control model.

[0078] As shown in Formula 1, the strategy evaluation function of the strategy network can be based on the current input control parameters for a known strategy, and calculate each state value function under the given strategy according to the given strategy, so as to calculate the loss value corresponding to each state value function, and search for the optimal parameters to achieve the goal in the parameter space. Furthermore, as shown in Formulas 2 to 5, the strategy evaluation function of the strategy network can be further combined with the update optimization of the strategy, screen the matching state, and calculate the loss value corresponding to the matching state value function to search for the strategy of the optimal parameters to achieve the goal in the parameter space. The goal is that the gene sequencing chip reaches and stabilizes at the target temperature in the shortest time, the temperature fluctuation is ≤2°C within 3s, there is no temperature overshoot, the heating time is shortened, and then the sequencing time is shortened, the sequencing accuracy is ensured, and the economic benefits are improved.

[0079] S14, in the gene sequencing reaction, the working surface temperature of the TEC for temperature regulation of the gene sequencing chip, the coolant temperature for temperature regulation of the semiconductor refrigerator, and the ambient temperature are collected in real time as the control input parameters of the temperature control model, and the control current of the TEC is output by the temperature control model to control the working state of the TEC.

[0080] In the above embodiment, the temperature control model for controlling the working state of TEC is trained by using a reinforcement learning algorithm, so that the optimal control scheme for the control current of the semiconductor refrigerator can be obtained with the goal of minimizing the time taken for the gene sequencing chip to reach and stabilize at the target temperature. The temperature control is more precise and the temperature rise control time is effectively shortened.

[0081] See also Figure 5 On the other hand, the present application also provides a gene sequencing temperature control system, including: a semiconductor refrigerator (TEC11), including a working surface close to the gene sequencing chip 20, for adjusting the temperature of the gene sequencing chip 20. A liquid cooling module 12, arranged on one side of the non-working surface of TEC11. A temperature controller, used to execute the gene sequencing temperature control method described in any embodiment of the present application, and the temperature controller controls the working state of TEC11 by outputting the control current.

[0082] Optionally, the gene sequencing temperature control system further includes: a uniform temperature heat conduction plate 14 arranged between the gene sequencing chip 20 and the working surface of TEC11; wherein the working surface temperature is the surface temperature of the uniform temperature heat conduction plate 14 and the gene sequencing chip 20 in contact with each other.

[0083] Optionally, the liquid cooling module 12 includes a liquid cooling plate 121 connected to the non-working surface of the TEC 11 by heat conduction, the liquid cooling plate 121 is provided with a liquid inlet and a liquid outlet, and a flow channel for containing coolant for the flow of the coolant is provided inside the liquid cooling plate 121, and the liquid cooling plate 121 is connected to an external liquid storage tank 122 through the liquid inlet and the liquid outlet.

[0084] In another aspect, the present application further provides a gene sequencing platform, comprising:

[0085] A workbench, used to carry gene sequencing chips;

[0086] A gene sequencing temperature control system as described in any embodiment of the present application.

[0087] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0088] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a server, a gene sequencer, etc.) to execute the methods of various embodiments of the present invention.

[0089] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A gene sequencing temperature control method, characterized in that: include: Acquire the real-time collected working surface temperature of the semiconductor refrigerator for temperature regulation of the gene sequencing chip, the coolant temperature for temperature regulation of the semiconductor refrigerator, and the ambient temperature; The working surface temperature, the coolant temperature and the ambient temperature corresponding to the current time step are used as control input parameters, the control input parameters are input into the trained temperature control model, and the control current corresponding to the next time step is output through the temperature control model; wherein the temperature control model is obtained by training the policy network based on the reinforcement learning algorithm using the training data set, and the reinforcement learning algorithm adopts the reinforcement learning algorithm based on the iteration of the state value function; The working state of the semiconductor refrigerator is controlled based on the control current.

2. The gene sequencing temperature control method according to claim 1, characterized in that: The loss function of the evaluation strategy of the policy network is as follows: δ t =R t+1 +γV(S t+λ )-V(S t ); Among them, δ t is the loss value, R t+1 is the reward of state t+1, γ is the discount factor, V(S t+λ ) refers to the value function of state t+λ, V(S t ) refers to the value function of state t.

3. The gene sequencing temperature control method according to claim 2, characterized in that: The evaluation strategy of the strategy network also includes updating the strategy based on the reward value; The reward value is calculated as follows: Where T is the time to reach the target temperature, T a That is, the working surface temperature, T min is the minimum time to reach the target temperature, η is the reward conversion coefficient, Tpt T is the target temperature, Tpt T +3 is the upper limit of the set temperature; The update of the said policy is as follows: Among them, S t is state t, S t+1 is state t+1, γ is the discount factor, V(S t+1 ) is S t+1 State value function, V(S t ) is S t State value function, δ t It refers to the loss value, σ is the weight, 0≤σ≤1, and ∑σ=1; α is the weight, E t (s) represents the fitness trajectory corresponding to state t, E t-1 (s) represents the fitness trajectory corresponding to state t-1.

4. The gene sequencing temperature control method according to claim 2, characterized in that: Also includes: Acquire a training data set; wherein the training data set includes training samples {I, Ta, Tb, Tc} formed by collecting control current I, working surface temperature Ta, coolant temperature Tb and ambient temperature Tc at a preset sampling time interval, with the optimization goal of minimizing the time (T1-T0) taken for the working surface temperature to rise to a preset temperature during the control of the semiconductor refrigerator; The training data set is used to train the policy network based on a reinforcement learning algorithm to obtain the temperature control model.

5. The gene sequencing temperature control method according to claim 4, characterized in that: The policy network is a convolutional neural network, the hidden layer of the convolutional neural network includes an activation layer, and the activation function used is as follows: Here, e is the base of natural logarithms.

6. The gene sequencing temperature control method according to claim 4, characterized in that: The pooling layer of the convolutional neural network adopts the maximum pooling function.

7. A gene sequencing temperature control system, characterized in that: include: A semiconductor refrigerator, including a working surface close to the gene sequencing chip, for adjusting the temperature of the gene sequencing chip; A liquid cooling module is arranged on one side of the non-working surface of the semiconductor refrigerator; A temperature controller, used to execute the gene sequencing temperature control method according to any one of claims 1 to 6, wherein the temperature controller controls the working state of the semiconductor refrigerator by outputting the control current.

8. The gene sequencing temperature control system according to claim 7, characterized in that: Also includes: A uniform temperature heat conduction plate is arranged between the gene sequencing chip and the working surface of the semiconductor refrigerator; wherein the working surface temperature is the surface temperature of the uniform temperature heat conduction plate and the gene sequencing chip in contact with each other.

9. The gene sequencing temperature control system according to claim 8, characterized in that: The liquid cooling module includes a liquid cooling plate connected to the non-working surface of the semiconductor refrigerator by heat conduction. The liquid cooling plate is provided with a liquid inlet and a liquid outlet and has a flow channel inside for accommodating coolant for the flow of the coolant. The liquid cooling plate is connected to an external liquid storage tank through the liquid inlet and the liquid outlet.

10. A gene sequencing platform, characterized in that: include: A workbench, used to carry gene sequencing chips; A gene sequencing temperature control system as described in any one of claims 7 to 9.

Images