Zebra fish phenotype deep learning analysis method for assisting drug discovery

Abstract

The invention relates to the technical field of drug research and development, in particular to a zebrafish phenotype deep learning analysis method for assisting drug discovery, which comprises the following steps: (1) culturing zebrafish and exposing drugs; (2) collecting behavior phenotype data; (3) establishing a zebra fish phenotypic feature library; (4) training a drug feature recognition model; and (5) model evaluation and drug action characteristic analysis. According to the method, drugs with similar action mechanisms can be effectively identified, and a new auxiliary decision-making tool is provided for early drug screening. The method verifies the application feasibility of the zebrafish phenotype analysis method based on deep learning in the drug preliminary screening stage, and can be used as a beneficial supplement of a traditional drug screening method.

Description

Technical Field

[0001] The present invention relates to the technical field of drug development, and in particular to a zebrafish phenotype deep learning analysis method for assisting drug discovery. The method combines zebrafish behavioral analysis with deep learning technology to assist in the early screening of psychotropic drugs. Background Art

[0002] At present, drug development mainly adopts two screening strategies: target-based and phenotype-based. In the early screening stage, a large number of candidate compounds need to be evaluated, which is a time-consuming and costly process. Although target-based screening is the most commonly used method, the target of action needs to be determined in advance. Although phenotype-based screening can directly observe the efficacy of the drug, traditional methods mainly rely on manual observation and simple motion analysis, and have the following problems: 1. Data analysis efficiency is low, and it is difficult to cope with the initial screening of a large number of compounds; 2. Manual observation is highly subjective and the results may be unstable; 3. There is a lack of means to quickly identify compounds of potential value; 4. The screening cost is high, and more decision-making tools are needed. Summary of the invention

[0003] To solve the above technical problems, the present invention provides a deep learning analysis method for assisting drug discovery. This method combines high-throughput in vivo experiments and artificial intelligence technology to quickly analyze the behavioral characteristics of zebrafish and provide a decision-making reference for early drug screening.

[0004] The technical solution of the present invention is as follows:

[0005] A zebrafish phenotype deep learning analysis method for assisting drug discovery, comprising the following steps:

[0006] 1. Zebrafish culture and drug exposure:

[0007] -Use wild-type AB strain zebrafish and culture at 28°C;

[0008] - Drug exposure begins within 2 hours of fertilization;

[0009] -Set up two concentration gradients (1 μg / L and 10 μg / L) in E3 culture medium;

[0010] -Use 48-well plates for exposure experiments, place one embryo in each well, and set up 47 replicate wells.

[0011] 2. Behavioral phenotyping data collection:

[0012] -Behavioral recordings were performed during a specific developmental period (5-6 dpf);

[0013] -Use camera system to record movement trajectory;

[0014] -Use ZebraLab software to extract the cumulative behavior trajectory diagram of fixed length.

[0015] 3. Establishment of zebrafish phenotypic feature library:

[0016] -Use Python program to convert motion trajectory into trajectory map with uniform resolution;

[0017] -Construct multiple datasets with different duration conditions;

[0018] -Perform data cleaning and balanced sampling;

[0019] - Remove erroneous data and missing data samples.

[0020] 4. Drug feature recognition model training:

[0021] -Use convolutional neural network architectures, including DenseNet, EfficientNet, etc.;

[0022] -Optimize parameters such as the number of network layers and learning rate;

[0023] -Set up Batch normalization layer and fully connected layer.

[0024] 5. Model evaluation and drug action characteristics analysis:

[0025] -Model classification accuracy evaluation;

[0026] -UMAP semantic feature dimensionality reduction visualization.

[0027] Beneficial effects of the present invention:

[0028] The present invention has the following advantages:

[0029] 1. Provides a new auxiliary screening tool that can quickly predict the possible action categories of compounds;

[0030] 2. The classification accuracy of DenseNet161 reached 67.36%, which can be used as an effective reference for early screening;

[0031] 3. It can complement the advantages of traditional screening methods and provide a new perspective for data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 The best accuracy of different deep learning models after training for 25 epochs on 4 datasets;

[0034] Figure 2 Dataset 2min 6dpf 10μg / L UMAP semantic feature dimensionality reduction visualization analysis. DETAILED DESCRIPTION

[0035] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0036] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0037] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0038] Example

[0039] 1. Zebrafish Culture and Drug Exposure

[0040] 1.1 Zebrafish culture

[0041] Wild-type AB strain zebrafish were selected and cultured at 28°C with a 12h:12h light-dark cycle. Artemia were fed at 10:00 and 18:00 every day. Healthy sexually mature zebrafish were selected for spawning.

[0042] 1.2 Drug exposure

[0043] Eleven psychotropic drugs were selected, including 2-valproic acid, nortriptyline, duloxetine, droperidol, haloperidol, chlorpromazine, clozapine, memantine, aripiprazole, amisulpride and ziprasidone. Exposure began within 2 hours after fertilization, and two concentration gradients of 1 μg / L and 10 μg / L were set. A 48-well plate was used for the exposure experiment, with one embryo placed in each well and 47 replicate wells set.

[0044] 2. Behavioral Phenotyping Data Collection

[0045] When zebrafish were exposed to 5dpf and 6dpf, an animal behavior tracking system equipped with a high-definition camera was used to record continuous video of each group of samples for 2 hours. The collected 2-hour video was processed using ZebraLab software, and the video was segmented and extracted into multiple 2-minute segments through a preset program. Each 2-minute segment generated a cumulative motion trajectory map, thereby obtaining a large amount of behavioral characteristic data for subsequent analysis.

[0046] 3. Establishment of zebrafish phenotypic feature library

[0047] The obtained zebrafish behavioral trajectory images were converted into images of uniform resolution using a Python program. Four behavioral trajectory phenotypic feature libraries were constructed, including 2min 5dpf 1μg / L (28621 images), 2min 5dpf 10μg / L (29950 images), 2min 6dpf 1μg / L (30299 images) and 2min 6dpf 10μg / L (31075 images) (Table 1).

[0048] 4. Drug feature recognition model training

[0049] DenseNet and EfficientNet were selected as the main deep learning frameworks, and 25 epochs were trained on each dataset. The model performance was optimized by adjusting parameters such as learning rate and batch size.

[0050] 5. Model evaluation and drug action characteristics analysis

[0051] 5.1 Model Accuracy Evaluation

[0052] On the 2min 6dpf 10μg / L dataset, DenseNet161 achieved a maximum classification accuracy of 67.36% ( Figure 1 ).

[0053] 5.2 Analysis of drug action characteristics

[0054] In the UMAP dimensionality reduction visualization analysis, for the sake of easy observation, the colors of drugs for treating similar diseases are set to be consistent, with only different shapes. From the results, it can be observed that drugs with the same efficacy show clustering in the feature space ( Figure 2). This shows that this method can effectively identify drugs with similar mechanisms of action, providing a new auxiliary decision-making tool for early drug screening. This result verifies the feasibility of the zebrafish phenotyping method based on deep learning in the initial drug screening stage, which can serve as a beneficial supplement to traditional drug screening methods.

[0055] Table 1 Dataset name and number of images

[0056]

[0057] 6. Method validation and application analysis

[0058] 6.1 Verification of decision-making support capabilities

[0059] A classification accuracy rate of 67.36% was achieved in known drug tests, which can help researchers quickly screen out compounds with higher potential value and provide a reference for subsequent in-depth evaluation.

[0060] 6.2 Limitations of the Method

[0061] This method is mainly used to assist decision making and cannot completely replace traditional evaluation methods; the results are for reference only and need to be verified through subsequent experiments; it is recommended to be used in conjunction with other screening methods.

[0062] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0063] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A zebrafish phenotype deep learning analysis method for drug discovery, characterized in that: The following steps are involved: (1) Zebrafish culture and drug exposure include: selecting wild-type AB strain zebrafish, culturing at 28°C, exposing to drugs within 2 hours after fertilization, selecting a variety of psychotropic drugs, and setting two concentration gradients of 1 μg / L and 10 μg / L; (2) Behavioral phenotypic data collection includes: using an animal behavior tracking system equipped with a camera to record the behavior of 5-6 dpf zebrafish, and using ZebraLab software to extract cumulative behavioral trajectory maps of fixed duration; (3) The establishment of zebrafish phenotypic feature library includes: converting behavioral trajectory graphs into uniform resolution images through Python program, performing data cleaning and balanced sampling; (4) Drug feature recognition model training: using DenseNet network architecture; (5) Model evaluation and analysis of drug action characteristics: including calculation of classification accuracy and feature dimensionality reduction using UMAP.

2. A zebrafish phenotype deep learning analysis method for assisting drug discovery according to claim 1, characterized in that: The step (3) requires constructing multiple data sets with different duration conditions, while deleting erroneous data and missing data samples.

3. A zebrafish phenotype deep learning analysis method for assisting drug discovery according to claim 1, characterized in that: The step (4) needs to optimize the parameters such as the number of network layers and learning rate, and set the batch normalization layer and the fully connected layer at the same time.

Images