Theophylline aptamer as well as virtual screening method and application thereof

Abstract

The invention belongs to the technical field of molecular biology, and particularly discloses a theophylline aptamer and a virtual screening method and application thereof. The theophylline aptamer is selected from one or more of MUT9, MUT16, MUT67, MUT78 or MUT89. The invention further discloses a preparation method of the theophylline aptamer. The invention discloses a theophylline aptamer as well as a virtual screening method and application thereof, the theophylline aptamer is high in affinity and specificity, the virtual screening method remarkably improves the accuracy and efficiency of virtual screening, the theophylline aptamer can be quickly obtained at low cost, and a powerful calculation tool is provided for rational design and application of the aptamer.

Description

Technical Field

[0001] This invention belongs to the field of molecular biology technology, specifically relating to a theophylline aptamer and its virtual screening method and application. Background Technology

[0002] Aptamers are single-stranded DNA or RNA molecules obtained through systematic evolution of ligands by exponential enrichment (SELEX) technology. They can bind to target molecules with high affinity and specificity, including proteins, small molecules, metal ions, and even intact cells. Compared to traditional antibodies, aptamers have advantages such as simple synthesis, ease of modification, high stability, and low immunogenicity, showing broad application potential in fields such as biosensing, clinical diagnostics, targeted therapy, and drug delivery.

[0003] Theophylline, a commonly used bronchodilator, is widely used in the treatment of asthma and chronic obstructive pulmonary disease. However, its therapeutic window in vivo is narrow, and excessively high blood concentrations can easily cause toxic reactions such as nausea, palpitations, and even arrhythmias. Therefore, rapid and sensitive monitoring of theophylline concentrations is crucial for guiding clinical medication and ensuring medication safety. Traditional detection methods, such as high-performance liquid chromatography (HPLC) and enzyme-linked immunosorbent assay (ELISA), have limitations such as cumbersome operation, expensive equipment, or the need for specific antibodies. Aptamers, as novel recognition elements, hold promise for development into highly sensitive, low-cost, and easily detectable biosensors for rapid quantitative analysis of theophylline.

[0004] However, traditional SELEX technology suffers from long screening cycles, high costs, and low success rates, especially for small molecule targets. Due to the lack of well-defined binding domains similar to proteins, the screening process is highly random, making it difficult to obtain aptamers with both high affinity and high specificity. In recent years, computationally assisted aptamer screening methods have gradually emerged, mainly including molecular docking-based structural simulation methods and machine learning-based sequence prediction methods. Molecular docking methods assess binding free energy by simulating the three-dimensional interaction between aptamers and small molecules, but their prediction results are often unstable due to limitations in the accuracy of nucleic acid structure prediction, the flexibility of small molecule conformations, and high computational resource consumption. Machine learning methods rely on known aptamer-small molecule interaction data for training, but the current public databases have limited scale of such data, and there is a lack of unified and reliable strategies for constructing negative samples (non-binding pairs), leading to problems such as underfitting and weak generalization ability, making it difficult to achieve ideal results in practical screening. Summary of the Invention

[0005] This invention aims to provide a theophylline aptamer, its virtual screening method, and its application. The theophylline aptamer has high affinity and high specificity. The virtual screening method significantly improves the accuracy and efficiency of virtual screening, enabling the rapid and low-cost acquisition of theophylline aptamers. It provides a powerful computational tool for the rational design and application of aptamers.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A theophylline aptamer, wherein the theophylline aptamer is selected from one or more of MUT9, MUT16, MUT67, MUT78 or MUT89.

[0007] Preferably, the nucleotide sequence of MUT9 is shown in SEQ ID NO.1, the nucleotide sequence of MUT16 is shown in SEQ ID NO.2, the nucleotide sequence of MUT67 is shown in SEQ ID NO.3, the nucleotide sequence of MUT78 is shown in SEQ ID NO.4, and the nucleotide sequence of MUT89 is shown in SEQ ID NO.5.

[0008] The present invention also provides a virtual screening method for the theophylline aptamers, comprising the following steps: S1. Construct an aptamer-small molecule interaction dataset, which includes a positive dataset, a negative dataset, and an independent test set independent of the positive and negative datasets; S2. Extract k-mer features from aptamer sequences and molecular descriptors from small molecules; S3. Based on the k-mer features extracted in S2 and the molecular descriptors of small molecules, multiple single models are constructed using machine learning or deep learning algorithms. S4. Construct a multi-model combination by randomly selecting the single model obtained in S3, and determine the optimal model combination based on the independent test set in S1. S5. Using the optimal model combination from S4, predict the binding potential of candidate aptamer sequences with theophylline and screen out theophylline aptamers with high affinity.

[0009] Preferably, in S1, the negative dataset is constructed in at least one of the following ways: random pairing of positive samples, hierarchical clustering of small molecule targets, cross-pairing of protein aptamers and small molecules, and pairing based on aptamer sequence similarity.

[0010] Preferably, in S2, the k-mer feature includes the frequencies of all consecutive nucleotide subsequences when k=1, 2, 3, 4, which are combined to form multiple sets of aptamer descriptors.

[0011] Preferably, in S4, a different number of models are randomly selected from all single models to form a combination, the performance of each combination is evaluated using an independent test set, and the combination with the highest ROC-AUC value is selected as the optimal model combination.

[0012] Preferably, S5 includes the following steps: S51. Using the known theophylline aptamer structure as a template, perform traversal mutations on specific regions of the aptamer to construct a mutant aptamer sequence library. S52. Use the optimal model combination in S4 to predict the binding potential of all sequences in the mutant sequence library constructed in S51, and screen out candidate sequences that meet the preset conditions. S53. Perform molecular docking simulation on the candidate sequences screened in S52, and sort them according to their calculated binding free energy with theophylline, retaining the sequence with the optimal binding free energy. S54. Select compounds with structures similar to theophylline as interfering agents. Using the optimal model combination or molecular docking simulation in S4, exclude sequences that can bind to the interfering agents to obtain the final candidate sequences of theophylline aptamers with high affinity and high specificity.

[0013] Preferably, in S52, the preset condition is: (1) Only the sequences that are consistently predicted as positive by all single models in the optimal model combination in S4 are retained; (2) For each mutation sequence, calculate its cumulative ranking of binding probability in different single models of S3, and finally select a predetermined number of candidate sequences with high ranking from the mutation sequence library constructed in S51.

[0014] Preferably, in S53, based on the similarity between RNA and DNA in base pairing and structure prediction, the deoxythymidine in the sequence is replaced with uracil nucleotide, which is converted into the corresponding RNA sequence, and its three-dimensional structure is predicted based on the replaced RNA sequence.

[0015] The present invention also provides the application of the theophylline aptamer in the preparation of biosensors, reagent kits or detection devices for detecting or monitoring theophylline concentration.

[0016] Compared with the prior art, the present invention has the following advantages and technical effects: This invention discloses a theophylline aptamer, its virtual screening method, and its application. Using this virtual screening method, several novel theophylline aptamers were successfully screened from a mutant library of template aptamers. Experimental verification shows that these aptamers all exhibit superior affinity for theophylline compared to the original template. Furthermore, all aptamers show extremely low cross-reactivity to adenine, a structural analogue of theophylline, demonstrating excellent selectivity and specificity.

[0017] The high-affinity, high-specificity theophylline aptamer obtained in this invention can be directly used as a recognition probe for developing rapid, sensitive, and low-cost theophylline biosensors, detection kits, or portable detection devices. This provides a superior alternative for clinical drug monitoring, food safety testing, and environmental analysis, demonstrating clear application value and market potential.

[0018] Furthermore, this invention effectively overcomes the underfitting problem caused by data limitations of a single model by employing a multi-model combination strategy and utilizing a rigorously constructed independent test set for evaluation and optimization. The final model combination achieves an ROC-AUC value of 0.8862, significantly improving the accuracy and reliability of predictions. This method is not limited to specific targets; its data construction, model combination, and selection logic are universal. Simultaneously, the high-performance aptamers obtained through this method can be directly used as core recognition elements in the fields of biosensing and diagnostics.

[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0020] Figure 1 Mass spectrum of the mutant aptamer MUT9; Figure 2 Mass spectrum of the mutant aptamer MUT16; Figure 3 Mass spectrum of the mutant aptamer MUT67; Figure 4 Mass spectrum of the mutant aptamer MUT78; Figure 5 Mass spectrum of mutant aptamer MUT89; Figure 6 The binding affinity curves of different aptamers with theophylline are shown, among which, Figure 6 In the figure, A represents the 8K0T binding affinity curve. Figure 6 B in the figure represents the MUT9 binding affinity curve. Figure 6 C in the figure represents the binding affinity curve for MUT16. Figure 6 D in the figure represents the binding affinity curve for MUT67. Figure 6 E in the figure represents the binding affinity curve for MUT78. Figure 6 F in the figure represents the binding affinity curve of MUT89; Figure 7 The fluorescence recovery ratio of the aptamer to theophylline and adenine is given, where F represents the fluorescence intensity after the addition of the target and F0 represents the fluorescence intensity of the substrate without the addition of the target. Detailed Implementation

[0021] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0022] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0023] Source of experimental materials: Sodium chloride (Greagent, catalog number G81793J), anhydrous magnesium chloride (Maclean, M875550), tris(hydroxymethyl)aminomethane (Sinopharm, catalog number 30188360), graphene oxide (Bailingwei, catalog number: 909428), theophylline (Maclean, catalog number: T819162), adenine (Maclean, catalog number: A6279).

[0024] Instruments: Multifunctional plate reader (Shanghai Bai'ao Biotechnology Co., Ltd., equipment model: Synergy H1), constant temperature shaker (Hangzhou Aosheng Instrument Co., Ltd., equipment model: MSC-100).

[0025] In this invention, unless otherwise specified, all other test materials and instruments are conventional test materials in the field and can be purchased through commercial channels.

[0026] Example 1 A virtual screening method for small molecule target aptamers includes the following steps: S1. Collect 521 aptamer-small molecule interaction pairs from the public databases AptaDB and UTexas Aptamer Database as positive datasets. Then, generate 521 negative datasets by randomly pairing positive samples, hierarchical clustering of small molecule targets, cross-pairing of protein aptamers and small molecule targets, and pairing based on aptamer sequence similarity.

[0027] Nine sets of negative datasets were constructed using multiple strategies, and the specific construction methods are as follows: 1. Random pairing of positive samples: The aptamers and ligands in the positive dataset are randomly recombined, and 521 pairs are randomly selected from them as the negative dataset N1.

[0028] 2. Small molecule target hierarchical clustering pairing: Small molecule descriptors are generated using RDkit, and hierarchical clustering is performed based on these descriptors to classify all small molecule ligands into different categories. Based on the classification results, negative datasets N2 and N3 are constructed in two ways: (1) Pair the aptamer with small molecule ligands of a different class than the original binding ligand, and randomly select 521 pairs of non-redundant negative data pairs as the negative dataset N2.

[0029] (2) For each small molecule ligand, randomly select one from the different categories of aptamers of the known small molecule ligand as the negative aptamer to pair with it, and construct a negative dataset N3 containing 521 pairs of non-redundant negative data pairs.

[0030] 3. For each ligand, a negative aptamer is defined as a sequence with low similarity to the original positive aptamer. For each small molecule ligand, an aptamer is randomly selected from the corresponding aptamer library according to similarity thresholds (≤0.5, ≤0.4, ≤0.3, ≤0.2, and minimum similarity) to generate negative datasets N4-N8, each containing 521 pairs of non-redundant negative data.

[0031] 4. Protein aptamer-small molecule target cross-pairing: Each small molecule ligand is randomly cross-paired with an aptamer from the AptaDB database targeting the protein to construct a negative dataset N9 containing 521 pairs of non-redundant negative data pairs.

[0032] 80% of the positive samples (aptamers that can bind to the corresponding small molecules) and 80% of the negative samples (aptamers that cannot bind to the corresponding small molecules) were randomly selected as the training set. To simulate a real application environment, an independent test set was constructed with a positive to negative sample ratio of 1:20. The negative samples were generated by randomly pairing positive samples, and it was ensured that this part of the negative samples did not overlap with the previously constructed negative dataset. S2. The nucleotide composition features of aptamers were extracted using the k-mer method. For each aptamer sequence, the frequencies of all k-mers (continuous nucleotide subsequences of length k) were calculated for k=1, 2, 3, and 4. Fifteen sets of aptamer descriptors of different dimensions were constructed by combining features of different k values. For small molecule targets, 209-dimensional small molecule descriptors were generated using RDkit (version: 2023.9.6). S3. Based on the k-mer features extracted in S2 and the molecular descriptors of small molecules, multiple single models are constructed using machine learning or deep learning algorithms, with the following characteristics: Eight machine learning algorithms were tested using scikit-learn (version 1.5.0), including Gradient Boosting Decision Tree (GBDT), K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), Naïve Bayes (NB), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), and Adaptive Boosting (AdaBoost). Additionally, three deep learning algorithms were tested using PyTorch (version 2.3.0+CPU), including Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and Bidirectional Recurrent Neural Network (Bi-RNN). S4. Based on the single models constructed in S3, a combined model was constructed using a random sampling strategy. n models were randomly selected from all single models (n taking values ​​of 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29 respectively), forming a model set. For each set value of n, the above random sampling process was repeated 100,000 times to obtain combined models covering different combinations and model compositions. The model performance was evaluated using an independent test set, and the optimal model combination consisting of 9 single models was finally determined (as shown in Table 1), with an area under the receiver operating characteristic curve (ROC-AUC) of 0.8862. Table 1. Single Model Composition in Optimal Multi-Model Combination

[0033] S5. Utilize the optimal model combination from S4 to predict the binding potential of candidate aptamer sequences to small molecule targets, and screen out high-affinity small molecule target aptamers.

[0034] The specific steps for screening theophylline aptamers with high affinity and high specificity are as follows: S51. Construction of mutant sequence library: Template aptamer: A DNA aptamer (PDB ID: 8K0T, sequence as shown in SEQ ID NO.6) that binds to theophylline and has a complete structure was retrieved from the Protein Structure Database (PDB) and used as a mutation template.

[0035] SEQ ID NO. 6: GCGGTGGTCTATTCATAGGCGTCCGCCGC.

[0036] RNAfold was used to predict the secondary structure of the mutant template, identify the hairpin loop region, and keep the aptamer stem structure sequence unchanged. A traversal mutation was performed at each base site of the hairpin loop, resulting in mutations to A, T, C, and G, respectively; ultimately generating 1,073,741,824 (4 15 ) mutant aptamer sequences, with a specific mutation site of 15 bases (specific mutation sites are 8-22).

[0037] S52. Use the optimal model combination in S4 to predict the binding potential of all sequences in the mutation sequence library constructed in S51, and screen out candidate sequences that meet the preset conditions. During the prediction process, each single model in the combined model independently predicts each candidate sequence and outputs the corresponding binding probability value.

[0038] The preset conditions are: (1) Only the sequences that are consistently predicted as positive by all single models in the optimal model combination in S4 are retained; (2) For each mutation sequence, calculate its cumulative ranking of binding probability in different single models of S3, and finally select the top 100 candidate sequences from the mutation sequence library constructed in S51.

[0039] S53. Molecular docking simulations were performed on the candidate sequences screened in S52. Based on the similarity between RNA and DNA in base pairing and structural prediction, deoxythymidine (T) in the sequences was replaced with uracil (U), converting them into corresponding RNA sequences. The three-dimensional structure of the replaced RNA sequences was then predicted. AutoDock Vina was used for molecular docking simulations. The optimal conformations and their corresponding binding free energies were compared. The mutant aptamers were ranked in descending order of their binding free energies with theophylline, and the top 5 sequences with the best binding free energies were retained.

[0040] S54. Adenine, which has a similar structure to theophylline, was selected as an interfering agent. The optimal model combination in S4 was used to predict the binding of adenine to the top 5 mutant aptamers of the above molecular docking. Mutant sequences that could bind to the interfering compound were excluded, and finally 5 aptamer sequences (MUT9, MUT16, MUT67, MUT78, MUT89) were retained for subsequent experimental verification.

[0041] The sequences of MUT9, MUT16, MUT67, MUT78, and MUT89 from Example 1 above were synthesized by Shanghai Jierui Biotechnology Co., Ltd., and their binding free energy information is shown in Table 2 below.

[0042] Table 2. Mutant aptamer sequences targeting theophylline

[0043] The quality of the synthesized DNA aptamers was assessed using a linear quadrupole ion trap (LTQ) liquid chromatography-mass spectrometry (LC-MS / MS). The results are as follows: Figure 1-5 As shown.

[0044] Depend on Figures 1-5 As can be seen, the mass spectrum shows only one sharp and symmetrical main peak, with no obvious impurity peaks. Furthermore, the deviation between the measured main peak mass and the theoretical molecular weight shown in the mass spectrum is ≤±0.05%, further demonstrating the successful synthesis of the aptamer.

[0045] The specific experimental protocol for verifying the binding affinity of the above candidate aptamer sequences to theophylline is as follows: Aptamer powder was dissolved to 100 μM using TE buffer, and then diluted to 100 nM using 20 mM Tris-HCl buffer (pH=7.4). The 100 nM aptamer solution was incubated at 90 °C for 10 min to allow the aptamers to fully expand and form single strands, followed by slow cooling to 25 °C to form hairpin structures. Subsequently, 100 μL of 100 nM FAM-labeled aptamers and 100 μL of theophylline standard solutions of different concentrations were incubated at 37 °C for 60 min. The fluorescence of unbound aptamers was quenched by adding 100 μL of optimal graphene oxide solution, and the fluorescence intensity was measured using a microplate reader after incubation for 30 min. The dissociation constant Kd was calculated by fitting the binding curve using the Langmuir binding model to assess affinity. The results are as follows: Figure 6 As shown.

[0046] To verify the aptamer specificity, the fluorescence response of the aptamer to 100 μM theophylline and its structural analogue adenine was detected under the same conditions. The binding specificity of the aptamer to theophylline was evaluated by the signal difference. The results are as follows: Figure 7 As shown.

[0047] The positive control aptamer showed a significant fluorescence recovery signal with increasing concentration, and its dissociation constant Kd was 101.11 μM. Figure 6 The presence of "A" indicates successful construction of the fluorescent sensor. Compared to the positive control, the affinity enhancement of MUT16 (Kd=31.10μM) and MUT67 (Kd=25.04μM) was particularly significant, increasing by approximately 3.25-fold and 4.20-fold, respectively. Figure 6 C and Figure 6 The affinity of MUT9 (Kd=43.46μM), MUT78 (Kd=47.33μM), and MUT89 (Kd=73.77 μM) was increased by approximately 2.33 times, 2.14 times, and 1.37 times, respectively. Figure 6 B in Figure 6 E and Figure 6 (F in the original text). This demonstrates the rationale for screening high-affinity aptamers using mutant sequences obtained through mutated hairpin loop structures. The above results strongly confirm the accuracy and application value of multi-model combined prediction. This strategy can screen high-affinity aptamers from a large aptamer sequence mutation library, providing reliable computational support for aptamer design and screening.

[0048] aptamer specificity assay results as follows Figure 7 As shown.

[0049] Depend on Figure 7 The results show that the positive control 8K0T exhibited a significantly higher fluorescence response to theophylline than adenine, demonstrating good specificity. The fluorescence recovery ratios (RRRs) of MUT16, MUT67, and MUT89 for theophylline were 0.642, 0.431, and 0.299, respectively, significantly higher than those for adenine. MUT78 also showed a higher response to theophylline than adenine, with only a weak cross-reactivity. These results indicate that both the positive control and the mutant can specifically recognize theophylline, while exhibiting extremely low response to the structural analogue adenine.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A theophylline aptamer, characterized in that, The theophylline aptamer is selected from one or more of MUT9, MUT16, MUT67, MUT78 or MUT89.

2. The theophylline aptamer according to claim 1, characterized in that, The nucleotide sequence of MUT9 is shown in SEQ ID NO.1, the nucleotide sequence of MUT16 is shown in SEQ ID NO.2, the nucleotide sequence of MUT67 is shown in SEQ ID NO.3, the nucleotide sequence of MUT78 is shown in SEQ ID NO.4, and the nucleotide sequence of MUT89 is shown in SEQ ID NO.

5.

3. A virtual screening method for theophylline aptamers as described in any one of claims 1 or 2, characterized in that, Includes the following steps: S1. Construct an aptamer-small molecule interaction dataset, which includes a positive dataset, a negative dataset, and an independent test set independent of the positive and negative datasets; S2. Extract k-mer features from aptamer sequences and molecular descriptors from small molecules; S3. Based on the k-mer features extracted in S2 and the molecular descriptors of small molecules, multiple single models are constructed using machine learning or deep learning algorithms. S4. Construct a multi-model combination by randomly selecting the single model obtained in S3, and determine the optimal model combination based on the independent test set in S1. S5. Using the optimal model combination from S4, predict the binding potential of candidate aptamer sequences with theophylline and screen out theophylline aptamers with high affinity.

4. The method according to claim 3, characterized in that, In S1, the negative dataset is constructed in at least one of the following ways: random pairing of positive samples, hierarchical clustering of small molecule targets, cross-pairing of protein aptamers and small molecules, and pairing based on aptamer sequence similarity.

5. The method according to claim 3, characterized in that, In S2, the k-mer features include the frequencies of all consecutive nucleotide subsequences when k=1, 2, 3, 4, which are combined to form multiple sets of aptamer descriptors.

6. The method according to claim 3, characterized in that, In S4, different numbers of models are randomly selected from all single models to form combinations. The performance of each combination is evaluated using an independent test set, and the combination with the highest ROC-AUC value is selected as the optimal model combination.

7. The method according to claim 3, characterized in that, S5 includes the following steps: S51. Using the known theophylline aptamer structure as a template, perform traversal mutations on specific regions of the aptamer to construct a mutant aptamer sequence library. S52. Use the optimal model combination in S4 to predict the binding potential of all sequences in the mutant sequence library constructed in S51, and screen out candidate sequences that meet the preset conditions. S53. Perform molecular docking simulation on the candidate sequences screened in S52, and sort them according to their calculated binding free energy with theophylline, retaining the sequence with the optimal binding free energy. S54. Select compounds with structures similar to theophylline as interfering agents. Using the optimal model combination or molecular docking simulation in S4, exclude sequences that can bind to the interfering agents to obtain the final candidate sequences of theophylline aptamers with high affinity and high specificity.

8. The method according to claim 7, characterized in that, In S52, the preset condition is: (1) Only the sequences that are consistently predicted as positive by all single models in the optimal model combination in S4 are retained; (2) For each mutation sequence, calculate its cumulative ranking of binding probability in different single models of S3, and finally select a predetermined number of candidate sequences with high ranking from the mutation sequence library constructed in S51.

9. The method according to claim 7, characterized in that, In S53, based on the similarity between RNA and DNA in base pairing and structure prediction, the deoxythymidine in the sequence is replaced with uracil nucleotide, which is converted into the corresponding RNA sequence, and its three-dimensional structure is predicted based on the replaced RNA sequence.

10. The use of the theophylline aptamer as described in claim 1 or 2 in the preparation of biosensors, kits, or detection devices for detecting or monitoring theophylline concentration.

Images