Exosome tangential flow filtration process parameter adaptive optimization method

Abstract

The present application relates to the field of biological medicine separation and purification, and particularly relates to a tangential flow filtration process parameter adaptive optimization method for exosome. The method comprises the following steps: after obtaining a steady-state benchmark transmembrane pressure, a sweep disturbance signal with continuously changing frequency is superimposed, a characteristic frequency triggering a permeation flux peak is extracted as an initial resonant frequency and a limited amplitude alternating oscillation is performed, a tentative frequency signal is injected periodically during washing and filtering, an optimal resonant frequency is dynamically updated based on a flux differential change gradient extremum search, and the method is adaptively stopped when a set condition of a decay slope of an impurity concentration characteristic signal is met. The present application utilizes water power resonance to destroy the polarization gel layer of physically trapped impurities, improves the mass transfer efficiency under the premise of ensuring the integrity of the exosome vesicle structure, reduces the consumption of expensive buffer and the process time, and realizes intelligent closed-loop control of the whole washing and filtering purification process.

Description

Technical Field

[0001] This invention relates to the field of biopharmaceutical separation and purification, and in particular to an adaptive optimization method for exosome tangential flow filtration process parameters. Background Technology

[0002] Exosomes, as nanoscale lipid vesicles secreted by cells, have great potential applications in cutting-edge fields such as precision drug delivery and cell-free therapy. Tangential flow filtration, with its advantages of being gentle and linearly scalable, has become a core technology for the separation and purification of exosomes. In the tangential flow filtration stage, a high-concentration exosome feed solution flows parallel to the filter membrane surface. Small free proteins pass through the filter membrane, while exosomes are retained and recycled. As filtration continues, significant concentration polarization occurs on the filter membrane surface due to the continuous accumulation of solute, resulting in a dense hydrodynamic gel layer boundary at the fluid-membrane interface.

[0003] Existing washing and filtration processes typically employ a constant flow rate in a laminar flow mode. This laminar flow state easily leads to the physical trapping and entrainment of free impurity proteins within the dense polarized gel layer, resulting in extremely low mass transfer efficiency of impurities permeating through the filter membrane. To achieve pharmaceutical purity, current technologies are forced to increase the volume of washing buffer replacement significantly, which not only prolongs the process time but also wastes expensive buffer solutions. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides an adaptive optimization method for exosome tangential flow filtration process parameters, which aims to improve the problems of impurities being encapsulated in the gel layer due to advection filtration, resulting in low mass transfer efficiency and wasted buffer solution.

[0005] This invention provides the following technical solution: an adaptive optimization method for exosome tangential flow filtration process parameters, comprising:

[0006] S1. When the exosomes enter the tangential flow filtration stage, control the liquid inlet actuator to run at a preset tangential flow rate, and collect the reference transmembrane pressure and reference permeation flux of the system under the advection steady state.

[0007] S2. The fluid regulating component at the return end is controlled to superimpose a frequency sweeping disturbance signal with continuously varying frequency on the basis of the reference transmembrane pressure, and the real-time permeation flux corresponding to each disturbance frequency is collected simultaneously.

[0008] S3. Based on the mapping relationship between each disturbance frequency and the synchronously acquired real-time infiltration flux, extract the characteristic frequency corresponding to the peak value of the real-time infiltration flux and lock it as the initial hydrodynamic resonant frequency.

[0009] S4. Control the fluid regulating component at the return end to oscillate alternately at the initial hydrodynamic resonant frequency, and limit the transmembrane pressure fluctuation amplitude caused by the alternating oscillation to within a preset amplitude threshold.

[0010] S5. During the continuous washing and filtration process, a test frequency signal is injected into the fluid regulating component at the return end according to the set replacement cycle, and the current optimal hydrodynamic resonant frequency is dynamically updated based on the differential change gradient of the real-time permeation flux.

[0011] S6. During operation at the current optimal hydrodynamic resonant frequency, the impurity concentration characteristic signal at the permeation end is monitored synchronously. When the attenuation slope of the impurity concentration characteristic signal reaches the preset termination condition, the tangential flow washing stage is adaptively stopped.

[0012] Preferably, in step S1, the step of obtaining the reference transmembrane pressure and reference permeation flux under advection steady state includes:

[0013] Real-time monitoring of pressure signals at the inlet and outlet ends; confirmation that the fluctuation variance of the pressure signal within a preset time window is lower than the steady-state threshold; and determination that the system has entered advection steady state.

[0014] Based on the system pressure data at each end after entering the advection steady state, the average transmembrane pressure during the steady state time is calculated and extracted as the reference transmembrane pressure.

[0015] Based on the flow rate detection data at the permeation end and the effective filtration area, the permeate volume per unit time is calculated, and the baseline permeation flux is generated and recorded.

[0016] Preferably, step S2, the step of superimposing a frequency-sweeping disturbance signal with continuously varying frequency, includes:

[0017] Construct a frequency sweep control vector that includes frequency sweep boundary parameters and step resolution to limit the frequency sweep disturbance signal to be within a preset tolerance frequency band;

[0018] The frequency sweep control vector is converted into a drive signal to control the return fluid regulator to generate a continuous opening oscillation at a corresponding frequency.

[0019] During the continuous opening oscillation, the output power of the liquid inlet actuator is dynamically compensated to maintain the preset tangential flow velocity from interference by the superposition of the frequency sweep disturbance signal.

[0020] Preferably, in step S3, the step of mapping the perturbation frequencies to the synchronously acquired real-time permeation flux includes:

[0021] The collected real-time permeation flux is filtered to remove background fluid noise introduced by the mechanical operation of the system.

[0022] Based on the time synchronization identifier, the filtered real-time penetration flux and the output perturbation frequencies are matched and aligned in the time domain to construct a two-dimensional sample set;

[0023] A smoothing fitting operation is performed on the two-dimensional sample set to generate a continuous frequency-flux response curve, thus constructing the mapping relationship.

[0024] Preferably, in step S3, the step of extracting the characteristic frequency corresponding to the peak value that triggers the real-time permeation flux includes:

[0025] Calculate the derivative change characteristics of the mapping relationship and search for candidate extreme points where the first derivative is zero and the second derivative is negative;

[0026] When the real-time permeation flux corresponding to the candidate extreme point is greater than the benchmark permeation flux, the candidate extreme point is verified as an effective resonant peak.

[0027] Extract the frequency coordinates corresponding to the effective resonance peak, use them as the characteristic frequency, and lock and record them as the initial hydrodynamic resonance frequency.

[0028] Preferably, in step S4, the step of limiting the transmembrane pressure fluctuation amplitude caused by the alternating oscillation to within a preset amplitude threshold includes:

[0029] Based on the reference transmembrane pressure and the preset elastic proportional coefficient, the allowable pressure boundary for fluctuation is calculated and generated as the preset amplitude threshold.

[0030] During the execution of the alternating oscillation, the fluctuation envelope signal of the current system transient transmembrane pressure is continuously extracted;

[0031] When the fluctuation envelope signal touches the allowable pressure boundary, the amplitude of the alternating oscillation is adaptively attenuated to keep it within the preset amplitude threshold.

[0032] Preferably, in step S5, the step of injecting a test frequency signal into the return fluid regulator according to a set replacement cycle includes:

[0033] The accumulated amount of liquid at the permeation end is continuously monitored. When the accumulated amount of liquid meets the set replacement cycle, a frequency test action is triggered.

[0034] Using the current optimal hydrodynamic resonance frequency as a reference point, positive and negative test perturbation signals with preset frequency step lengths are generated respectively.

[0035] The positive and negative test perturbation signals are alternately output to the return fluid regulating element, and the average permeation flux in both directions within the test time window is recorded independently.

[0036] Preferably, in step S5, the step of updating the current optimal hydrodynamic resonant frequency includes:

[0037] Extract the direction and magnitude of the difference in real-time permeation flux caused by injecting the test frequency signal in different frequency offset directions, and calculate the current differential change gradient by combining the offset step size of the test frequency signal;

[0038] By introducing a preset search gain parameter, the differential gradient is transformed into a frequency correction step size that approaches the target extreme point;

[0039] The frequency correction step size is used to superimpose and compensate the current optimal hydrodynamic resonant frequency, driving the frequency to slide towards the gradient rising side, thereby updating and obtaining a new current optimal hydrodynamic resonant frequency.

[0040] Preferably, in step S6, the step of adaptively stopping the tangential flow washing and filtration stage includes:

[0041] Using a concentration detection unit configured at the permeation end, a sequence signal characterizing the content of free impurities is continuously acquired as the impurity concentration feature signal;

[0042] A sliding analysis window is applied to the impurity concentration characteristic signal, and the linear rate of change of the signal intensity within the window is calculated and extracted as the attenuation slope.

[0043] When the absolute value of the attenuation slope remains below the zero tolerance parameter for a preset time period, a stop command is generated to adaptively stop the tangential flow washing and filtration stage.

[0044] The present invention has the following beneficial effects:

[0045] 1. In this invention, the optimal hydrodynamic resonant frequency of the pipeline system is dynamically tracked by frequency sweeping, thereby driving the fluid regulating component to generate micro-amplitude alternating oscillations. The resonant waves are used to destroy the polarized gel layer that encapsulates impurities, allowing free proteins to be rapidly released and permeated. This improves the washing and filtration mass transfer efficiency without increasing macroscopic shear force and saves on expensive buffer consumption.

[0046] 2. In this invention, the transient transmembrane pressure fluctuation envelope signal is continuously extracted during the application of alternating oscillation and compared with the allowable pressure boundary set based on vesicle elasticity. When the pressure fluctuation touches the boundary, the amplitude is adaptively attenuated, and the disturbance is strictly limited within the safe threshold. This effectively avoids physical damage to the exosomal lipid bilayer caused by alternating stress and improves the structural integrity of the product.

[0047] 3. In this invention, the impurity sequence signal is acquired in real time by the permeation end concentration detection unit, and the attenuation slope of the signal intensity is continuously calculated by the sliding analysis window. When the absolute value of the slope is continuously lower than the zero-approaching parameter, the shutdown command is automatically issued, avoiding the traditional flow consumption mode that relies on a fixed washing volume. This realizes intelligent and unmanned closed-loop control of the entire exosome purification process. Attached Figure Description

[0048] Figure 1 This is a flowchart of an adaptive optimization method for exosome tangential flow filtration process parameters proposed in this invention;

[0049] Figure 2 This is a flowchart illustrating the implementation of the adaptive control for tangential flow filtration of exosomes proposed in this invention. Detailed Implementation

[0050] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Example 1

[0052] In a first embodiment of the present invention, the present invention provides an adaptive optimization method for exosome tangential flow filtration process parameters, such as... Figure 1 As shown, it includes:

[0053] S1. When the exosomes enter the tangential flow filtration stage, control the liquid inlet actuator to run at a preset tangential flow rate, and collect the reference transmembrane pressure and reference permeation flux of the system under the advection steady state.

[0054] Further, step S1, which involves acquiring the reference transmembrane pressure and reference permeation flux of the system under advection steady state, includes:

[0055] Real-time monitoring of pressure signals at the inlet and outlet ends; confirmation that the fluctuation variance of the pressure signal within a preset time window is lower than the steady-state threshold; and determination that the system has entered advection steady state.

[0056] Based on the system pressure data at each end after entering the advection steady state, the average transmembrane pressure during the steady state time is calculated and extracted as the reference transmembrane pressure.

[0057] Based on the flow rate detection data at the permeation end and the effective filtration area, the permeate volume per unit time is calculated, and a baseline permeation flux is generated and recorded.

[0058] Specifically, when the exosomes enter the tangential flow filtration stage, the main control system sends a control command with a preset tangential flow rate to the inlet actuator, driving the exosome feed solution into the tangential flow filtration membrane. During the feed solution circulation process, a high-frequency pressure sensor installed on the pipeline continuously collects the pressure signals at the inlet and outlet ends at a set sampling frequency.

[0059] To determine the fluid dynamics state, the main control system establishes a preset time window and records discrete pressure sampling data within the window in real time. The system extracts the pressure fluctuation variance within this preset time window and determines whether it is below the steady-state threshold. The formula for calculating the pressure fluctuation variance is:

[0060] ;

[0061] In the formula, This represents the variance of pressure fluctuations within a preset time window. This represents the total number of discrete pressure sampling points within a preset time window. Indicates the first The real-time pressure value at the inlet or outlet of the liquid at each sampling time. This represents the average pressure value within a preset time window. In this scheme, the preset time window is preferably 3 seconds, the sampling frequency of the high-frequency pressure sensor is preferably 500 Hz, and the corresponding total number of discrete sampling points is... The preferred value is 1500, and the preferred steady-state threshold is 0.005 bar squared. When the fluctuation variance of the pressure signal at the inlet and the pressure signal at the return end is consistently lower than the steady-state threshold set by the system, the main control system determines that the flow field distribution in the tangential flow circulation pipeline has reached equilibrium, that is, confirms that the system has entered the advection steady state.

[0062] After the system enters a steady-state advection phase, the main control system calculates the average transmembrane pressure based on the pressure data at each end during this steady-state period and stores it as the reference transmembrane pressure. The formula for calculating the reference transmembrane pressure is:

[0063] ;

[0064] In the formula, This represents the reference transmembrane pressure under steady-state advection. This represents the total number of samples taken during the duration of the advection steady state. Indicates the first Real-time pressure at the inlet end at each sampling time Indicates the first Real-time pressure at the return end at each sampling time. Indicates the first The real-time pressure at the permeation end at each sampling time. In this scheme, the preferred duration of the advection steady state is 10 seconds, corresponding to the total number of sampling times. The preferred value is 5000.

[0065] Simultaneously, the main control system receives the mass flow rate data transmitted from the permeate flow meter and, combined with the known physical parameters of the tangential flow filtration membrane pack, calculates and generates a baseline permeation flux. The conversion formula for the baseline permeation flux is:

[0066] ;

[0067] In the formula, This represents the baseline permeation flux under advection steady-state conditions. This represents the average mass flow rate of the permeate at the permeate end during the advection steady-state period. This indicates the fluid density of the permeate. This represents the effective filtration area of ​​the tangential flow filtration membrane pack. In this scheme, the fluid density of the permeate is... The preferred value is 1000 kg / m³, with an effective filtration area of The preferred value is 0.1 square meters. The main control system writes the calculated baseline transmembrane pressure and baseline permeation flux into the storage unit as a reference for subsequent parameter optimization.

[0068] This step uses a variance model to objectively quantify the flow field state mathematically, avoiding errors caused by extracting data under unstable conditions, and providing accurate initial reference parameters for subsequent search of hydrodynamic resonant frequencies.

[0069] S2. Control the fluid regulating element at the return end to superimpose a sweeping perturbation signal with continuously varying frequency on the basis of the reference transmembrane pressure, and simultaneously collect the real-time permeation flux corresponding to each perturbation frequency.

[0070] Further, step S2, the step of superimposing a frequency-sweeping disturbance signal with continuously varying frequency, includes:

[0071] Construct a frequency sweep control vector that includes frequency sweep boundary parameters and step resolution to limit the frequency sweep disturbance signal to be within a preset tolerance frequency band;

[0072] The sweep frequency control vector is converted into a drive signal to control the fluid regulating component at the return end to generate continuous opening oscillations at the corresponding frequency.

[0073] During continuous opening oscillation, the output power of the liquid inlet actuator is dynamically compensated to maintain the preset tangential flow velocity from interference by the superposition of frequency sweep disturbance signals.

[0074] Specifically, based on the pre-set physical tolerance limits of exosome vesicles, the main control system extracts the start and end frequencies to construct a sweep frequency control vector that includes sweep frequency boundary parameters and step resolution. The system uses a linear sweep frequency mode to generate a continuously varying sweep frequency perturbation signal, ensuring that this perturbation signal strictly remains within a preset tolerance frequency band. The mathematical expression for the sweep frequency perturbation signal is:

[0075] ;

[0076] In the formula, Indicates at time The generated opening perturbation increment This represents the preset valve opening oscillation amplitude scalar value. Indicates the starting frequency of the tolerance band. Indicates the end frequency of the tolerance band. This represents the total duration of a single complete frequency sweep cycle. In this scheme, the valve opening oscillation amplitude is a scalar value. Preferably, it is 5% of the reference opening, and the starting frequency is... Preferably 1 Hz, stop frequency Preferably 20 Hz, the total duration of a single complete frequency sweep cycle The preferred setting is 60 seconds.

[0077] The main control system converts the aforementioned frequency sweep control vector into a standard voltage drive signal and superimposes it onto the steady-state reference opening voltage required to maintain the reference transmembrane pressure. The superimposed integrated drive signal is transmitted to the fluid regulator at the return end, controlling the valve core of the proportional control valve to generate continuous opening oscillations at the corresponding frequency, thereby generating alternating hydrodynamic pressure waves in the return pipeline. During the continuous oscillation frequency sweep of the valve core, the main control system synchronously collects the real-time permeation flux corresponding to each disturbance frequency node through a high-frequency data bus and writes the frequency-permeation flux data pairs into a dynamic buffer.

[0078] Because continuous oscillations in the return flow angle cause periodic and rapid changes in system flow resistance, the main control system activates closed-loop dynamic compensation control logic during the oscillation process to prevent macroscopic attenuation of the main tangential flow field. The system acquires the actual tangential flow velocity measured by the inlet flowmeter in real time and uses a proportional-integral-derivative control algorithm to dynamically compensate the output power of the inlet actuator, thus maintaining the preset tangential flow velocity unaffected by the superposition of frequency sweep disturbance signals. The output power compensation formula for the inlet actuator is:

[0079] ;

[0080] In the formula, Indicates the time of liquid inlet actuator Target output power, This represents the reference power required to maintain a preset tangential flow velocity under advection steady-state conditions. This indicates the difference between the preset tangential velocity and the actual tangential velocity at time [time value missing]. The flow velocity deviation value, Indicates proportional compensation gain. Indicates integral compensation gain. This represents the differential compensation gain. This represents the integral variable. In this scheme, the proportional compensation gain... The preferred value is 1.2, with integral compensation gain. The preferred value is 0.5, with differential compensation gain. The preferred value is 0.1.

[0081] This step, by introducing continuous-frequency microfluidic disturbances and supplementing them with a dynamic tangential velocity compensation mechanism, traversed the hydrodynamic frequency domain characteristics of the system without compromising the stability of the tangential flow field, providing a complete data sample for subsequently locking the optimal resonance point.

[0082] S3. Based on the mapping relationship between each disturbance frequency and the synchronously acquired real-time infiltration flux, extract the characteristic frequency corresponding to the peak value of the real-time infiltration flux and lock it as the initial hydrodynamic resonant frequency.

[0083] Furthermore, step S3, which involves establishing a mapping relationship between each disturbance frequency and the synchronously acquired real-time permeation flux, includes:

[0084] The collected real-time permeation flux is filtered to remove background fluid noise introduced by the mechanical operation of the system.

[0085] Based on the time synchronization identifier, the filtered real-time penetration flux is matched and aligned with the output perturbation frequencies in the time domain to construct a two-dimensional sample set;

[0086] A smoothing fitting operation is performed on the two-dimensional sample set to generate a continuous frequency-flux response curve, thus constructing a mapping relationship.

[0087] Further, step S3, the step of extracting the characteristic frequency corresponding to the peak value that triggers the real-time permeation flux, includes:

[0088] Calculate the derivative change characteristics of the mapping relationship and search for candidate extreme points where the first derivative is zero and the second derivative is negative;

[0089] When the real-time permeation flux corresponding to the candidate extreme point is greater than the baseline permeation flux, the candidate extreme point is verified as an effective resonant peak.

[0090] Extract the frequency coordinates corresponding to the effective resonance peak, use them as the characteristic frequency, and lock and record them as the initial hydrodynamic resonance frequency.

[0091] Specifically, the main control system reads the continuously acquired real-time permeation flux data sequence from the dynamic buffer. To eliminate high-frequency noise caused by the mechanical pulsation of the inlet actuator rotor and the inherent resonance of the pipeline, the system performs low-pass filtering on the acquired real-time permeation flux to filter out background fluid noise introduced by the system's mechanical operation. The main control system uses a digital Butterworth filter to smooth the permeation flux data sequence; the filtering difference equation expression is as follows:

[0092] ;

[0093] In the formula, Indicates time Pure real-time permeation flux after filtering. Indicates time The raw, real-time permeation flux collected includes background fluid noise. This represents the feedforward tap coefficients of the filter. This represents the feedback tap coefficients of the filter. and These represent the filter orders for the feedforward and feedback, respectively. In this scheme, the low-pass filter is preferably a second-order Butterworth filter, with a feedforward order of... With feedback order The preferred value is 2.

[0094] After filtering, the main control system, based on the time synchronization identifier generated by the underlying hardware clock, performs time-domain matching and alignment between the filtered real-time permeation flux and the output perturbation frequencies, pairing frequency values ​​and flux values ​​at the same timestamp to construct a two-dimensional sample set. Subsequently, the system performs a smoothing fitting operation on the two-dimensional sample set to generate a continuous frequency-flux response curve, thereby establishing the mapping relationship between frequency and flux. The system uses a least-squares polynomial fitting algorithm to construct the mapping relationship; the fitting function expression is:

[0095] ;

[0096] In the formula, This indicates that the frequency-flux response curve generated after fitting is at a given perturbation frequency. The continuous permeation flux mapping value under the following conditions This represents the coefficients obtained through polynomial regression. This represents the preset order of the fitting polynomial. In this scheme, the preset order of the fitting polynomial is... The preferred value is 6.

[0097] After obtaining the mapping relationship, the main control system calculates the derivative change characteristics of the mapping relationship and searches for candidate extrema points where the first derivative is zero and the second derivative is negative. The system performs discrete differential operations on the frequency-flux response curve, and the mathematical criteria that candidate extrema points must satisfy are:

[0098] ;

[0099] In the formula, It represents the characteristic of the change of the first derivative corresponding to the frequency-flux response curve, reflecting the gradient direction of the permeation flux as a function of frequency; The characteristic of the second derivative change corresponding to the frequency-flux response curve is used to confirm that the gradient stationary point is a local maximum. The main control system records the frequency points that satisfy the above conditions within the scanning range as candidate extreme points.

[0100] To eliminate spurious disturbance peaks, the main control system compares the real-time seepage flux corresponding to the candidate extreme point with the baseline seepage flux recorded under steady-state advection. When the real-time seepage flux corresponding to the candidate extreme point is significantly greater than the baseline seepage flux, the main control system verifies that the candidate extreme point is a valid resonant peak. The system extracts the frequency coordinates corresponding to this valid resonant peak, uses it as the characteristic frequency, and writes it into the main control storage register to lock it as the initial hydrodynamic resonant frequency.

[0101] This step eliminates sensor sampling errors through digital filtering and polynomial fitting, and uses rigorous derivative features to accurately identify the hydrodynamic resonance point that maximizes the permeability of the filter membrane, providing a precise frequency basis for subsequent micro-amplitude oscillation filtration.

[0102] S4. Control the fluid regulating component at the return end to oscillate alternately at the initial hydrodynamic resonant frequency, and limit the transmembrane pressure fluctuation amplitude caused by the alternating oscillation to be within a preset amplitude threshold.

[0103] Further, step S4, the step of limiting the transmembrane pressure fluctuation amplitude caused by alternating oscillation within a preset amplitude threshold, includes:

[0104] Based on the reference transmembrane pressure and the preset elastic proportional coefficient, the allowable pressure boundary for fluctuation is calculated and generated as the preset amplitude threshold.

[0105] During the alternating oscillation execution, the fluctuation envelope signal of the current system transient transmembrane pressure is continuously extracted;

[0106] When the fluctuation envelope signal reaches the allowable pressure boundary, the amplitude of the alternating oscillation is adaptively decayed to keep it within the preset amplitude threshold.

[0107] Specifically, the main control system acquires the baseline transmembrane pressure recorded under advection steady-state conditions and reads a pre-configured elastic scaling factor characterizing the deformation tolerance of the exosome vesicle lipid bilayer. Based on the baseline transmembrane pressure and the preset elastic scaling factor, the system calculates and generates an allowable pressure boundary that permits transmembrane pressure fluctuations, and uses this allowable pressure boundary as a preset amplitude threshold. The formula for calculating the allowable pressure boundary is:

[0108] ;

[0109] In the formula, This represents the preset amplitude threshold for the transmembrane pressure deviation from the reference value during alternating oscillations. This represents the reference transmembrane pressure under steady-state advection. This represents the preset elasticity ratio coefficient. In this scheme, the elasticity ratio coefficient... The preferred value is 0.1.

[0110] The main control system generates an alternating drive electrical signal corresponding to the initial hydrodynamic resonant frequency, controlling the fluid regulating components at the return end to perform alternating oscillations. During the alternating oscillation, the main control system acquires high-frequency pressure data at each end of the system to calculate the transient transmembrane pressure, and continuously extracts the fluctuation envelope signal of the current system's transient transmembrane pressure using a sliding time window algorithm. The formula for extracting the fluctuation envelope signal is:

[0111] ;

[0112] In the formula, Indicates at discrete time Extracted transient transmembrane pressure fluctuation envelope signal, This represents the length of the sliding time window that covers at least one complete alternating oscillation cycle. Indicates at time The system calculates the transient transmembrane pressure in real time. In this scheme, the sliding time window length... The preferred value is 100.

[0113] The main control system performs a high-frequency comparison between the real-time extracted fluctuation envelope signal and the allowable pressure boundary. When the fluctuation envelope signal reaches the allowable pressure boundary, the main control system immediately triggers the amplitude attenuation control mechanism. By updating the amplitude parameters of the return fluid regulating component drive signal, it adaptively attenuates the amplitude of the alternating oscillation, forcing the pressure fluctuation caused by the alternating oscillation to retreat and remain within the preset amplitude threshold. The adaptive attenuation update formula for the alternating oscillation amplitude is:

[0114] ;

[0115] In the formula, This indicates the amplitude of the driving signal at the next moment after the decay update. This represents the amplitude of the driving signal at the current moment. This represents the adaptive decay step size coefficient. In this scheme, the adaptive decay step size coefficient... The preferred value is 0.15.

[0116] This step, through rigorous monitoring of transient transmembrane pressure envelope and closed-loop attenuation regulation, ensures that hydrodynamic alternating oscillations only peel off the polarized gel layer on the membrane surface, effectively avoiding the physical rupture of exosome vesicles caused by macroscopic alternating pressure waves.

[0117] S5. During the continuous washing and filtration process, a test frequency signal is injected into the fluid regulating component at the return liquid end according to the set replacement cycle, and the current optimal hydrodynamic resonant frequency is dynamically updated based on the differential change gradient of the real-time permeation flux.

[0118] Further, step S5, the step of injecting a test frequency signal into the return fluid regulating element according to a set replacement cycle, includes:

[0119] The accumulated amount of liquid at the permeation end is continuously monitored. When the accumulated amount of liquid meets the set replacement cycle, a frequency test action is triggered.

[0120] Using the current optimal hydrodynamic resonance frequency as a reference point, positive and negative test perturbation signals with preset frequency step lengths are generated respectively.

[0121] The positive and negative test perturbation signals are alternately output to the fluid control device at the return end, and the average permeation flux in both directions within the test time window is recorded independently.

[0122] Furthermore, step S5, the step of updating the current optimal hydrodynamic resonant frequency, includes:

[0123] Extract the direction and magnitude of the difference in real-time permeation flux caused by injecting test frequency signals in different frequency offset directions, and calculate the current differential gradient by combining the offset step size of the test frequency signals.

[0124] By introducing a preset search gain parameter, the differential gradient is transformed into a frequency correction step size that approaches the target extreme point;

[0125] The current optimal hydrodynamic resonant frequency is superimposed and compensated by using the frequency correction step size, driving the frequency to slide towards the gradient rising side, and then updating and obtaining a new current optimal hydrodynamic resonant frequency.

[0126] Specifically, the main control system continuously monitors the accumulated liquid volume at the permeation end by performing continuous integral calculations on the instantaneous mass flow rate at the permeation end. When the calculated increase in the accumulated liquid volume meets the set replacement cycle volume, the system automatically triggers a frequency probing action under steady-state filtration conditions. The formula for determining the accumulated liquid volume is:

[0127] ;

[0128] In the formula, This represents the timestamp of the last completed frequency update. Indicates the current continuous monitoring time. Indicates time Instantaneous mass flow rate at the permeation end, This indicates the fluid density of the washing permeate. This indicates the set volume of a single-trigger replacement cycle. In this scheme, the volume of a single-trigger replacement cycle... The preferred washing and filtration replacement volume is 0.5 times.

[0129] After triggering the frequency test, the main control system uses the current optimal hydrodynamic resonant frequency as a reference point to generate positive and negative test perturbation signals with preset frequency step lengths. The system alternately outputs these signals to the return fluid regulating component, driving the valve to briefly and alternately change the frequency of its opening oscillation. The permeation flow sensor at the permeation end independently records the average permeation flux in both test directions within a preset test time window. The formula for calculating the average permeation flux in both directions is:

[0130] ;

[0131] ;

[0132] In the formula, and These represent the average permeation flux recorded during the positive and negative testing phases, respectively. This indicates the duration of the unidirectional probing time window. This indicates the effective filtration area of ​​the tangential flow filtration membrane package. and These represent the instantaneous mass flow rates collected during the injection of positive and negative probe frequency signals, respectively. In this scheme, the duration of the single-direction probe time window... The preferred setting is 15 seconds.

[0133] After obtaining the average permeation flux in two directions, the main control system extracts the direction and magnitude of the difference in real-time permeation flux caused by injecting test frequency signals in different frequency offset directions. Combined with the offset step size of the test frequency signals, it calculates the differential gradient reflecting the current pipeline fluid resonance characteristics. The main control system introduces a preset search gain parameter to convert the differential gradient into a frequency correction step size that approaches the target extreme point. This frequency correction step size is used to superimpose compensation on the current optimal hydrodynamic resonance frequency, driving the frequency to slide towards the gradient's upward side, thereby updating and obtaining a new current optimal hydrodynamic resonance frequency for subsequent control. The mathematical evolution formula for dynamic frequency updating is:

[0134] ;

[0135] In the formula, This indicates the new, currently optimal hydrodynamic resonant frequency obtained through the slip update. This indicates the current optimal hydrodynamic resonant frequency before the update operation. This represents the preset search gain parameter. This represents the offset step size of the probe frequency signal relative to the reference point. This represents the current differential gradient obtained from the solution. In this scheme, the search gain parameter... The preferred value is 0.05, and the offset step size is... The preferred setting is 0.5 Hz.

[0136] This step uses a closed-loop testing mechanism to track the resonant point shift caused by changes in the physical properties of the feed liquid in real time, ensuring that the filtration system always operates at the optimal frequency boundary of the barrier-free gel layer during the long continuous washing and filtration process.

[0137] S6. During operation at the current optimal hydrodynamic resonant frequency, the impurity concentration characteristic signal at the permeation end is monitored synchronously. When the attenuation slope of the impurity concentration characteristic signal reaches the preset termination condition, the tangential flow washing stage is adaptively stopped.

[0138] Furthermore, step S6, the step of adaptively stopping the tangential flow washing and filtering stage, includes:

[0139] Using a concentration detection unit configured at the permeation end, a sequence signal characterizing the content of free impurities is continuously acquired as an impurity concentration feature signal;

[0140] A sliding analysis window is applied to the impurity concentration characteristic signal to calculate the linear rate of change of the signal intensity within the window, which is then extracted as the attenuation slope.

[0141] When the absolute value of the attenuation slope remains below the zero tolerance parameter for a preset time period, a stop command is generated to adaptively stop the tangential flow washing and filtration stage.

[0142] Specifically, during the tangential flow washing and filtration stage at the current optimal hydrodynamic resonant frequency, the main control system uses an optical ultraviolet absorption detector configured in the permeation end pipeline as a concentration detection unit to continuously collect absorbance sequence signals characterizing the content of free impurity proteins, and stores them as impurity concentration characteristic signals in the system memory.

[0143] To assess the attenuation trend of impurity elution in real time, the main control system applies a sliding analysis window to the impurity concentration characteristic signal. The linear rate of change of signal intensity within this window is calculated using the least squares method, and extracted as the attenuation slope. The formula for calculating the attenuation slope is:

[0144] ;

[0145] In the formula, Indicates the current data collection time The calculated attenuation slope, This indicates the total number of discrete sampling points contained within the sliding analysis window. Indicates the first in the sliding analysis window The timestamp of each sampling point Representation and timestamp The corresponding impurity concentration characteristic signal intensity, This represents the average of all timestamps within the sliding analysis window. This represents the average value of the impurity concentration characteristic signal intensity within the sliding analysis window. In this scheme, the sampling period of the concentration detection unit is preferably 1 second, and the total number of discrete sampling points included in the sliding analysis window is... The preferred value is 120.

[0146] The main control system refreshes and outputs the decay slope for each sampling cycle, and continuously compares its absolute value with the system's preset zero-tolerance parameter to determine whether the impurity elution process has approached its limit. The logical evaluation formula for stopping the process is as follows:

[0147] ;

[0148] In the formula, This indicates the total number of consecutive sampling points included within a preset time period. This represents the continuous decay slope sequence within the current time and the backtracking historical window. This represents the zero-tolerance parameter indicating that the concentration no longer shows a substantial decrease. In this scheme, it refers to the total number of consecutive sampling points included within the preset time period. The preferred value is 3, a zero-tolerance parameter. The preferred value is 0.001.

[0149] When the conditions of the above logical evaluation formula are met, that is, when the maximum absolute value of the attenuation slope remains below the zero tolerance parameter for a preset time period, the main control system determines that the free impurities have been completely washed away. At this time, the main control system generates a digital stop command and sends it to the inlet actuator and the return fluid regulator to cut off the fluid power and close the valve, thereby adaptively stopping the tangential flow washing and filtration stage.

[0150] This step upgrades the fixed-volume empirical rinsing mode to an objective closed-loop control driven by signal data, avoiding the ineffective replacement of expensive buffer solutions.

[0151] Example 2

[0152] The following are examples of applications in the purification process of mesenchymal stem cell-derived exosomes, such as... Figure 2 As shown. The tangential flow filtration hardware selected for this scenario includes a hollow fiber membrane pack with a molecular weight cutoff of 500 kDa and an effective membrane area of ​​0.1 square meters, a variable frequency peristaltic pump as the inlet actuator, a high-frequency pneumatic proportional valve as the return fluid regulating device, and a Coriolis mass flow meter and a UV280 ultraviolet absorption detector connected in series at the permeation end.

[0153] When the exosomes enter the tangential flow filtration stage, the main control system controls the variable frequency peristaltic pump to start at a preset flow rate of 150 mL / min. After 3 minutes of circulation, the system pressure variance is below 0.005, indicating that it has entered a steady-state flow. At this time, the average baseline transmembrane pressure extracted is 0.80 bar, and the baseline permeate flux is 25.0.

[0154] After reaching steady state, the system is set to a frequency sweep range of 1 to 20 Hz, the proportional valve performs continuous opening oscillation for 60 seconds, and the peristaltic pump synchronously and dynamically compensates for the tangential flow velocity. After filtering and fitting the acquired instantaneous permeation flux, the system identifies that the derivative characteristics satisfy the extreme value condition when the disturbance frequency is 8.5 Hz, and the permeation flux reaches a peak value of 32.5 Hz. Subsequently, 8.5 Hz is locked as the initial hydrodynamic resonant frequency.

[0155] The proportional valve then performs alternating oscillations at 8.5 Hz. Combined with the exosome elastic proportional setting system, the allowable pressure boundary is set to 0.08 bar. When high-frequency sampling detects that the transient transmembrane pressure fluctuation deviates from the reference value by 0.085 bar and touches the boundary, the system immediately reduces the proportional valve oscillation amplitude by 15%, stabilizing subsequent pressure fluctuations within 0.075 bar and successfully preventing pressure waves from damaging the exosome structure.

[0156] During the filtration process, the system is set to trigger an extreme value search every 0.5 filtration volumes. When 1.5 filtration volumes are reached, the system probes at 8.0 Hz and 9.0 Hz, finding that the flux is higher at 9.0 Hz, and then shifts and updates the optimal hydrodynamic resonant frequency to 8.8 Hz. When filtration reaches 4.8 filtration volumes, the decay slope of the impurity concentration collected by the UV detector is continuously below the zero tolerance of 0.001, indicating that the free impurities have been washed away, and the system automatically issues a command to smoothly terminate the process.

[0157] Under identical experimental conditions, maintaining the same hardware fluid pipeline configuration and using the same batch and concentration of mesenchymal stem cell-derived exosome initial feed solution, the adaptive resonant filtration scheme of this invention was compared with the traditional constant flow rinsing process in parallel. The detailed comparison of its key process indicators is shown in Table 1:

[0158] Table 1. Comparison of purification effects between the present invention's scheme and the traditional advection flushing mode.

[0159] Comparison indicators Traditional methods Invention Solution Average permeation flux 16.5 LMH 24.2 LMH Filter replacement volume requirements 7.5DV 4.8DV Overall process time 210 minutes 135 minutes

[0160] As can be seen from the above comparative data, this invention, by adaptively tracking the hydrodynamic resonant frequency and applying pressure-controlled micro-amplitude alternating oscillations, disrupts the polarized gel layer attached to the filter membrane surface, reducing the mass transfer resistance of encapsulated impurities. This increases the average permeate flux from 16.5 LMH in the traditional advection flushing mode to 24.2 LMH. Simultaneously, thanks to the adaptive shutdown determination mechanism based on the impurity concentration decay slope, fluid displacement is avoided in traditional methods, reducing the required washing and filtration displacement volume from the traditional 7.5 DV to 4.8 DV. With the dual benefits of increased flux and reduced washing and filtration volume, the overall process time is shortened from the traditional 210 minutes to 135 minutes. This demonstrates that the method of this invention can effectively improve the efficiency of exosome tangential flow filtration, achieving high efficiency and intelligence in the purification process while reducing pharmaceutical buffer consumption and time costs.

[0161] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive optimization method for exosome tangential flow filtration process parameters, characterized in that, include: S1. When the exosomes enter the tangential flow filtration stage, control the liquid inlet actuator to run at a preset tangential flow rate, and collect the reference transmembrane pressure and reference permeation flux of the system under the advection steady state. S2. The fluid regulating component at the return end is controlled to superimpose a frequency sweeping disturbance signal with continuously varying frequency on the basis of the reference transmembrane pressure, and the real-time permeation flux corresponding to each disturbance frequency is collected simultaneously. S3. Based on the mapping relationship between each disturbance frequency and the synchronously acquired real-time infiltration flux, extract the characteristic frequency corresponding to the peak value of the real-time infiltration flux and lock it as the initial hydrodynamic resonant frequency. S4. Control the fluid regulating component at the return end to oscillate alternately at the initial hydrodynamic resonant frequency, and limit the transmembrane pressure fluctuation amplitude caused by the alternating oscillation to within a preset amplitude threshold. S5. During the continuous washing and filtration process, a test frequency signal is injected into the fluid regulating component at the return end according to the set replacement cycle, and the current optimal hydrodynamic resonant frequency is dynamically updated based on the differential change gradient of the real-time permeation flux. S6. During operation at the current optimal hydrodynamic resonant frequency, the impurity concentration characteristic signal at the permeation end is monitored synchronously. When the attenuation slope of the impurity concentration characteristic signal reaches the preset termination condition, the tangential flow washing stage is adaptively stopped.

2. The adaptive optimization method for exosome tangential flow filtration process parameters according to claim 1, characterized in that, Step S1, which involves the acquisition system being in a steady-state state with reference transmembrane pressure and reference permeation flux, includes: Real-time monitoring of pressure signals at the inlet and outlet ends; confirmation that the fluctuation variance of the pressure signal within a preset time window is lower than the steady-state threshold; and determination that the system has entered advection steady state. Based on the system pressure data at each end after entering the advection steady state, the average transmembrane pressure during the steady state time is calculated and extracted as the reference transmembrane pressure. Based on the flow rate detection data at the permeation end and the effective filtration area, the permeate volume per unit time is calculated, and the baseline permeation flux is generated and recorded.

3. The adaptive optimization method for exosome tangential flow filtration process parameters according to claim 1, characterized in that, Step S2, the step of superimposing a frequency-sweeping disturbance signal with continuously varying frequency, includes: Construct a frequency sweep control vector that includes frequency sweep boundary parameters and step resolution to limit the frequency sweep disturbance signal to be within a preset tolerance frequency band; The frequency sweep control vector is converted into a drive signal to control the return fluid regulator to generate a continuous opening oscillation at a corresponding frequency. During the continuous opening oscillation, the output power of the liquid inlet actuator is dynamically compensated to maintain the preset tangential flow velocity from interference by the superposition of the frequency sweep disturbance signal.

4. The adaptive optimization method for exosome tangential flow filtration process parameters according to claim 1, characterized in that, Step S3, the step of mapping the perturbation frequencies to the synchronously acquired real-time permeation flux, includes: The collected real-time permeation flux is filtered to remove background fluid noise introduced by the mechanical operation of the system. Based on the time synchronization identifier, the filtered real-time penetration flux and the output perturbation frequencies are matched and aligned in the time domain to construct a two-dimensional sample set; A smoothing fitting operation is performed on the two-dimensional sample set to generate a continuous frequency-flux response curve, thus constructing the mapping relationship.

5. The adaptive optimization method for exosome tangential flow filtration process parameters according to claim 1, characterized in that, Step S3, the step of extracting the characteristic frequency corresponding to the peak value that triggers the real-time permeation flux, includes: Calculate the derivative change characteristics of the mapping relationship and search for candidate extreme points where the first derivative is zero and the second derivative is negative; When the real-time permeation flux corresponding to the candidate extreme point is greater than the benchmark permeation flux, the candidate extreme point is verified as an effective resonant peak. Extract the frequency coordinates corresponding to the effective resonance peak, use them as the characteristic frequency, and lock and record them as the initial hydrodynamic resonance frequency.

6. The adaptive optimization method for exosome tangential flow filtration process parameters according to claim 1, characterized in that, In step S4, the step of limiting the transmembrane pressure fluctuation amplitude caused by the alternating oscillation within a preset amplitude threshold includes: Based on the reference transmembrane pressure and the preset elastic proportional coefficient, the allowable pressure boundary for fluctuation is calculated and generated as the preset amplitude threshold. During the execution of the alternating oscillation, the fluctuation envelope signal of the current system transient transmembrane pressure is continuously extracted; When the fluctuation envelope signal touches the allowable pressure boundary, the amplitude of the alternating oscillation is adaptively attenuated to keep it within the preset amplitude threshold.

7. The adaptive optimization method for exosome tangential flow filtration process parameters according to claim 1, characterized in that, Step S5, the step of injecting a test frequency signal into the return fluid regulator according to a set replacement cycle, includes: The accumulated amount of liquid at the permeation end is continuously monitored. When the accumulated amount of liquid meets the set replacement cycle, a frequency test action is triggered. Using the current optimal hydrodynamic resonance frequency as a reference point, positive and negative test perturbation signals with preset frequency step lengths are generated respectively. The positive and negative test perturbation signals are alternately output to the return fluid regulating element, and the average permeation flux in both directions within the test time window is recorded independently.

8. The adaptive optimization method for exosome tangential flow filtration process parameters according to claim 1, characterized in that, Step S5, the step of updating the current optimal hydrodynamic resonant frequency, includes: Extract the direction and magnitude of the difference in real-time permeation flux caused by injecting the test frequency signal in different frequency offset directions, and calculate the current differential change gradient by combining the offset step size of the test frequency signal; By introducing a preset search gain parameter, the differential gradient is transformed into a frequency correction step size that approaches the target extreme point; The frequency correction step size is used to superimpose and compensate the current optimal hydrodynamic resonant frequency, driving the frequency to slide towards the gradient rising side, thereby updating and obtaining a new current optimal hydrodynamic resonant frequency.

9. The adaptive optimization method for exosome tangential flow filtration process parameters according to claim 1, characterized in that, In step S6, the step of adaptively stopping the tangential flow washing and filtration stage includes: Using a concentration detection unit configured at the permeation end, a sequence signal characterizing the content of free impurities is continuously acquired as the impurity concentration feature signal; A sliding analysis window is applied to the impurity concentration characteristic signal, and the linear rate of change of the signal intensity within the window is calculated and extracted as the attenuation slope. When the absolute value of the attenuation slope remains below the zero tolerance parameter for a preset time period, a stop command is generated to adaptively stop the tangential flow washing and filtration stage.

Images