A method for recognizing a thalidomide r-form chiral molecule

By combining ECD spectroscopy with CNN convolutional neural network to construct a recognition model, the R-type chiral molecule of thalidomide can be automatically identified. This solves the problems of high cost, low throughput and human error in existing technologies, and achieves efficient, low-cost and environmentally friendly recognition results.

CN122337374APending Publication Date: 2026-07-03HAINAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610314072.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-16
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

Existing technologies for identifying thalidomide R-type chiral molecules suffer from high cost, low throughput, complex operation, and susceptibility to human error.

Method used

An identification model was constructed by combining ECD spectroscopy with a CNN convolutional neural network. By building a dataset and training the identification model, the thalidomide R-type chiral molecule was automatically identified. The CNN automatically captured the local Cotton effect, avoiding errors caused by manual judgment.

Benefits of technology

This method achieves high-throughput, high-speed, low-cost, and environmentally friendly recognition of thalidomide R-type chiral molecules, avoiding the complex operation and human error of traditional methods, and improving recognition efficiency and accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122337374A_ABST
    Figure CN122337374A_ABST
Patent Text Reader

Abstract

This invention discloses a method for identifying thalidomide R-type chiral molecules. The method includes the following steps: constructing a dataset, dividing the dataset into a training set and a test set according to a preset ratio; the dataset includes ECD spectra of thalidomide R-type chiral molecules, ECD spectra of diazepam molecules, ECD spectra of phenobarbital molecules, and corresponding label data; constructing a recognition model based on a CNN convolutional neural network, inputting the training set and test set into the recognition model for training and performance evaluation to obtain a trained recognition model; inputting the ECD spectrum of the thalidomide chiral molecule to be detected into the trained recognition model to obtain a recognition result, wherein the recognition result is thalidomide R-type chiral molecules or other molecules. This invention enables simple and rapid identification of thalidomide R-type chiral molecules.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent recognition technology, and in particular to a method for recognizing thalidomide R-type chiral molecules. Background Technology

[0002] Thalidomide, also known as thalidomide, is a centrally acting depressant discovered during the development of antibacterial drugs. It was once widely used in Europe and Japan as an antiemetic for morning sickness, but shortly after its introduction, data showed increased miscarriage and phocomelia rates among pregnant women using the drug, leading to its withdrawal from the market – an event known as the thalidomide incident. Later research revealed that the drug has two enantiomers: the R-type chiral molecule of thalidomide is safe, while the S-type chiral molecule has teratogenic effects. Electron dichroism (ECD) spectroscopy, first observed in 1847 by Austrian mineralogist Wilhelm Karl von Haidinger in amethyst crystals as the "difference in absorption of circularly polarized light," or circular dichroism, has been applied to various scientific research fields with continuous advancements in science and technology. ECD spectroscopy has progressed through five stages: "phenomenon → instrumentation → quantum theory → calculation → engineering," and is now one of the core tools for identifying the absolute configuration of chiral drugs, natural products, and functional materials.

[0003] Currently, domestic and international techniques for distinguishing thalidomide chiral molecules include high-performance liquid chromatography (HPLC) and gas chromatography (GC). HPLC involves: 1) Sample pretreatment: precipitating proteins or directly dissolving them and passing them through a 0.22µm membrane; 2) Chiral separation: using a CHIRALPAK AD-RH column, acetonitrile-methanol-citric acid buffer gradient, flow rate 0.5 mL·min⁻¹, 25°C; 3) Detection and characterization: UV 220nm; the R-type chiral molecule is approximately 17.6 min (S-type 13.5 min), and a retention time deviation of <2% is considered an R-type chiral molecule of thalidomide. However, this HPLC method suffers from high cost, short lifespan of chiral columns, and high mobile phase consumption. Single runs are limited to 10–20 min, resulting in low throughput. Strict temperature and pH control is required, making it a high-barrier-to-operation method. Gas chromatography (GC): 1) Derivatization: Thalidomide is converted into a volatile derivative using TFA reagent (trifluoroacetic acid); 2) Injection and separation: The derivative is dissolved in n-hexane and injected in 1µL split; the chiral capillary column is programmed to rise from 100 to 180°C, with He as the carrier gas at 1.2 mL / min⁻¹. R / S derivatives elute sequentially due to differences in inclusion cohesion; 3) Detection: ECD / MS recording; the R-type peak elutes 0.3–0.5 min later than the S-type peak, which can be confirmed by referring to a standard. However, this GC method requires derivatization first, involves multiple steps, and is prone to degradation. Chiral columns are expensive and have slow heating times (15–30 min per cycle), making continuous high-throughput difficult. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a method for identifying thalidomide R-type chiral molecules. This method avoids the traditional manual identification and extraction of molecular signals and can quickly identify large batches of samples in practical applications. This is because thalidomide R-type chiral molecules exhibit strong positive and negative Cotton effects at specific wavelengths, which are their unique characteristic signals. This method is very suitable for the simple and rapid identification of thalidomide R-type chiral molecules.

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

[0006] A method for recognizing thalidomide R-type chiral molecules includes the following steps:

[0007] A dataset is constructed, which is divided into a training set and a test set according to a preset ratio. The dataset includes the ECD spectra of thalidomide R-type chiral molecules, the ECD spectra of diazepam molecules, the ECD spectra of phenobarbital molecules, and the corresponding label data.

[0008] Construct a recognition model based on a CNN convolutional neural network, input the training set and test set into the recognition model for training and performance evaluation, and obtain a well-trained recognition model;

[0009] The ECD spectrum of the thalidomide chiral molecule to be detected is input into the trained recognition model to obtain the recognition result, which is thalidomide R-type chiral molecule or other molecules.

[0010] Preferably, constructing the dataset includes the following steps:

[0011] ECD spectral data of thalidomide R-type chiral molecule, diazepam molecule, and phenobarbital molecule were obtained and labeled accordingly;

[0012] The acquisition method includes: using Spartan software to search for the initial conformation of the molecule, using GaussView and Gaussian software to optimize the configuration, correct the dispersion and calculate the excited state of the initial conformation, and using SpecDis software to perform spectral fitting to obtain ECD spectral data.

[0013] Preferably, the method further includes the following steps: converting the ECD spectral data of thalidomide R-type chiral molecules, diazepam molecules, and phenobarbital molecules into one-dimensional group numbers using UltraEdit software and Excel software, respectively.

[0014] Preferably, the recognition model includes an input layer, a first convolutional block, a second convolutional block, a third convolutional block, a global average pooling layer, a fully connected layer, and an output layer connected in sequence.

[0015] Preferably, the first convolutional block includes a convolutional layer with a kernel size of 3*1 and a kernel count of 32, a batch normalization layer, a ReLU activation layer, and a max pooling layer connected in sequence; the second convolutional block includes a convolutional layer with a kernel size of 3*1 and a kernel count of 64, a batch normalization layer, a ReLU activation layer, and a max pooling layer connected in sequence; and the third convolutional block includes a convolutional layer with a kernel size of 3*1 and a kernel count of 128, a batch normalization layer, and a ReLU activation layer connected in sequence.

[0016] Based on the above technical solution, the beneficial effects of this invention are as follows: This invention discloses a method for identifying thalidomide R-type chiral molecules. It uses ECD spectroscopy combined with a CNN convolutional neural network to identify thalidomide R-type chiral molecules. Compared with traditional HPLC (liquid chromatography) and GC (gas chromatography) techniques, it has advantages such as high throughput, high speed, low cost, low environmental impact, and no need for manual peak selection. The CNN automatically captures the local Cotton effect, avoiding subjective errors caused by "human eye comparison." Traditional ECD (circular electronic dichroism spectroscopy) relies on semi-empirical rules or TDDFT (time-dependent density functional theory) comparison, requiring professional experience. The method of identifying thalidomide R-type chiral molecules using ECD spectroscopy combined with a CNN convolutional neural network is superior in speed, cost, environmental friendliness, and automation. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart of a method for recognizing thalidomide R-type chiral molecules in one embodiment;

[0018] Figure 2 This is a schematic diagram of the process for acquiring spectral data in a method for identifying thalidomide R-type chiral molecules in one embodiment;

[0019] Figure 3 This is a schematic diagram of the structure of the recognition model in a method for recognizing thalidomide R-type chiral molecules in one embodiment;

[0020] Figure 4 This is a schematic diagram of the training loss curve of the recognition model in one embodiment;

[0021] Figure 5 This is a schematic diagram of the accuracy curve of the recognition model training in one embodiment. Detailed Implementation

[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0023] like Figure 1 As shown, this embodiment provides a method for recognizing thalidomide R-type chiral molecules, including the following steps:

[0024] A dataset is constructed, which is divided into a training set and a test set according to a preset ratio. The dataset includes the ECD spectra of thalidomide R-type chiral molecules, the ECD spectra of diazepam molecules, the ECD spectra of phenobarbital molecules, and the corresponding label data.

[0025] Construct a recognition model based on a CNN convolutional neural network, input the training set and test set into the recognition model for training and performance evaluation, and obtain a well-trained recognition model;

[0026] The ECD spectrum of the thalidomide chiral molecule to be detected is input into the trained recognition model to obtain the recognition result, which is thalidomide R-type chiral molecule or other molecules.

[0027] In this embodiment, ECD spectral data is acquired, including the ECD spectra of thalidomide chiral molecules (R-type chiral molecules) and the ECD spectra of other molecules (Diazepam, phenobarbital, etc.). The acquisition process is as follows: Figure 2 As shown in the flowchart: 1) Molecular construction: Download the thalidomide R-type chiral molecule model from https: / / molview.org / . 2) Initial conformation search: Search for the initial conformation of the molecule using Spartan '14 V1.1.4 software, specifically utilizing semi-empirical method PM6 (parametric method 6 in semi-empirical methods, available in the simulation software, full name: ParametricMethod 6). The stable conformation number of the thalidomide R-type chiral molecule is two. 3) Preliminary and High-Precision Configuration Optimization: The obtained conformation numbers were optimized and calculated using GaussView 6.0 and Gaussian 16W software. Preliminary optimization was performed using the keyword "opt" and the basis set B3LYP / 6-31G(d) (a partial option in Gaussian software). The results were then optimized for high precision using density functional theory B3LYP / 6-311G(d,p) (a partial option in Gaussian software), with dispersion correction performed using em=gd3bj (a partial option in Gaussian software). 4) Excited State Calculation: The obtained results were used for excited state calculation using td=(nstates=50) (a partial option in Gaussian software). 5) Fitting Boltzmann Weighted Spectra: ECD (Electron Circular Dichroism) spectra were plotted using GaussView 6.0 software, and Boltzmann weighting was applied. The high-precision optimized data is then fed into SpecDis software to extract heat, followed by Boltzmann-weighted spectra, and finally spectral fitting. The spectrum is set to a wavelength range of 130nm to 360nm, and the output file is generated.

[0028] Recognition models built based on CNN convolutional neural networks: such as Figure 3 The process is as follows: 1) Convert the ECD spectral data output file (.xy) into a one-dimensional array using UltraEdit and Excel software. 2) Load and process the ECD spectral data, dividing it into two folders: "test" and "train". 3) Input layer: Input a 1×N vector, converted to a 4-D tensor acceptable to Matlab software. 4) Conv Block 1 (first convolutional block), containing convolutional layers, batch normalization layers, ReLU activation layers, and max pooling layers. Convolutional layers: 1-D convolution, simulated using a 2-D kernel with a height of 3 and a width of 1 using a sliding window, with a kernel size of 3*1 and 32 kernels. Batch normalization layer: accelerates convergence, mitigates internal covariate shifts, and provides some regularization effect. ReLU activation layer: non-linear activation. Max pooling layer: pooling window size of 2*1, with 2 non-overlapping downsampling points along the "height" direction, "absorbing" small translations and improving deformation robustness. 5) Conv Block 2 (Second Convolutional Block): Consists of a convolutional layer with 64 kernels (3x1 kernel size), a batch normalization layer, a ReLU activation layer, and a max pooling layer, connected sequentially. 6) Conv Block 3 (Third Convolutional Block): Consists of a convolutional layer with 128 kernels (3x1 kernel size), a batch normalization layer, and a ReLU activation layer, connected sequentially. 7) Global Average Pooling Layer: Reduces the dimensionality of the data, significantly reducing the number of parameters and preventing overfitting. 8) Fully Connected Layer: Compresses high-dimensional convolutional features, integrates the features, and performs classification. 9) Output Layer: Outputs the classification results. A confusion matrix is ​​used to evaluate the model performance. Figure 4 , 5 The diagrams show the model training loss curve and the model training accuracy curve of the recognition model built on the CNN convolutional neural network.

[0029] It should be understood that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart above may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0030] The above are merely preferred embodiments of the present application and are not intended to limit the embodiments of the present application. For those skilled in the art, the embodiments of the present application can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of the present application should be included within the protection scope of the embodiments of the present application.

Claims

1. A method for the recognition of a chiral molecule of the R form of thalidomide, characterized in that, Includes the following steps: A dataset is constructed, which is divided into a training set and a test set according to a preset ratio. The dataset includes the ECD spectra of thalidomide R-type chiral molecules, the ECD spectra of diazepam molecules, the ECD spectra of phenobarbital molecules, and the corresponding label data. Construct a recognition model based on a CNN convolutional neural network, input the training set and test set into the recognition model for training and performance evaluation, and obtain a well-trained recognition model; The ECD spectrum of the thalidomide chiral molecule to be detected is input into the trained recognition model to obtain the recognition result, which is thalidomide R-type chiral molecule or other molecules.

2. The method for recognizing thalidomide R-type chiral molecules according to claim 1, characterized in that, The construction of the dataset includes the following steps: ECD spectral data of thalidomide R-type chiral molecule, diazepam molecule, and phenobarbital molecule were obtained and labeled accordingly; The acquisition method includes: using Spartan software to search for the initial conformation of the molecule, using GaussView and Gaussian software to optimize the configuration, correct the dispersion and calculate the excited state of the initial conformation, and using SpecDis software to perform spectral fitting to obtain ECD spectral data.

3. The method for recognizing thalidomide R-type chiral molecules according to claim 2, characterized in that, The process also includes the following steps: converting the ECD spectral data of thalidomide R-type chiral molecules, diazepam molecules, and phenobarbital molecules into one-dimensional group numbers using UltraEdit and Excel software, respectively.

4. The method for recognizing thalidomide R-type chiral molecules according to claim 1, characterized in that, The recognition model includes an input layer, a first convolutional block, a second convolutional block, a third convolutional block, a global average pooling layer, a fully connected layer, and an output layer connected in sequence.

5. The method for recognizing thalidomide R-type chiral molecules according to claim 4, characterized in that, The first convolutional block includes a convolutional layer with a kernel size of 3*1 and a kernel count of 32, a batch normalization layer, a ReLU activation layer, and a max pooling layer connected in sequence; the second convolutional block includes a convolutional layer with a kernel size of 3*1 and a kernel count of 64, a batch normalization layer, a ReLU activation layer, and a max pooling layer connected in sequence; the third convolutional block includes a convolutional layer with a kernel size of 3*1 and a kernel count of 128, a batch normalization layer, and a ReLU activation layer connected in sequence.