Drug screening method based on single virus tracing and machine learning and application

Abstract

The invention provides a drug screening method based on single virus tracing and machine learning, which comprises the following steps: S1, attaching a biotinylated virus to a host cell membrane, and marking by using a quantum dot modified by streptavidin to obtain a cell sample combined with the virus marked by the quantum dot; s2, analyzing the image sequence of the biotinylated virus infected host cell to obtain a single virus infection track; s3, inputting the coordinate information of the plurality of single-virus infection tracks into a machine learning model for calculation and analysis to obtain single-virus infection fingerprints under corresponding processing; and S4, evaluating the antiviral effect of the drug according to the single-virus infection fingerprint difference of the drug treatment group and the control group. According to the invention, a rapid and visual drug screening technology is developed by using a single virus tracing technology and machine learning.

Description

Technical Field

[0001] The present invention belongs to the field of biomedicine technology, and in particular relates to a drug screening method and application based on single virus tracing and machine learning. Background Art

[0002] The emergence and spread of novel viruses pose a major threat to human health and the global socioeconomic landscape, creating a huge economic and social security burden. Accelerating the development of antiviral drugs can provide faster and more effective treatments to reduce the spread and impact of viruses. Despite the rapid development of modern scientific research, drug development often faces challenges such as long development cycles and high costs. Reusing approved drugs or candidate drugs that have undergone clinical evaluation is a viable strategy to shorten the drug development cycle. Compared with developing new drugs for specific indications, this approach has many advantages, such as reducing the risk of failure, shortening drug development time, and reducing investment. Currently, this strategy has been used to screen antiviral drugs for Zika virus, Ebola virus, hepatitis C virus, influenza virus, and dengue virus.

[0003] At present, there are two main approaches to drug repurposing. The first approach is based on the repurposing of virus-related research content, which relies on the understanding of disease-related pathways, drug molecular structures, or target mechanisms to screen drugs. The second approach does not require the understanding of targets and disease mechanisms, but directly screens phenotypes. For acute infectious diseases such as COVID-19, the rapid use of methods based on reported studies is limited due to the lack of understanding of emerging viruses. Therefore, phenotypic screening often helps to speed up the screening process of antiviral drugs. At present, phenotypic screening mainly uses the host cells of the target virus to interact with candidate compounds, and then evaluates the antiviral effect of candidate molecules by measuring cell viability after virus infection, pathway-related key molecule content, or viral replication level. Although these traditional phenotypic screening methods are helpful for screening drugs against emerging viruses, they often require cumbersome steps to simultaneously verify various aspects of the virus and cells, and lack intuitive evidence that the drug can prevent viral infection. In addition, for new viruses with high lethality and high transmission rates, the traditional phenotypic screening cycle still needs to be shortened. Therefore, there is still an urgent need to develop rapid and intuitive antiviral drug phenotypic screening methods to quickly identify candidate drugs. Summary of the invention

[0004] In view of this, the present invention aims to propose a drug screening method and application based on single virus tracing and machine learning to solve the problem in the prior art that it is impossible to quickly and intuitively perform phenotypic screening of antiviral drugs to quickly identify candidate drugs.

[0005] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0006] The first aspect of the present invention provides a drug screening method based on single virus tracing and machine learning, comprising the following steps:

[0007] S1: attaching biotinylated viruses to host cell membranes and labeling them with streptavidin-modified quantum dots (QDs) to obtain cell samples bound to the quantum dot-labeled viruses;

[0008] S2: Analyze the image sequences of biotinylated virus-infected host cells to obtain several single virus infection tracks;

[0009] S3: Input the coordinate information of several single virus infection trajectories into the machine learning model for calculation and analysis to obtain the single virus infection fingerprint (SVIF) under corresponding processing;

[0010] S4: Identify and evaluate the antiviral effect of drugs based on the differences in single virus infection fingerprints between the drug-treated group and the control group.

[0011] Furthermore, the S1 specifically includes the following steps:

[0012] The biotinylated virus was incubated with the host cells at 4°C for 5 minutes, and the unattached biotinylated virus was removed by washing. 3 nM streptavidin-modified quantum dots (SA-QDs) were added to label the attached biotinylated virus at 4°C for 5 minutes, and the unbound streptavidin-modified quantum dots were removed by washing.

[0013] Furthermore, the S2 is to image the process of biotinylated virus infecting host cells in the live cell online culture system of the confocal microscope, and to trace the virus entry and transport process in the live cell online culture system of the confocal microscope, thereby drawing the movement trajectory of a single virus.

[0014] Furthermore, the specific steps of S3 are: 4 The XY coordinate information of each virus trajectory is input into the logistic regression model. Through calculation and analysis, a matrix with 17 features is obtained. Multiple matrices constitute a single virus infection fingerprint (SVIF).

[0015] Furthermore, the 17 features in S3 are the proportion of hidden Markov state 0 (T0), the proportion of hidden Markov state 1 (T1), the proportion of hidden Markov state 2 (T2), the proportion of hidden Markov state 3 (T3), lifespan, average step length, kurtosis, Gaussianity, fractal dimension, efficiency, sleepiness, average MSD, diffusion coefficient D, MSD curve power law fitting quality P value, power law scaling coefficient alpha, MSD ratio, and number of frames N.

[0016] Furthermore, the specific steps of S4 are: calculating the average Euclidean distance between the drug treatment group and the control group, taking the Euclidean distance within the control group as the baseline, calculating the antiviral score, and thus determining the antiviral effect of the drug.

[0017] The Euclidean distance within the control group was 0.21.

[0018] Furthermore, the method for preparing the biotinylated virus is: incubating and purifying the virus with biotin to obtain the biotinylated virus.

[0019] The biotin is EZ-link Sulfo-NHS-LC-biotin, purchased from Thermo.

[0020] The virus is an influenza virus.

[0021] The incubation condition is 30 minutes at room temperature.

[0022] The specific steps of the purification are: using a gel filtration column and a filter to remove free biotin and virus aggregates.

[0023] The gel filtration column is of model NAP-5, purchased from GE Healthcare. The pore size of the filter is 0.22 μm.

[0024] Furthermore, before the host cells are combined with the biotinylated virus, the process also includes pre-treating the host cells with antiviral drugs at 37° C. for 30 minutes.

[0025] Furthermore, the preparation method of the modified quantum dots is:

[0026] CdZnSeS organic phase quantum dots of different sizes were prepared by changing the ratio of Zn / S, and the quantum dots dispersed in n-hexane were washed with methanol and resuspended in chloroform. Excess OPA was dissolved in chloroform and slowly added to the quantum dot solution, ultrasonically treated, and the solvent was removed by rotary evaporator. The quantum dots were redissolved in Britton-Robinson buffer solution and purified by ultracentrifugation to obtain OPA-coated quantum dots.

[0027] The OPA-coated quantum dots, 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC), and H2N-PEG2000-COOH were dissolved in phosphate buffer (PBS) and reacted at room temperature for 2.5 hours to obtain streptavidin-modified quantum dots (SA-QDs). The streptavidin-modified quantum dots were dissolved in phosphate buffer and stored at 4°C for later use.

[0028] The pH of the Britton-Robinson buffer solution is 9.0.

[0029] The molar ratio of the OPA-coated quantum dots, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride, and H2N-PEG2000-COOH is 1:3000:600.

[0030] The second aspect of the present invention provides the use of the method described in the first aspect of the present invention in drug screening.

[0031] Furthermore, the drug is an antiviral drug.

[0032] The third aspect of the present invention provides use of the method described in the first aspect of the present invention in drug repurposing.

[0033] Compared with the prior art, the drug screening method based on single virus tracing and machine learning and its application described in the present invention have the following advantages:

[0034] (1) The present invention provides a rapid and visual drug screening technology using quantum dot-based single virus tracing and machine learning. Thanks to the high brightness and excellent photostability of quantum dots (QDs), this method can generate unique single virus infection fingerprint data from high-fidelity virus trajectories and detect dynamic changes in virus behavior after drug administration. This method can successfully identify viral infection patterns at different infection stages within 90 minutes and predict the efficacy of antiviral drugs through machine learning, providing valuable support for evaluating the efficacy of antiviral drugs and providing broad application prospects for responding to future outbreaks of new viruses.

[0035] (2) The present invention uses single virus tracing technology (SVT) based on quantum dots (QDs), which can track the infection behavior of hundreds or thousands of individual viruses in living cells in real time at a single virus resolution. It can perform high-throughput quantitative analysis of the infection trajectory of a single virus with nanometer-level spatial resolution and millisecond-level temporal resolution, thereby analyzing the interaction between the virus and the host cell that is obscured by the ensemble averaging method. The dynamic infection characteristics extracted by SVT can serve as valuable "fingerprint" information. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0037] Figure 1 This is a flow chart of the single virus tracing technology;

[0038] Figure 2Flowchart for evaluating the efficacy of antiviral drugs using SVIF-based machine learning;

[0039] Figure 3 The infection trajectory of the virus in live MDCK cells within 20 minutes after infection;

[0040] Figure 4 is the cumulative displacement over time of 57 virus transport trajectories collected from 5 MDCK cells;

[0041] Figure 5 For Figure 3 The instantaneous speed of 6 randomly selected trajectories;

[0042] Figure 6 are three representative trajectories (their different degrees of directional deviations show similar fractal dimension values, and the color represents time);

[0043] Figure 7 are the state occupancy, state transition probability, and diffusion coefficient obtained through hidden Markov model (HMM) analysis;

[0044] Figure 8 for Figure 3 HMM distribution of viral transport trajectories in;

[0045] Fig. 9 HMM changes for representative viral transport trajectories;

[0046] Fig.10 for Fig. 9 The results of HMM state changes over time;

[0047] Fig.11 Representative trajectories of infection stage 0 (infection 0-2 minutes), infection stage 1 (infection 2-5 minutes), and infection stage 2 (infection 5-20 minutes);

[0048] Fig.12 are the average and distribution of instantaneous speeds in the three infection stages;

[0049] Fig.13 is the change of HMM state occupancy at different infection stages;

[0050] Fig.14 is the mean step length (meanSL) distribution of the three infection stages;

[0051] Fig.15 is the distribution of fractal dimensions of the three infection stages;

[0052] Fig.16 is the diffusion coefficient distribution of the three stages of virus infection;

[0053] Fig.17 The difficulty distribution of the three stages of virus infection;

[0054] Fig.18 is the Gaussian distribution of the three stages of virus infection;

[0055] Fig.19 The distribution of the duration frames of virus movement in the three stages of virus infection;

[0056] Fig. 20 is the radial histogram of SVIF at stage 0 of viral infection;

[0057] Fig.21 It is the radial histogram of SVIF in the first stage of viral infection;

[0058] Fig. 22 It is the radial histogram of SVIF in the second stage of viral infection;

[0059] Fig.23 Confusion matrix for the trained logistic regression model predicting the three stages of influenza virus infection;

[0060] Fig.24 The PCA three-dimensional projection of the SVIF of the three stages of influenza virus infection;

[0061] Fig.25 It is the LDA one-dimensional projection of the three-stage SVIF of influenza virus infection;

[0062] Fig.26 Confusion matrix for the trained logistic regression model predicting the paclitaxel-treated group and the three-stage infection group;

[0063] Fig. 27 Confusion matrix for the trained logistic regression model predicting the nocodazole-treated group and the three-stage infection group;

[0064] Fig.28 is the PCA three-dimensional projection of SVIF of the paclitaxel-treated group;

[0065] Fig.29 is the PCA three-dimensional projection of the SVIF of the nocodazole-treated group;

[0066] Fig.30 is the normalized distribution of the five most discriminative features between the paclitaxel-treated group and the infection stage II group;

[0067] Fig.31 is the normalized distribution of the five most discriminative features between the nocodazole-treated group and the infection stage II group;

[0068] Fig.32 Heat map of characteristic patterns of viral infection for all treatment groups;

[0069] Fig.33 Comparison of the similarity of SVIF in different drug treatment groups;

[0070] Fig.34 To rank different drugs for interfering with viral transport;

[0071] Fig.35 Representative images of MDCK cells immunofluorescently labeled for IAV nucleoprotein (NP; magenta) and nuclei (Hoechst 33342; cyan) 2 h after infection;

[0072] Fig.36 The total fluorescence intensity of viral nucleoprotein in the nucleus was detected by immunofluorescence in MDCK cells treated with drugs 2 hours after infection;

[0073] Fig.37 The viral RNA production in drug-treated MDCK cells was detected by real-time qPCR 2 hours after infection;

[0074] Fig.38 Real-time qPCR was used to detect viral RNA production in drug-treated MDCK cells 24 hours after infection. DETAILED DESCRIPTION

[0075] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0076] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0077] 1. High-throughput tracking of single viruses in living cells

[0078] The virus is labeled with quantum dots in a two-step method, and then the virus entry and transport process is tracked in a live cell online culture system of a confocal microscope, thereby drawing the movement track of a single virus. The cells are live MDCK cells.

[0079] We have demonstrated that the direction of viral movement is nearly linear and that the movement speeds alternate between fast and slow under normal physiological conditions, as follows:

[0080] Proof that the direction of motion is nearly linear:

[0081] In order to measure the differences in movement directions between virus subpopulations, the fractal dimension was used to quantify the linearity of the virus movement trajectory. The analysis showed that although the direction of the virus movement deviated by more than 90 degrees, the virus movement as a whole still conformed to linearity (e.g. Figure 6 shown).

[0082] Proof of alternating fast and slow movements:

[0083] We studied the cumulative displacement and instantaneous velocity changes of many viruses over time. The cumulative displacement curve is step-shaped and the instantaneous velocity curve is multi-peaked, indicating that there are alternating fast and slow states in the virus infection process (such as Figure 4 and Figure 5 In order to analyze the fast and slow movement states of the virus in more detail, we used the Hidden Markov Model (HMM) to analyze all movement trajectories during virus infection. The HMM divides the virus movement into seven states according to the instantaneous speed, and calculates the proportion of each state in the entire movement process and the transition probability between states. The results show that the diffusion coefficient is less than 0.01μm 2 / s state accounts for more than 80% of viral activity (such as Figure 7 The HMM state transitions of typical trajectories show that the state changes are related to the local cellular environment where the virus particles are located (as shown in Figure 8 and Fig.10 In addition, previous studies have shown that vesicles transporting viruses change speed during virus maturation. Therefore, changes in the local cellular environment and changes in viral structure and function can affect the HMM state. In summary, the transport process of viruses is complex and diverse. Although the transport properties of individual virus particles vary, the transport properties of virus groups under the same experimental conditions are predictable.

[0084] 2. Using 17 features to draw a single virus infection fingerprint (SVIF)

[0085] As influenza virus infection progresses, we studied the changes in viral transport patterns at different stages of infection. The viral infection process was divided into three stages according to the infection time: 0-2 minutes as stage 0, 2-5 minutes as stage 1, and 5-20 minutes as stage 2.

[0086] Figure 11-19 It indicates that the distribution trends of each feature in the three infection stages are different. In summary, the transport characteristics of the virus vary with the degree of infection.

[0087] To highlight these differences more intuitively, we selected a total of 17 features from different perspectives, such as velocity fluctuation, space filling characteristics, transport trajectory linearity, and duration of viral movement, to establish a single virus infection fingerprint (SVIF) (e.g. Figure 20-22 By comparing the infection fingerprints at three different infection stages, we can directly observe the differences in virus transport behavior caused by changes in the biological environment caused by different infection levels.

[0088] 3. Identifying Small Differences in SVIF Using Logistic Regression Models

[0089] To generate SVIF under certain conditions, a4 The time-consuming and laborious processing, calculation, and analysis process prompted people to explore machine learning techniques. Using machine learning techniques, by inputting the XY coordinate information of the trajectory and outputting the five features with the highest discrimination weights, SVIFs of different treatment groups can be quickly created, compared, and analyzed. In this study, a logistic regression classifier was trained using the SVIF containing the above 17 features. Subsequently, the classifier was used to predict the three infection stages of influenza virus with an accuracy of 93.1%, 84.4%, and 90.0%, respectively, and an F1 score of 87.1% (e.g. Fig.23 The projections based on the principal component analysis (PCA) and linear discriminant analysis (LDA) dimensionality reduction methods showed a clear distribution between groups, indicating that the model has the ability to effectively analyze two groups of virus trajectories with highly overlapping step length distributions, and highlighting the robust performance of the model in analyzing complex and similar groups (e.g. Fig.24 and Fig.25 shown).

[0090] In addition, we used machine learning to identify subtle changes in viral behavior after drug treatment and use this as a basis for determining the antiviral effect of the drug. We used two microtubule inhibitors, paclitaxel and nocodazole, to hinder the viral infection process. Figure 26-29 The results showed that machine learning could distinguish that the virus movement pattern changed after the two drugs took effect. The 17 descriptive features of the second-stage group and the inhibitor treatment group were sorted according to the discriminant weights. The results showed that the feature with the greatest impact on both treatment groups was the fractal dimension, indicating that the virus movement lost its directionality (such as Fig.30 and Fig.31 (shown) Fig.30 and Fig.31 The data are colored according to the variable with the highest value in the partition, and the numbers represent the normalized vector components in the LDA projection). The results of machine learning highlight its power in identifying the antiviral efficacy of drugs based on differences in SVIFs.

[0091] 4. Evaluating the efficacy of antiviral drugs through machine learning

[0092] We used SVIFs and machine learning to establish an antiviral drug screening platform to create a comprehensive evaluation system for evaluating the antiviral efficacy of various drugs. In this study, we tested 10 drugs, including microtubule inhibitors (such as paclitaxel, vincristine sulfate, and nocodazole), endosomal acidification inhibitors NH4Cl and bafilomycin A1, cholesterol-lowering drug methyl-β-cyclodextrin (MβCD), and four approved drugs with antiviral potential, diltiazem, amiloride, apigenin, and atorvastatin (these drugs were originally used for other diseases). Cells were treated with optimized drug concentrations, viral trajectories within 5-20 minutes of infection were extracted, and SVIFs under specific conditions were constructed.

[0093] The single virus infection fingerprint (SVIF) consists of 17 features. Features T0, T1, T2, T3, lifespan and mean step length (meanSL) show the changes in the speed of virus movement. Features such as kurtosis, Gaussianity, fractal dimension and efficiency describe the directionality and asymmetry of the virus transmission trajectory. At the same time, turbidity, average MSD, diffusion coefficient D, MSD curve power law fitting quality P value (Pval), power law scaling coefficient alpha and MSD ratio are used to describe the virus movement type and trajectory filling characteristics. Finally, feature N represents the duration of virus movement. A four-state HMM (T0, T1, T2 and T3) is used instead of a seven-state HMM to describe the virus transport process involving memoryless transitions between different states, which simplifies and ensures a clear distinction between the fast and slow states of virus transport while reducing computational complexity.

[0094] By at least 10 4 The z-value normalization of the matrix with 17 features can provide a preliminary visualization of the single virus infection fingerprint (SVIF) (e.g. Fig.32 The fingerprints of the control groups (stage 2, stage 1, and stage 0) are lightly colored and serve as a baseline for comparison with the other groups. The squares of the drug-treated groups are darker green and red, showing the effects of the drugs on viral infection from different angles.

[0095] By calculating the average Euclidean distance of SVIFs of different groups, the antiviral effect of the drug can be intuitively seen and the mechanism similarity between different drugs can be estimated (e.g. Fig.33 ). The intra-group and inter-group similarities were determined by the average Euclidean distance of the SVIFs of each group or paired group, respectively. The Euclidean distance matrix was calculated for each pair, and each circle represents the similarity between the viral movement patterns in the pair. In the similarity matrix, the larger the circle, the more obvious the trend from green to red, and the higher the similarity.

[0096] The evaluation of antiviral efficacy relies on the difference in SVIF between the treatment group and the control group. The average Euclidean distance of the SVIFs in the second phase of the drug-treated group and the DMSO-treated group was calculated and normalized by the minimum-maximum normalization as the antiviral score. The higher the score, the more obvious the inhibitory effect of the corresponding drug on viral transport. The effectiveness of the 10 molecules was judged based on the Euclidean distance of 0.21 between the groups in the second phase.

[0097] The results showed that amiloride, diltiazem and three microtubule inhibitors showed relatively strong therapeutic effects (e.g. Fig.34 shown).

[0098] To verify the machine learning results, we used immunofluorescence and qPCR techniques to detect the protein and RNA levels of progeny viruses after treatment with five effective drugs (such as Figure 35-38 As shown), the results are consistent with the machine learning results. Amiloride and diltiazem, which were originally approved as sodium channel inhibitors and calcium channel inhibitors for the treatment of diseases such as hypertension, showed the best antiviral effects. Vincristine and nocodazole showed the second highest efficacy by disrupting the microtubule polymerization required for rapid viral movement. Paclitaxel had limited antiviral effect. These findings are consistent with the results obtained using SVIF-based machine learning methods, further verifying the practicality of SVIF in antiviral drug screening and efficacy evaluation, and highlighting the potential of SVIF in assisting the development of multidimensional databases.

[0099] V. Conclusion

[0100] The research in this patent demonstrates the potential of SVIFs to reflect the perturbations experienced by the virus during the dynamic infection process. In addition, it also highlights the effectiveness of logistic regression models in analyzing complex and similar experimental groups. It also demonstrates the ability of SVIF combined with machine learning for antiviral drug screening and efficacy evaluation. The method of this patent can provide intuitive evidence of the drug's ability to resist viral infection in a very short analysis time. Therefore, it is expected to be widely used in responding to future new virus outbreaks.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A drug screening method based on single virus tracing and machine learning, characterized in that: The following steps are involved: S1: attaching biotinylated viruses to the host cell membrane and labeling them with streptavidin-modified quantum dots to obtain cell samples bound to the quantum dot-labeled viruses; S2: Analyze the image sequences of biotinylated virus-infected host cells to obtain several single virus infection tracks; S3: Input the coordinate information of several single virus infection trajectories into the machine learning model for calculation and analysis to obtain the single virus infection fingerprint under corresponding processing; S4: Identify and evaluate the antiviral effect of drugs based on the differences in single virus infection fingerprints between the drug-treated group and the control group.

2. A drug screening method based on single virus tracing and machine learning according to claim 1, characterized in that: The S1 specifically includes the following steps: The biotinylated virus was incubated with the host cells at 4°C for 5 minutes, and the unattached biotinylated virus was removed by washing. 3 nM streptavidin-modified quantum dots were added to label the attached biotinylated virus at 4°C for 5 minutes, and the unbound streptavidin-modified quantum dots were removed by washing.

3. The drug screening method based on single virus tracing and machine learning according to claim 1, characterized in that: The S2 is used to image the process of biotinylated virus infecting host cells in a live cell online culture system using a confocal microscope.

4. The drug screening method based on single virus tracing and machine learning according to claim 1, characterized in that: The 17 features in S3 are the proportion of hidden Markov state 0, the proportion of hidden Markov state 1, the proportion of hidden Markov state 2, the proportion of hidden Markov state 3, lifespan, average step length, kurtosis, Gaussianity, fractal dimension, efficiency, sleepiness, average MSD, diffusion coefficient D, MSD curve power law fitting quality P value, power law scaling coefficient alpha, MSD ratio, and number of frames N.

5. The drug screening method based on single virus tracing and machine learning according to claim 1, characterized in that: The specific steps of S4 are: calculating the average Euclidean distance between the drug treatment group and the control group, taking the Euclidean distance within the control group as a baseline, calculating the antiviral score, and thus determining the antiviral effect of the drug.

6. The drug screening method based on single virus tracing and machine learning according to claim 1, characterized in that: The method for preparing the biotinylated virus comprises: incubating and purifying the virus with biotin to obtain the biotinylated virus.

7. A drug screening method based on single virus tracing and machine learning according to claim 6, characterized in that: The incubation condition is 30 minutes at room temperature.

8. The drug screening method based on single virus tracing and machine learning according to claim 6, characterized in that: The specific steps of the purification are: using a gel filtration column and a filter to remove free biotin and virus aggregates.

9. Use of the method according to any one of claims 1 to 8 in drug screening.

10. Use of the method according to any one of claims 1 to 8 in drug repurposing.

Images