Antibacterial applications of dasupin
By integrating machine learning models and databases, the antibacterial activity and structural novelty of compounds were discovered, solving the problem of developing novel antibacterial drugs. Dalsupine, which has low toxicity and multiple inhibitory activities, was discovered and is suitable for preparing drugs against drug-resistant bacteria.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- KUNMING UNIV OF SCI & TECH
- Filing Date
- 2021-11-03
- Publication Date
- 2026-07-21
AI Technical Summary
Existing technologies lack effective methods for discovering and validating novel antibacterial active compounds, and the development of antibacterial drugs is progressing slowly, especially in China, resulting in a shortage of antibacterial drugs and making it difficult to cope with the bacterial resistance crisis.
Machine learning methods were employed to construct a prediction model for antibacterial activity by building a support vector machine and a random forest prediction model, combined with Pybel and PyDPI to generate molecular descriptors. Data was collected from the ChEMBL and DrugBank databases to predict the antibacterial activity and evaluate the structural novelty of the compounds.
Accurate prediction of the antibacterial activity of compounds was achieved, and novel compounds were screened. Dasupine showed inhibitory activity against a variety of bacteria and fungi, had low toxicity, and was suitable for the preparation of drugs against drug-resistant bacteria, with an accuracy rate of over 91%.
Smart Images
Figure CN115359854B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of pharmaceutical technology, and in particular to the antibacterial application of daspergillin and methods for predicting its antibacterial activity and evaluating its structural novelty. Background Technology
[0002] Antibiotics are the cornerstone of modern medicine; their discovery and use have saved countless lives. Alexander Fleming's discovery of penicillin in 1929 and its use in treating various infectious diseases marked the beginning of the antibiotic era. Between 1960 and 1970, scientists developed numerous antibacterial drugs and widely applied them in medicine.
[0003] Controlling the spread of antibiotic resistance and developing novel antibacterial drugs has become a societal consensus. However, developing novel antibacterial drugs faces multiple challenges. First, a vast number of antibiotics have already been identified, making it difficult to find new antibacterial drugs. A team of 20 scientists at GlaxoSmithKline spent five years screening 500,000 compounds using whole-cell screening experiments, obtaining only three lead compounds. Compared to the United States, my country's progress in antibacterial drug development is slow, with only one-seventh the number of antibacterial drugs being developed, and companies in China facing a severe shortage of antibacterial drugs in their pipelines. Discovering novel antibacterial active compounds from hundreds of millions of compounds, including those with vast chemical structures, is crucial for antibacterial drug development. However, currently, there are no readily available methods or frameworks to achieve this goal, making it a challenging task.
[0004] Disopyramide is an antiarrhythmic drug used to treat premature atrial contractions, paroxysmal atrial tachycardia, atrial fibrillation, and premature ventricular contractions. Currently, there are no reports of antibacterial activity from disopyramide.
[0005] Secondly, there is currently no existing technology that can discover novel antibacterial active compounds and experimentally verify these compounds. Therefore, we propose a method for predicting antibacterial activity and evaluating structural novelty, which systematically discovers and evaluates the antibacterial potential of all marketed drugs. Summary of the Invention
[0006] (a) Purpose of the invention
[0007] In view of this, the purpose of this invention is to propose a method for the antibacterial application of dasuprine and the prediction of its antibacterial activity and the evaluation of its structural novelty, so as to predict the antibacterial activity of the compound and evaluate its structural novelty.
[0008] (II) Technical Solution
[0009] To achieve the above-mentioned technical objectives, this invention provides a method for the antibacterial application of dapoxetine, as well as a method for predicting its antibacterial activity and evaluating its structural novelty:
[0010] It includes predicting the antibacterial activity of compounds and evaluating the structural novelty of compounds. The prediction of the antibacterial activity of compounds specifically includes the following steps.
[0011] Step 1: Collect high-throughput data related to antibacterial activity and cytotoxicity, and organize, filter and analyze this data to form an antibacterial activity benchmark database;
[0012] Step 2: Use Pybel and PyDPI to generate the composition descriptor, topological descriptor, molecular connectivity, molecular charge descriptor, and Daylight molecular fingerprint features of the compound to be predicted.
[0013] Step 3: Select the corresponding test features from the features generated in Step 2 using the feature selection module in scikit-learn;
[0014] Step 4: Construct the Support Vector Machine prediction model and the Support Random Forest prediction model;
[0015] Step 5: Use 10 rounds of 5-fold cross-validation to evaluate the classification performance of all models built in Step 4.
[0016] Step 6: After evaluation in step 5, the support vector machine and random forest prediction models with good prediction performance are combined to form an antibacterial prediction model;
[0017] Step 7: Predict the target features in Step 3 using an antibacterial prediction model. If both models predict that the compound has antibacterial activity, then the compound is selected as a candidate antibacterial compound.
[0018] Preferably, the data collected in step 1 specifically includes a benchmark dataset of antibacterial activity and a dataset of all marketed small molecule drugs;
[0019] The benchmark dataset for antibacterial activity is all the antibacterial activity data downloaded from the ChEMBL database;
[0020] The dataset of all marketed small molecule drugs is a collection of all marketed small molecule drugs and their corresponding information downloaded from the DrugBank drug database.
[0021] Preferably, the evaluation in step 5 uses five metrics: ROC curve, accuracy, precision, recall, and F1 score. The calculation formulas are as follows:
[0022]
[0023]
[0024]
[0025]
[0026] Among them, TP is a true positive, TN is a true negative, FP is a false positive, and FN is a false negative.
[0027] Preferably, in step 4, the support vector machine prediction model is constructed by using libsvm27, which is encapsulated in the Python-based machine learning module library Scikit-learn.
[0028] Preferably, in step 4, constructing a model that supports random forest prediction specifically involves using the random forest classifier in the Python-based machine learning library Scikit-learn to train and predict samples, thereby constructing a model that supports random forest prediction.
[0029] Preferably, the evaluation of the structural novelty of the compound specifically includes the following steps:
[0030] Step 1: Calculate the overall structural similarity between the candidate compound and all known antibacterial drugs using Pybel, and measure it using the Tanimoto coefficient (TC). The formula for calculating the TC value is:
[0031] TC = C(i, j) / U(i, j), where C(i, j) represents the number of common features in the molecular fingerprints of two small molecules i and j, and U(i, j) represents the total number of features in the molecular fingerprints of two small molecules i and j;
[0032] Step 2: Use Pybel to generate FP2 molecular fingerprints and calculate TC values;
[0033] Step 3: Determine whether the calculated TC value is lower than 0.5. If it is lower than 0.5, the two small molecules have very low similarity, and the selected compound has novel structure.
[0034] Preferably, the evaluation of the structural novelty of the compound further includes the following steps:
[0035] Step 1: Construct a substructure library of the active groups of all known marketed antibacterial drugs, and then use fmcsR to search for the substructures of newly discovered candidate antibacterial compounds.
[0036] Step 2: If the candidate compound does not contain the active substructure of a known antibacterial drug and the overall similarity is less than 0.5, then the compound has structural novelty.
[0037] In addition, we provide an application of dasuprine, namely, using dasuprine in the preparation of antibacterial drugs.
[0038] As can be seen from the above technical solutions, this application has the following beneficial effects:
[0039] This invention develops a novel method for accurately predicting the antibacterial activity of compounds and evaluating their structural novelty through machine learning and integrated antibacterial data, enabling the discovery of novel antibacterial compounds. This method achieves an accuracy exceeding 91% and can rapidly screen compounds in libraries containing hundreds of millions of compounds. Screening against the DrugBank database can re-identify novel antibacterial drugs from all marketed medications. Since these are marketed drugs, they have typically undergone safety testing and exhibit low toxicity.
[0040] This invention, through prediction and experimental verification, reveals that disopyramide, in addition to its inhibitory activity against various bacteria and fungi, safety profile, and low toxicity, also possesses the following characteristics that distinguish it from existing antibacterial drugs. Disopyramide differs chemically from existing marketed antibacterial drugs; it does not contain the active substructures of known antibacterial drugs and has an overall similarity of less than 0.2. Disopyramide was originally classified as an antiarrhythmic drug and can be used for atrial premature beats, paroxysmal atrial tachycardia, atrial fibrillation, and ventricular premature beats. Currently, there are no reports of disopyramide exhibiting antibacterial activity. It holds promise for applications in the preparation of drugs against drug-resistant bacteria.
[0041] In summary, this invention not only provides important ideas and guidance for the development of novel antibacterial drugs, but more importantly, it provides a novel, low-toxicity antibacterial compound, Disopyramide, to address the growing crisis of bacterial resistance. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0043] Figure 1This is a flowchart illustrating a method for predicting the antibacterial activity of compounds and evaluating their structural novelty, provided by the present invention.
[0044] Figure 2 The chemical structure diagram of Disopyramide provided by this invention.
[0045] Figure 3 The graph shows the performance of the support vector machine model and random forest model provided by this invention in predicting antibacterial activity.
[0046] Figure 4 The diagram shows the structural similarity calculation and activity prediction results of Disopyramide provided by this invention.
[0047] Figure 5 The diagram shows the novelty evaluation results of the Disopyramide structure provided by this invention.
[0048] Figure 6 The experimental data on the antibacterial activity of Disopyramide provided by this invention. Detailed Implementation
[0049] The following description is exemplary in nature and is not intended to limit the scope, application, or use of this disclosure. It should be understood that in all these figures, the same or similar reference numerals indicate the same or similar parts and features. The figures are merely schematic representations of the concept and principles of embodiments of this disclosure and do not necessarily show the specific dimensions and scale of the various embodiments of this disclosure. Certain details or structures of embodiments of this disclosure may be exaggerated in particular portions of certain figures.
[0050] Reference Figure 1-6 :
[0051] Example 1
[0052] A method for predicting the antibacterial activity of a compound and evaluating its structural novelty includes predicting the antibacterial activity of the compound and evaluating its structural novelty. The method for predicting the antibacterial activity of the compound specifically includes the following steps.
[0053] Step 1: Collect high-throughput data related to antibacterial activity and cytotoxicity, and organize, filter and analyze this data to form an antibacterial activity benchmark database;
[0054] Step 2: Use Pybel and PyDPI to generate the composition descriptor, topological descriptor, molecular connectivity, molecular charge descriptor, and Daylight molecular fingerprint features of the compound to be predicted.
[0055] Step 3: Select the corresponding test features from the features generated in Step 2 using the feature selection module in scikit-learn;
[0056] Step 4: Construct the Support Vector Machine prediction model and the Support Random Forest prediction model;
[0057] Step 5: Use 10 rounds of 5-fold cross-validation to evaluate the classification performance of all models built in Step 4.
[0058] Step 6: After evaluation in Step 5, the support vector machine and random forest prediction models with good prediction performance are combined to form an antibacterial prediction model;
[0059] Step 7: Predict the target features in Step 3 using an antibacterial prediction model. If both models predict that the compound has antibacterial activity, then the compound is selected as a candidate antibacterial compound.
[0060] In addition, the data collected in step 1 specifically includes a benchmark dataset of antibacterial activity and a dataset of all marketed small molecule drugs.
[0061] The baseline dataset for antibacterial activity consisted of all antibacterial activity data downloaded from the ChEMBL database. Specifically, active and inactive compounds were filtered based on their half-inhibition concentration (IC50). Compounds with an IC50 less than 1000 nM (10 μM) were classified as antibacterial active, while those with an IC50 greater than 10000 nM (100 μM) were classified as inactive. This baseline dataset contained a total of 1097 active antibacterial compounds and 578 inactive compounds. Since the negative data samples were fewer than the positive data samples in this dataset, we used random sampling from the positive dataset to obtain a balanced dataset with the same number of negative data samples. This process was repeated multiple times to ensure that the prediction model did not deviate significantly between each repetition.
[0062] The dataset of all marketed small molecule drugs was downloaded from the DrugBank drug database, containing all marketed small molecule drugs and their corresponding information. It includes a total of 4196 marketed small molecule drugs, of which 427 are antibacterial drugs.
[0063] Specifically, the evaluation in step 5 uses five metrics: ROC curve, accuracy, precision, recall, and F1 score. The calculation formulas are as follows:
[0064]
[0065]
[0066]
[0067]
[0068] Among them, TP is a true positive, TN is a true negative, FP is a false positive, and FN is a false negative.
[0069] It should be noted that in step 4, the support vector machine prediction model is constructed using libsvm27, which is encapsulated in the Python-based machine learning module library Scikit-learn.
[0070] Specifically, the area under the ROC curve (AUC) was used to select the optimal model and parameters. The kernel function "rbf" was ultimately determined to be the optimal model, with a penalty parameter C of 50, and other parameters set to default.
[0071] Building a model that supports random forest prediction involves using the random forest classifier from the Python-based machine learning library Scikit-learn to train and predict samples, thereby constructing a model that supports random forest prediction.
[0072] Specifically, the parameter settings are as follows: (1) The number of decision trees is 950, which is selected using the area under the ROC curve (AUC); (2) Other parameters are set to default.
[0073] Example 2
[0074] A method for predicting the antibacterial activity of a compound and evaluating its structural novelty, based on Example 1, specifically includes the following steps for evaluating the structural novelty of the compound:
[0075] Step 1: Calculate the overall structural similarity between the candidate compound and all known antibacterial drugs using Pybel, and measure it using the Tanimoto coefficient (TC). The formula for calculating the TC value is:
[0076] TC = C(i, j) / U(i, j), where C(i, j) represents the number of common features in the molecular fingerprints of two small molecules i and j, and U(i, j) represents the total number of features in the molecular fingerprints of two small molecules i and j;
[0077] Step 2: Use Pybel to generate FP2 molecular fingerprints and calculate TC values;
[0078] Step 3: Determine if the calculated TC value is lower than 0.5. If it is lower than 0.5, the two small molecules have very low similarity, and the selected compound has a novel structure.
[0079] In addition, as a preferred method, the evaluation of the structural novelty of the compound specifically includes the following steps:
[0080] Step 1: Construct a substructure library of the active groups of all known marketed antibacterial drugs, and then use fmcsR to search for the substructures of newly discovered candidate antibacterial compounds.
[0081] Step 2: If the candidate compound does not contain the active substructure of a known antibacterial drug and the overall similarity is less than 0.5, then the compound has structural novelty.
[0082] Specifically, the substructure library includes the substructures of the active groups of all known marketed antibacterial drugs, such as sulfonamides, penicillins, cephalosporins, carbapenems, chloramphenicol, tetracyclines, aminoglycosides, macrolides, glycopeptides, quinolones, linezolids, and lipoglycopeptides.
[0083] Using the above methods to predict and evaluate daboyramide, calculations showed that the maximum similarity between daboyramide and 427 existing antibacterial drugs was 0.44, while the overall average similarity was only 0.17. Figure 4 Meanwhile, Disopyramide does not contain the active substructure of the 10 common classes of antibacterial drugs. Figure 5 Overlap less than 1 indicates that substructures are not included.
[0084] Example 3
[0085] We used the method described above for predicting the antibacterial activity of compounds and evaluating their structural novelty to predict and evaluate dasuprine. The results showed that dasuprine possesses both antibacterial activity and structural novelty. Below, we will verify its antibacterial activity through experiments. The specific steps are as follows:
[0086] Step 1: Take 20 μl each of *Escherichia coli*, *Candida albicans*, *Bacillus subtilis*, and *Staphylococcus aureus* and add them separately to LB liquid medium. Incubate at 37°C with shaking until the culture becomes cloudy (OD600 approximately 0.6). Centrifuge the cloudy bacterial culture (5000 rpm, 5 min), reserving 100 μl of liquid medium. Mix the bacterial cells thoroughly by pipetting and spread evenly on the surface of solid LB medium. Place a sterilized 5 mm diameter filter paper on the surface of the medium and add 10 μl of Disopyramide sample to the center of the filter paper. After the sample is completely absorbed, invert the container and incubate at 37°C for 10-12 hours. Observe the formation of inhibition zones; the presence of inhibition zones indicates antibacterial activity.
[0087] Step 2: Following the same procedure as above, using ampicillin and fluconazole as positive controls, add 4 μl of the positive control solution to the center of a 5 mm diameter filter paper disc.
[0088] Step 3: The antibacterial activity value of the sample to be tested (relative to the positive control) is calculated according to the following formula:
[0089] The activity value (%) of the sample is calculated as follows: (diameter of the inhibition zone of the test sample / diameter of the inhibition zone of the positive control) * 100
[0090] Step 4, the antibacterial activity of disopyramide was determined as follows: Figure 6 . Figure 6 The inhibitory activities of disopyramide against various bacteria and fungi were listed, showing that disopyramide has significant inhibitory effects on Escherichia coli, Staphylococcus aureus, Bacillus subtilis, and Candida albicans. The inhibitory activity of disopyramide against Escherichia coli is 87% that of ampicillin, and its inhibitory activity against Candida albicans is 113.8% that of fluconazole.
[0091] The above experimental results indicate that the chemical structure of disopyramide differs from that of marketed antibacterial drugs and that it possesses inhibitory activity against a variety of bacteria. Therefore, disopyramide is a novel antibacterial compound that can be used in the preparation of antibacterial drugs.
[0092] Therefore, we provide an application of dasuprine, namely, using dasuprine in the preparation of antibacterial drugs.
[0093] This invention develops a novel method for accurately predicting the antibacterial activity of compounds and evaluating their structural novelty through machine learning and integrated antibacterial data, enabling the discovery of novel antibacterial compounds. This method achieves an accuracy exceeding 91% and can rapidly screen compounds in libraries containing hundreds of millions of compounds. Screening against the DrugBank database can re-identify novel antibacterial drugs from all marketed medications. Since these are marketed drugs, they have typically undergone safety testing and exhibit low toxicity.
[0094] This invention, through prediction and experimental verification, reveals that disopyramide, in addition to its inhibitory activity against various bacteria and fungi, safety profile, and low toxicity, also possesses the following characteristics that distinguish it from existing antibacterial drugs. Disopyramide differs chemically from existing marketed antibacterial drugs; it does not contain the active substructures of known antibacterial drugs and has an overall similarity of less than 0.2. Disopyramide was originally classified as an antiarrhythmic drug and can be used for atrial premature beats, paroxysmal atrial tachycardia, atrial fibrillation, and ventricular premature beats. Currently, there are no reports of disopyramide exhibiting antibacterial activity. It holds promise for applications in the preparation of drugs against drug-resistant bacteria.
[0095] In summary, this invention not only provides important ideas and guidance for the development of novel antibacterial drugs, but more importantly, it provides a novel, low-toxicity antibacterial compound, Disopyramide, to address the growing crisis of bacterial resistance.
[0096] The exemplary implementation of the solution proposed in this disclosure has been described in detail above with reference to preferred embodiments. However, those skilled in the art will understand that various modifications and alterations can be made to the above specific embodiments without departing from the spirit of this disclosure, and various combinations can be made to the various technical features and structures proposed in this disclosure without exceeding the protection scope of this disclosure, which is determined by the appended claims.
Claims
1. An application of dalsupine, characterized in that, The dasuprine was used to prepare an anti-Candida albicans drug.