Brain-targeted peptide screening method

Abstract

The invention discloses a brain-targeted peptide screening method, which comprises the following steps: introducing LSTM, trRosetttaX, AlphaFod2, Point Transformer deep learning technology and Autolock Vina molecular docking tool, generating a brain-targeted peptide sequence, predicting a deep learning peptide structure, docking protein and peptide, predicting spatial affinity, screening and increasing biochemical characteristic information of a target spot and peptide through Point Transformer deep learning, and screening and identifying the brain-targeted peptide according to the biochemical characteristic information of the target spot and peptide. A dual structure of molecular docking binding free energy scoring and Point Transform spatial affinity scoring is used for screening, and brain-targeted peptide screening is achieved. According to the method, the screening process of the brain-targeted peptide can be accelerated, the experimental period is shortened, the screening cost is reduced, rapid screening is achieved, the binding precision and screening accuracy of the peptide and a target spot are improved, and candidate peptide with low affinity is effectively screened out.

Description

Technical Field

[0001] The invention relates to the technical field of targeting peptide screening, and in particular to a method for screening brain targeting peptides. Background Art

[0002] Traditional brain-targeting peptide screening technology mainly relies on animal experiments and cell experiments to identify targeting peptides that can cross the blood-brain barrier. However, this method is time-consuming, costly, and has experimental errors and ethical issues. In practical applications, screening for peptides that cross the blood-brain barrier remains a technical difficulty.

[0003] Although existing computer screening methods can accelerate the molecular screening process through affinity prediction to a certain extent, they still face many limitations in the screening of brain-targeted peptides. Traditional molecular docking methods (such as Autodock Vina) are limited by the insufficient simulation ability of molecular flexibility and cannot effectively predict the binding sites and binding forces between proteins and ligands, resulting in insufficient accuracy in target identification. Although virtual screening based on artificial intelligence (AI) has advantages in efficiency and automation, there is still room for improvement in specificity and accuracy. Especially in the screening of brain-targeted peptides, due to the limitations of training data diversity and coverage, the AI ​​model's specific recognition of targets is relatively limited and complex structures cannot be reliably identified. In addition, the existing deep learning models are difficult to accurately capture the interactions between complex molecules in molecular structure prediction, and cannot efficiently restore the binding mode of protein-ligand complexes, which in turn limits the identification and optimization of high-affinity peptides. These technologies have not yet achieved ideal results in the application of brain-targeted peptide screening. Therefore, further optimizing the method system that integrates molecular modeling and deep learning, improving the prediction accuracy of target protein binding sites, and improving the accuracy of high-throughput screening are important directions for promoting brain-targeted peptide research. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the object of the present invention is to provide a method for screening brain targeting peptides to improve the efficiency and accuracy of brain targeting peptide screening.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0006] A method for screening brain targeting peptides is provided, comprising the following steps:

[0007] LSTM, trRosettaX, AlphaFold2, Point Transformer deep learning technology and Autodock Vina molecular docking tool were introduced to generate brain-targeted peptide sequences, predict deep learning peptide structures, dock proteins and peptides, and predict spatial affinity. The biochemical characteristic information of targets and peptides was increased through Point Transformer deep learning screening. The dual structure of molecular docking combined with free energy scoring and Point Transformer spatial affinity scoring was used for screening to achieve brain-targeted peptide screening.

[0008] Furthermore, the method for generating a brain-targeting peptide sequence by pe p_LSTM specifically includes the following sub-steps:

[0009] (1) Based on the LSTM network, the pep_LSTM model is trained on the active peptide database to enable it to generate a variety of peptide sequences;

[0010] (2) fine-tuning the data using peptides related to the blood-brain barrier to make the model more accurate in predicting results;

[0011] (3) The target CD98hc sequence information was input into the model to generate a virtual peptide sequence with potential brain targeting;

[0012] (4) The target CD98hc and peptide were docked by using Docking technology to screen out relatively accurate candidate peptides.

[0013] Furthermore, the method for structure prediction specifically includes the following sub-steps:

[0014] (1) The peptide sequence generated by the pep_LSTM model is used to predict the three-dimensional structure of PDB or PDBQT using trRosettaX and Alpha Fold2 technology;

[0015] (2) The folding conformation of the peptide is predicted through deep learning models and co-evolution information, and the sequence is converted into PDB or PDBQT file format to provide a basis for subsequent peptide docking.

[0016] Furthermore, the method of Autodock Vina molecular docking specifically includes the following sub-steps:

[0017] (1) Using Vina, the processed peptide structure was molecularly docked with the target CD98hc to calculate the binding free energy;

[0018] (2) Vina selects peptide sequences with lower binding energy through energy scoring as candidate peptides with potential targeting and blood-brain barrier crossing capabilities.

[0019] Furthermore, the Point Transformer deep learning screening method specifically includes the following sub-steps:

[0020] (1) Obtain the protein-ligand interaction datasets from the RCSB PDB and BrainPep databases and preprocess the datasets;

[0021] (2) Divide the data set into a training set:test set ratio of 8:2;

[0022] (3) Through training and optimization on the Point Transformer model with the training set, the biochemical characteristic information of the hydrophobicity, polarity, and charge distribution of the target and peptide is increased to make it more suitable for the needs of brain targeting peptide screening;

[0023] (4) Put the target data and peptide data set into the model to perform affinity prediction;

[0024] (5) Point Transformer evaluates peptide-target interactions by analyzing three-dimensional spatial affinity, hydrophobicity, polarity, and charge distribution, and screens out peptide sequences with higher targeting and specificity.

[0025] Furthermore, the dual structure screening using molecular docking combined with free energy scoring and Point Transformer spatial affinity scoring also includes the following steps: predicting peptide toxicity of candidate peptides using Ad met and Toxi n Pred, and deleting candidate peptides predicted to be toxic.

[0026] Furthermore, it also includes: marking the glycosylation sites of the target CD98hc, searching for N-glycosylation through the formula Asn-X-Ser / Thr (X≠Pro), and searching for O-glycosylation through the formula Ser / Thr. In addition, there are glycosylation sites reported in the literature. After all glycosylation sites are marked and recorded, regular expressions are used to screen whether the results of pocket docking contain active sites. If so, they are deleted.

[0027] The beneficial effects of the present invention are:

[0028] 1. Improvement of Point Transformer algorithm: In addition to the three-dimensional spatial position of the algorithm itself, the biochemical characteristics of the target and peptide are also added: hydrophobicity, polarity, charge distribution, etc., so that Point Transformer not only relies on geometric information, but also considers the biochemical characteristics between molecules, thereby improving the accuracy of screening.

[0029] 2. Multi-level affinity screening mechanism: The present invention is the first to apply the dual structural screening mechanism of molecular docking combined with free energy scoring and Point Transformer spatial affinity scoring, which effectively improves the specificity and screening accuracy of brain-targeting peptides for the target CD98hc.

[0030] 3. Enhanced binding effect: The presence of glycosylation sites will lead to increased steric hindrance and charge distribution and polar interference, affecting the binding of the target and the peptide. The present invention avoids the influence of glycosylation sites on the binding force between the target and the peptide through an algorithm, making it easier to bind. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Prediction flow chart for the generation of virtual peptides for the blood-brain barrier in step 1;

[0032] Figure 2 The process and scoring of peptide structure transformation in step 2;

[0033] Figure 3 This is the Point Transformer prediction and screening flow chart in step 5;

[0034] Figure 4 Confusion matrix (left) and ROC curve (right) for PointTransformer virtual screening in step 5

[0035] Figure 5 This is the improved part of the Point Transformer algorithm in step 5;

[0036] Figure 6 This is the Autodock Vina selection flow chart for step 4;

[0037] Figure 7 This is a screening flow chart for combining the two virtual screening methods in step six;

[0038] Figure 8 The active site (left) and glycosylation site (right) of the target in step six;

[0039] Fig. 9 The structure of FITC connected to the N-terminus of the polypeptide in step 7;

[0040] Figure 10 The result diagram of cell transcytosis efficiency in step 7;

[0041] Fig.11 This is the fluorescence data of the mouse brain in step 7. DETAILED DESCRIPTION

[0042] The specific implementation modes of the present invention are described below to facilitate those skilled in the art to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0043] This paper proposes an AI-enabled brain-targeted peptide screening method, which introduces deep learning technologies such as LSTM, trRosettaX, AlphaFold2, Point Transformer, and the Autodock Vina molecular docking tool, and realizes the functions of related peptide generation model, deep learning peptide structure prediction, protein and peptide docking, and spatial affinity prediction, and constructs a full-process intelligent screening system. The specific steps are as follows:

[0044] Step 1: Generate targeted peptide sequences through pe p_LSTM

[0045] (1) Based on the LSTM network, the pep_LSTM model is trained on a large number of active peptide databases to enable it to generate a variety of peptide sequences to improve the diversity and coverage of the initial screening;

[0046] (2) fine-tuning the data using peptides related to the blood-brain barrier to make the model more accurate in predicting results;

[0047] (3) The target CD98hc sequence information was input into the model to generate a virtual peptide sequence with potential brain targeting;

[0048] (4) The target CD98hc and peptide were docked by using Docking technology to screen out relatively accurate candidate peptides.

[0049] Step 2: Structure prediction

[0050] (1) The peptide sequences generated by the pep_LSTM model were used to predict the three-dimensional structure of PDB or PDBQT using trRosettaX and Alpha Fold2 technologies. Sequences with a length greater than 20 were predicted using AlphaFold2 technology, and sequences with a length between 0 and 20 were predicted using trRosettaX technology;

[0051] (2) The folding conformation of the peptide is predicted by using a deep learning model and co-evolution information, and the sequence is converted into PDB or PDBQT file format. At the same time, a score image between different residues is generated (the specific prediction result score is in Figure 2 The result at the bottom) provides a basis for subsequent peptide docking.

[0052] Step 3: Open Babel Processing

[0053] The PDB or PDBQT structures of the target CD98hc and the generated peptides were further processed using Open Babel to add hydrogen atoms and electrons to ensure the integrity of the molecular structure; this processing step ensures that the peptides have reasonable atomic features in the subsequent molecular docking process and improves the reliability of the docking results.

[0054] Step 4: Autodock Vina molecular docking

[0055] (1) Using Vina, the processed peptide structure was molecularly docked with the target CD98hc to calculate the binding free energy;

[0056] (2) Vina selects peptide sequences with lower binding energy through energy scoring as candidate peptides with potential targeting and blood-brain barrier crossing capabilities.

[0057] Step 5: Point Transformer Deep Learning Screening

[0058] (1) Obtain the protein-ligand interaction datasets from the RCSB PDB and BrainPep databases and preprocess the datasets;

[0059] (2) The data set is divided into training set: test set = 8:2;

[0060] (3) By training and optimizing the Point Transformer model with the training set, the biochemical characteristic information of the target and peptide, such as hydrophobicity, polarity, and charge distribution, is added to make it more suitable for the screening of brain-targeted peptides;

[0061] (4) Put the target data and peptide data set into the model to perform affinity prediction;

[0062] (5) Point Transformer evaluates peptide-target interactions by analyzing three-dimensional spatial affinity, hydrophobicity, polarity, and charge distribution, and screens out peptide sequences with higher targeting and specificity.

[0063] like Figure 4 As shown, on the left: Confusion matrix analysis can provide a deeper understanding of the model's performance on different categories. The confusion matrix data on the test set is shown below:

[0064] Top left: low affinity predicted as low affinity (true negative class): 145;

[0065] Top right: Low affinity predicted as high affinity (false negative): 56v

[0066] Bottom left: high affinity predicted as low affinity (false positive): 39;

[0067] Bottom right: High affinity predicted as high affinity (true class): 160.

[0068] The confusion matrix can be used to calculate the precision and recall.

[0069] Accuracy: Among all the pairs predicted to have high affinity, the model correctly identified 160 true positive samples, so the accuracy is:

[0070]

[0071] Recall: Among all the samples with high affinity, the model successfully identified 160 true positive samples, so the recall is:

[0072]

[0073] Right: AUC (Area Under Curve) is the core indicator for measuring the classification effect of the model. Its value range is between 0.5 and 1. The higher the value, the better the classification performance.

[0074] Step 6: Final screening

[0075] (1) Combine the molecular docking screening results of Autodock Vina with the deep learning screening results of Point Transformer to comprehensively rank the binding energy and affinity scores;

[0076] (2) Peptide toxicity prediction of candidate peptides was performed using Ad met and Toxi n Pred, and candidate peptides predicted to be toxic were deleted;

[0077] (3) Annotate the glycosylation sites of the target CD98hc. N-glycosylation is searched by the formula Asn-X-Ser / Thr (X≠Pro), and O-glycosylation is searched by the formula Ser / Thr. In addition, there are glycosylation sites reported in the literature. After all glycosylation sites are annotated and recorded, regular expressions are used to filter whether the pocket docking results contain active sites. If so, they are deleted;

[0078] (4) Finally, the optimal peptide sequence targeting the blood-brain barrier was screened out.

[0079] Step 7:

[0080] (1) Labeling the screened candidate peptides with FITC and synthesizing them;

[0081] (2) Construct an in vitro Transwell model to verify the targeting selection and blood-brain barrier-crossing efficiency of candidate peptides;

[0082] (3) Verify whether the targeting peptide can cross the blood-brain barrier by tail vein injection in mice.

[0083] Verification results such as Figure 9-11 As shown. In summary, the method of the present invention can speed up the screening process of brain targeting peptides, shorten the experimental cycle, reduce the screening cost, achieve rapid screening, improve the binding accuracy of peptides to targets and the screening accuracy, and effectively screen out candidate peptides with low affinity. The screening process of the present invention has good versatility and can be expanded to meet the screening needs of other targeting peptides, providing a new idea for the development of targeting peptides for various diseases.

[0084] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered exemplary and non-restrictive in all respects, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present invention.

[0085] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for screening brain targeting peptides, characterized in that: The following steps are involved: LSTM, trRosettaX, AlphaFold2, Point Transformer deep learning technology and AutodockVina molecular docking tool were introduced to generate brain-targeted peptide sequences, predict deep learning peptide structures, dock proteins and peptides, and predict spatial affinity. The biochemical characteristic information of targets and peptides was increased through Point Transformer deep learning screening. The dual structure of molecular docking combined with free energy scoring and Point Transformer spatial affinity scoring was used for screening to achieve brain-targeted peptide screening.

2. The brain targeting peptide screening method according to claim 1, characterized in that: The method for generating brain-targeting peptide sequences by pep_LSTM specifically includes the following sub-steps: (1) Based on the LSTM network, the pep_LSTM model is trained on the active peptide database to enable it to generate a variety of peptide sequences; (2) fine-tuning the data using peptides related to the blood-brain barrier to make the model more accurate in predicting results; (3) The target CD98hc sequence information was input into the model to generate a virtual peptide sequence with potential brain targeting; (4) The target CD98hc and peptide were docked by using Docking technology to screen out relatively accurate candidate peptides.

3. The brain targeting peptide screening method according to claim 1, characterized in that: The structure prediction method specifically includes the following sub-steps: (1) The peptide sequence generated by the pep_LSTM model is used to predict the three-dimensional structure of PDB or PDBQT using trRosettaX and Alpha Fold2 technology; (2) The folding conformation of the peptide is predicted through deep learning models and co-evolution information, and the sequence is converted into PDB or PDBQT file format to provide a basis for subsequent peptide docking.

4. The method for screening brain targeting peptides according to claim 1, characterized in that: The molecular docking method of Autodock Vina specifically includes the following sub-steps: (1) Vina was used to perform molecular docking between the processed peptide structure and the target CD98hc and calculate the binding free energy; (2) Vina selected peptide sequences with lower binding energy through energy scoring as candidate peptides with potential targeting and blood-brain barrier crossing capabilities.

5. The brain targeting peptide screening method according to claim 1, characterized in that: The Point Transformer deep learning screening method specifically includes the following sub-steps: (1) Obtain the protein-ligand interaction datasets from the RCSB PDB and BrainPep databases and preprocess the datasets; (2) The data set is divided into training set: test set = 8:2; (3) Through training and optimization on the Point Transformer model with the training set, the biochemical characteristic information of the hydrophobicity, polarity, and charge distribution of the target and peptide is increased to make it more suitable for the needs of brain targeting peptide screening; (4) Put the target data and peptide data set into the model to perform affinity prediction; (5) Point Transformer evaluates peptide-target interactions by analyzing three-dimensional spatial affinity, hydrophobicity, polarity, and charge distribution, and screens out peptide sequences with higher targeting and specificity.

6. The method for screening brain targeting peptides according to claim 1, characterized in that: The screening using the dual structure of molecular docking combined with free energy score and spatial affinity score of Point Transformer also includes the following steps: peptide toxicity prediction of candidate peptides by Admet and ToxinPred, and candidate peptides predicted to be toxic are deleted.

7. The method for screening brain targeting peptides according to claim 6, characterized in that: Also includes: The glycosylation sites of the target CD98hc were annotated. N-glycosylation was searched using the formula Asn-X-Ser / Thr (X≠Pro), and O-glycosylation was searched using the formula Ser / Thr. In addition, there were glycosylation sites reported in the literature. After all glycosylation sites were annotated and recorded, regular expressions were used to filter out whether the pocket docking results contained active sites. If so, they were deleted.

Images