Calculation-based phosphorylated protein analysis method

Abstract

The present invention relates to a calculation-based method for analyzing phosphorylated proteins, comprising: a phosphorylated protein sequence prediction step (20) for predicting the probability of presence of each phosphorylation combination of a target protein using a calculation model, and selecting a phosphorylated protein in which the probability of presence is greater than or equal to a prescribed probability; a phosphorylated protein three-dimensional structure prediction step (30): respectively replacing phosphorylated serine and phosphorylated threonine in the phosphorylated protein sequence obtained by the phosphorylated protein sequence prediction step with aspartic acid and glutamic acid, and predicting the three-dimensional structure of the replaced protein by using a calculation model; and a classification and clustering step (40): classifying and clustering the phosphorylated proteins according to the three-dimensional structure obtained in the phosphorylated protein three-dimensional structure prediction step. According to the method, the three-dimensional structure of the phosphorylated protein can be predicted comprehensively, quickly and at low cost only by means of calculation of the AI model without an experiment, and then the phosphorylated protein capable of being used as a biomarker or a drug action target is screened based on the obtained three-dimensional structure.

Description

Technical Field

[0001] The present invention relates to a computationally based analysis method for phosphorylated proteins, and more specifically, to a method for analyzing the three-dimensional structure of phosphorylated proteins by calculation using an AI model, and then screening phosphorylated proteins that can serve as biomarkers, drug targets, etc. based on the obtained three-dimensional structure. Background Art

[0002] Protein is a biological macromolecule composed of one or more polypeptide chains. It is an important component of the cell structure in the body and the main bearer of life activities. Each polypeptide chain constituting a protein is composed of twenty to hundreds of amino acid residues connected by peptide bonds. There are about 20 kinds of amino acids constituting the peptide chain, which are called basic amino acids. The order of arrangement of amino acid residues in the protein peptide chain is called the primary structure of the protein. The primary structure (amino acid sequence) of the protein determines the three-dimensional structure of the protein (including secondary structure, tertiary structure and quaternary structure), and the function of the protein is closely related to its three-dimensional structure. The change of the three-dimensional structure of the protein is related to the disease, and similar pathological conditions generally have similar protein structures (see non-patent literature 1).

[0003] The reversible process of protein phosphorylation and dephosphorylation regulates all life activities including cell proliferation, development, differentiation, signal transduction, apoptosis, neural activity, muscle contraction and tumorigenesis. Protein phosphorylation mainly occurs on two amino acids, serine and threonine, and tyrosine. These residues have free hydroxyl groups and are not charged. When these residues are phosphorylated, the protein has a charge, which changes the three-dimensional structure and further causes changes in protein activity. For example, phosphorylation of neutral amino acids serine (S) and threonine (T) will form negatively charged phosphorylated serine (pS) and phosphorylated threonine (pT), which will cause changes in the three-dimensional structure of the protein and change the function of the protein.

[0004] Abnormal phosphorylation of proteins often leads to the occurrence of diseases. For example, Alzheimer's disease is closely related to hyperphosphorylated Tau protein (microtubule-associated protein tau) (see patent documents 1 and 4), hereditary Parkinson's disease is related to phosphorylated LRRK2 (see non-patent document 2), and even the reproduction of the new coronavirus in the body and the damage to the body are closely related to protein phosphorylation (see non-patent document 3).

[0005] Therefore, the research and detection of phosphorylated proteins has important clinical significance. Phosphorylated amino acid residues can be used as biomarkers for disease diagnosis. For example, phosphorylated proteins T181, S202, and T217 can be used as biomarkers for the diagnosis of Alzheimer's disease (see Patent Document 1), phosphorylated DCBLD2 Y750 It can be used as a biomarker to determine the malignancy of glioma (see Patent Document 2). Phosphorylated amino acid residues can also be used as biomarkers to determine drug responsiveness, for example, phosphorylated Akt can be used as a biomarker to determine the resistance of tumors to microtubule targeting agents (see Patent Document 3), or can also be used as a target for drug therapy, such as specific antibodies that recognize the phosphorylated epitope of Tau protein (see Patent Document 4), antibodies against the phosphorylated tyrosine residue at position 74 of TOPK, etc. (see Patent Document 5).

[0006] In the prior art, phosphorylated proteins are usually detected by experimental methods. Specifically, the sequence and phosphorylation sites of the target protein are analyzed, the phosphorylated protein is obtained by experimental methods, and then the structure of the phosphorylated protein is identified by experimental methods such as X-ray, MR, and Cryo-EM.

[0007] However, the experimental method is time-consuming and laborious, and it is difficult to study all phosphorylated proteins. For example, Tau protein has 441 amino acids, of which 80 are serine (S) and threonine (T). These serine and threonine can be phosphorylated. Therefore, theoretically, there are 2 possible phosphorylated Tau protein states. 80 (about 10 24 ) species, even if only considering the phosphorylation state of less than 2 (i.e. 1 or 2) amino acid sites, there are 3241 species, so it is almost impossible to comprehensively study the structure of phosphorylated Tau protein by experimental methods alone. In addition, there are about 20,000 proteins in the human proteome, and almost every one of them is phosphorylated, so it is impossible to thoroughly study them all by simple experimental methods. In addition, some proteins are difficult to extract and purify, and the structure of some proteins is difficult to identify by experimental methods, so they are not suitable for research by experimental methods.

[0008] In view of the above situation, methods for analyzing phosphorylated proteins based on deep learning AI models have received widespread attention. For example, various methods for calculating and predicting protein phosphorylation sites have been proposed (see non-patent document 6, patent documents 6 to 10).

[0009] In addition, for the prediction of protein three-dimensional structure, two AI models proposed in 2021, AlphaFold2 (see non-patent document 4) and RosseTTAFold (see non-patent document 5), have received widespread attention. Both models are algorithms that predict the three-dimensional structure of proteins based on amino acid sequences. These algorithms provide fast and relatively accurate prediction results for protein structure prediction, but these algorithms cannot yet predict the three-dimensional structure of phosphorylated proteins.

[0010] Prior art literature

[0011] Patent Literature

[0012] Patent document 1: CN114761807A

[0013] Patent document 2: CN104083775A

[0014] Patent document 3: CN103502467A

[0015] Patent document 4: CN103502272A

[0016] Patent document 5: CN105646714A

[0017] Patent Document 6: JP Patent Publication No. 2018-206374A

[0018] Patent document 7: CN113539364A

[0019] Patent document 8: CN113450872A

[0020] Patent document 9: CN113096722A

[0021] Patent document 10: CN101710365A

[0022] Non-patent literature

[0023] Non-patent literature 1: Shi, Y. et al. Structure-based classification of tauopathies. Nature 598, 359–363 (2021).

[0024] Non-patent literature 2: Kyle B. Fraser et al., Neurology, 86(11), March 15, 2016; DOI: doi.org / 10.1212 / WNL.00000000000002436.

[0025] Non-patent literature 3: Ryan M. Hekman et al.; Actionable Cytopathogenic Host Responses of Human Alveolar Type 2 Cells to SARS-CoV-2. Molecular Cell, 2020; DOI: 10.1016 / j.molcel.2020.11.028.

[0026] Non-patent literature 4: Jumper et al., Highly accurate protein structure prediction with AlphaFold, Nature. / 10.1038 / s41586-021-03819-2 (2021).

[0027] Non-patent document 5: Baek, et al., Accurate prediction of protein structures and interactions using a three-track neural network; Science, 15, eabj8754. / 10.1126 / science.abj8754(2021).

[0028] Non-patent document 6: Fenglin Luo et al., DeepPhos: prediction of proteinphosphorylation sites with deep learning, Bioinformatics, 35(16), 2019, 2766–2773. Summary of the invention

[0029] Problems to be solved by the invention

[0030] As mentioned above, the state of the art is that no AI model or calculation method capable of predicting the three-dimensional structure of phosphorylated proteins has been established so far.

[0031] The present invention is completed in response to this technical problem. The purpose of the present invention is to provide a computationally based analysis method for phosphorylated proteins, which computationally analyzes the three-dimensional structure of phosphorylated proteins and then classifies and clusters the three-dimensional structure to identify phosphorylated protein biomarkers or drug targets associated with diseases.

[0032] Specifically, the present invention replaces phosphorylated serine (pS) and phosphorylated threonine (pT) with negatively charged amino acids, such as aspartic acid and glutamic acid, respectively, by using a phosphorylation simulation method, thereby converting the phosphorylated protein into an amino acid sequence that can be analyzed using an AI model of the prior art, such as AlphaFold2 or Rosset TFAold algorithm, and then predicting the three-dimensional structure of the replaced protein, thereby obtaining a prediction result of the three-dimensional structure of the phosphorylated protein whose structure is similar to that of the replaced protein. The present invention provides a simple, fast and relatively accurate prediction method for the prediction of the three-dimensional structure of phosphorylated proteins.

[0033] Furthermore, by comparing the three-dimensional structure of the phosphorylated protein with the three-dimensional structure of the non-phosphorylated protein, for example, through classification and cluster analysis, based on the comparison results, phosphorylated proteins with large structural differences before and after phosphorylation are selected to form a candidate library, and then combined with relevant information of the disease, phosphorylated proteins associated with the disease are identified as biomarkers or drug targets.

[0034] Methods for solving problems

[0035] The present invention provides the following technical solutions.

[0036] <1> A method for analyzing phosphorylated proteins based on computation, comprising:

[0037] Phosphorylated protein sequence prediction step: using a computational model to predict the probability of existence of each phosphorylation combination of the target protein, and selecting phosphorylated proteins whose probability of existence is greater than or equal to the specified probability;

[0038] Phosphorylated protein three-dimensional structure prediction step: replacing the phosphorylated serine and phosphorylated threonine in the phosphorylated protein sequence obtained by the phosphorylated protein sequence prediction step with aspartic acid and glutamic acid, respectively, and using a computational model to predict the three-dimensional structure of the replaced protein; and

[0039] Classification and clustering steps: classify and cluster the phosphorylated proteins according to the three-dimensional structure obtained in the phosphorylated protein three-dimensional structure prediction step.

[0040] <2> The analysis method according to <1>, wherein the phosphorylated protein sequence prediction step comprises:

[0041] List all possible phosphorylation combinations containing phosphorylation sites;

[0042] Predicting the probability of the presence of a phosphorylated protein by calculation; and

[0043] Select phosphorylated proteins whose presence probability is greater than or equal to the specified probability.

[0044] <3> The analysis method according to <1> or <2>, wherein the computational model used in the phosphorylated protein sequence prediction step is the DeepPhos algorithm.

[0045] <4> The analysis method according to any one of <1> to <3>, wherein the computational model used in the phosphorylated protein three-dimensional structure prediction step is RosseTTAFold algorithm or AlphaFold2 algorithm.

[0046] <5> The analysis method according to any one of <1> to <3>, wherein the computational model used in the phosphorylated protein three-dimensional structure prediction step is a combination algorithm of RosseTTAFold algorithm and AlphaFold2 algorithm.

[0047] <6> The analysis method according to any one of <1> to <5>, wherein the method of classifying and clustering the phosphorylated proteins in the classifying and clustering step is a hierarchical clustering method.

[0048] <7> The analysis method according to any one of <1> to <6>, wherein each category obtained in the classification and clustering steps is used as a possible candidate phosphoprotein group to generate a candidate phosphoprotein list.

[0049] <8> The analysis method according to any one of <1> to <7>, wherein the predetermined probability is 90%.

[0050] <9> A method for screening a phosphorylated protein as a biomarker, comprising:

[0051] The three-dimensional structure of the candidate phosphorylated protein obtained by the analysis method described in any one of <1> to <7> is compared with the three-dimensional structure of the non-phosphorylated protein, and the phosphorylated protein with a large structural difference is selected as a biomarker.

[0052] <10> A method for screening a phosphorylated protein as a drug target, comprising:

[0053] The three-dimensional structure of the candidate phosphorylated protein obtained by the analysis method described in any one of <1> to <7> is compared with the three-dimensional structure of the non-phosphorylated protein, and the phosphorylated protein with a large structural difference is selected as the drug target.

[0054] Effects of the Invention

[0055] Through the present invention, it is possible to comprehensively, quickly and inexpensively predict the three-dimensional structure of phosphorylated proteins without relying on experiments but only relying on the calculation of AI models, and then it is possible to screen out phosphorylated proteins that can serve as biomarkers or drug targets based on the obtained three-dimensional structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 The figure is a schematic diagram showing the steps of the phosphorylated protein analysis method of the present invention.

[0057] Figure 2 The diagram shows the structural comparison of phosphorylated serine (pS) and aspartic acid (D), and phosphorylated threonine (pT) and glutamic acid (E).

[0058] Figure 3 Schematic diagram of the steps of analyzing phosphorylated Tau protein using the method of the present invention in Example 1.

[0059] Figure 4 Schematic diagram of the prediction steps and prediction results of the phosphorylated Tau protein sequence with high probability of existence in Example 1, wherein: Figure 4 (a) is a schematic diagram of the prediction steps. Figure 4 (b) is a schematic diagram of the prediction results.

[0060] Figure 5 The schematic diagram of the prediction steps and prediction results of the three-dimensional structure of the phosphorylated Tau protein in Example 1, wherein: Figure 5 (a) is a schematic diagram of the prediction steps. Figure 5 (b) is a schematic diagram of the prediction results.

[0061] Figure 6 The figure shows the experimental results of the three-dimensional structure of the Tau protein fragment in Example 1, and a visual comparison of the AlphaFold2 prediction results before and after amino acid replacement using the method of the present invention.

[0062] Figure 7 The schematic diagram of the clustering steps and clustering results of classifying and clustering the three-dimensional structure of phosphorylated Tau protein in Example 1, wherein: Figure 7 (a) is a schematic diagram of clustering steps. Figure 7 (b) is a schematic diagram of the clustering results.

[0063] FIG8 is an excel sheet showing relevant data comparing the predicted results and experimental results of the 11 phosphorylated Tau protein sequences before and after the amino acid substitution in Example 2. DETAILED DESCRIPTION

[0064] The following is a detailed description of the specific embodiments of the present invention. It should be noted that the following description of the embodiments is only for the purpose of illustrating the inventive concept of the present invention, and is not intended to limit the present invention.

[0065] <Analysis Method of Phosphorylated Protein>

[0066] One embodiment of the present invention relates to a method for analyzing phosphorylated proteins based on calculation (hereinafter, sometimes referred to as "the method of the present invention"), which comprises:

[0067] Phosphorylated protein sequence prediction step: using a computational model to predict the probability of existence of each phosphorylation combination of the target protein, and selecting phosphorylated proteins whose probability of existence is greater than or equal to the specified probability;

[0068] Phosphorylated protein three-dimensional structure prediction step: replacing the phosphorylated serine and phosphorylated threonine in the phosphorylated protein sequence obtained by the phosphorylated protein sequence prediction step with aspartic acid and glutamic acid, respectively, and using a computational model to predict the three-dimensional structure of the replaced protein; and

[0069] Classification and clustering steps: classify and cluster the phosphorylated proteins according to the three-dimensional structure obtained in the phosphorylated protein three-dimensional structure prediction step.

[0070] Figure 1 Schematic diagram showing the steps of the method for analyzing phosphorylated protein biomarkers of the present invention. Figure 1 As shown, the phosphorylated protein analysis method of the present invention includes a step of selecting a target protein (10), a step of predicting a phosphorylated protein sequence (20), a step of predicting a phosphorylated protein three-dimensional structure (30), and a step of classification and clustering (40).

[0071] (Step (10) of selecting target protein)

[0072] The target protein is not particularly limited and may be any protein to be studied, such as various kinases and phosphatases associated with diseases, and proteins that may undergo abnormal phosphorylation mutations.

[0073] (Prediction step of phosphorylated protein sequence (20))

[0074] In the step (20) of predicting the phosphorylated protein sequence, the computational model is used to predict the probability of existence of each phosphorylation combination of the target protein, and phosphorylated proteins whose probability of existence is greater than or equal to a specified probability are selected.

[0075] Regarding the phosphorylation combination, it refers to the combination of all phosphorylated proteins when one, two or more of all possible phosphorylated serine and threonine residue sites in the amino acid sequence of the protein are phosphorylated. For example, Tau protein itself is a phosphate-containing protein. Studies have shown that Tau protein molecules in normal mature brains contain 2 to 3 phosphate groups, while Tau proteins in the brains of patients with Alzheimer's disease (senile dementia) are abnormally overphosphorylated, and each molecule of Tau protein may contain 5 to 9 phosphate groups. Therefore, for Tau protein, for example, the combination of all phosphorylated proteins containing 2, 3, 4, 5, 6, 7, 8 and 9 serine and threonine residue sites respectively can be used as research objects to analyze the probability of the existence of each phosphorylated protein.

[0076] As for the computational model used in the phosphorylated protein sequence prediction step (20), the DeepPhos algorithm described in Non-Patent Document 6 may be used, but is not limited thereto. For example, the methods described in Patent Documents 6 to 10 may also be used.

[0077] DeepPhos is a novel multi-layer CNN (Convolutional Neural Networks) architecture that can accurately predict phosphorylation sites with protein sequence information. Unlike previous deep learning methods, DeepPhos consists of densely connected convolutional neural network (DC-CNN) blocks, in which the convolutional layers are simultaneously interconnected through intra-block cascade layers (Intra-BCL), effectively enhancing the flow of phosphorylation information. At the same time, multiple DC-CNN blocks with different window and filter sizes are used to automatically capture important sequence representations of protein phosphorylation sites, and are further integrated through inter-block cascade layers (Inter-BCL) for final prediction. DeepPhos outperforms previous methods in general phosphorylation prediction and also has good performance in kinase-specific phosphorylation site prediction. Predictions using DeepPhos are not affected by group, family, subfamily, and individual kinase levels.

[0078] Regarding the existence probability, from the perspective of the model calculation amount and the expected number of candidate phosphorylated proteins, it is preferred to select phosphorylated proteins with an existence probability of 70% or more, for example, 75% or more, 80% or more, 85% or more, 90% or more, 95% or more, etc.

[0079] (Steps for predicting the three-dimensional structure of phosphorylated proteins (30))

[0080] In the step (30) of predicting the three-dimensional structure of the phosphorylated protein, the phosphorylated serine and phosphorylated threonine in the phosphorylated protein sequence with a high probability of existence obtained through the phosphorylated protein sequence prediction step are replaced with aspartic acid and glutamic acid, respectively, and then the three-dimensional structure of the replaced protein is predicted using a computational model.

[0081] The above method of replacing the amino acid at the phosphorylation site with an amino acid with a carboxyl side chain, thereby replacing the phosphorylated amino acid with a non-phosphorylated amino acid, is similar to phosphorylation simulation (Phosphomimetics) used in the experimental method. Figure 2 As shown in the figure, aspartic acid (D) and glutamic acid (E) are similar in structure to phosphorylated serine (pS) and phosphorylated threonine (pT), respectively, and can be substituted for each other in the experiment. The phosphorylation of serine and threonine is the phosphorylation of neutral amino acids serine and threonine into negatively charged amino acid groups, which makes the protein have a charge and changes in the three-dimensional structure. Therefore, the three-dimensional structure of the phosphorylated protein will be different from that of the protein before phosphorylation. The negatively charged amino acids aspartic acid or glutamic acid can be used to replace phosphoserine or phosphothreonine. The former can simulate the phosphorylation state of the protein (Pearlman, SM, Serber, Z. & Ferrell, JEA Mechanism for the Evolution of Phosphorylation Sites. Cell 147, 934–946 (2011).).

[0082] Regarding the computational model used in the step (30) of predicting the three-dimensional structure of the phosphorylated protein, for example, the RoseTTAFold algorithm or the AlphaFold2 algorithm can be used, but it is not limited to these two algorithms, and other algorithms used in the art for predicting the three-dimensional structure of proteins based on amino acid sequences can also be used. The RoseTTAFold algorithm, the AlphaFold2 algorithm, or other current algorithms cannot directly predict the three-dimensional structure of phosphorylated proteins based on amino acid sequences, but by using a phosphorylation simulation method, the phosphorylated amino acid residues are replaced with negatively charged amino acids, and the three-dimensional structure of the simulated protein after the replacement can be predicted by an AI model, thereby indirectly analyzing the three-dimensional structure of the phosphorylated protein.

[0083] RoseTTAFold uses the nlp (natural language processing) method to directly learn co-evolution information from MSA (multiple sequence alignment). The MSA sequence will be cropped according to the length of 260, because the sequence is too long and the attention operation in the network will increase the memory burden of the GPU. The pre-processing of the MSA feature includes: in the residue dimension, the 21 residues in the MSA (including alignment gaps) are embedded into a vector of a specified length, and then a sinusoidal positional encoding is added to provide position information; in the sequence dimension, the relationship information between the MSA and the sequence to be predicted (query sequence) is added. The information provided by the searched template is the 2D residue pair information of the alignment domain, including the distance and orientation between residue pairs. In addition, some indicators generated by the template through the HHSearch search will also be added to the 2D template information. The template information will be pre-processed by axial attention (template direction and residue direction), pixelattention, and the embedding and positional encoding information of the query sequence will be added.

[0084] Compared with RoseTTAFold, AlphaFold2's MSA will have more processing. According to the identity distance with the query sequence, the MSA sequence will be randomly deleted in blocks in the sequence direction. This is done to save computing memory and maintain the diversity of the sequence. Then, the remaining sequences will be clustered, and only the sequences at the cluster center will be retained as the MSA input of the model. The sequences not at the cluster center will be randomly extracted as extra MSA input networks. Here, AlphaFold2 also directly uses MSA to learn co-evolution information, but AlphaFold2 uses a mask strategy. Each position has a 15% probability of being masked, and then replaced according to a specific rule, and finally the recovery degree of the masked residue is observed. This is expected to further explore the co-evolution information and do embedding better. The impact of this loss can also be further reflected in the loss function it gives. AlphaFold2 also performs cropping on the sequence. In its embedding process, pair features (i.e. pair representation) are directly initialized by query sequence information, and then the distance information of the template is added with the help of pointwise Attention. MSA features (i.e., MSA representations) are obtained by jointly projecting the query sequence and MSA, and the angle information of the template is added. In addition, the position encoding in AlphaFold2 is a relative position encoding, which can better process long-range text signals.

[0085] Since AlphaFold2 and RoseTTAFold each have their own advantages, RoseTTAFold is faster in calculation, but AlphaFold2 is more accurate in calculation. Therefore, they can be combined in actual use. For example, Ian R. Humphreys et al. have made useful attempts in this regard (Ian R. Humphreys et al., Computed structures of core eukaryotic protein complexes, SCIENCE, 11 Nov 2021, Vol 374, Issue 6573; DOI: 10.1126 / science.abm4805). Of course, AlphaFold2 or RoseTTAFold can also be combined with other algorithms.

[0086] (Classification and clustering steps (40))

[0087] In the classification and clustering step (40), the phosphorylated proteins are classified and clustered according to the three-dimensional structure obtained in the phosphorylated protein three-dimensional structure prediction step.

[0088] Phosphorylated proteins can be classified and clustered by, for example, hierarchical clustering.

[0089] Hierarchical clustering is an unsupervised learning clustering method that divides data sets into different groups according to their similarity and forms a tree structure. Hierarchical clustering is divided into two methods: agglomerative clustering and divisive clustering. Agglomerative clustering starts from each data point and continuously combines similar data points to form larger groups until all data points are in the same group. Divisive clustering starts from all data points and continuously divides dissimilar data points into new groups until each group has only one data point. In hierarchical clustering, it is necessary to define a similarity measure. For example, methods such as Euclidean distance, Manhattan distance, and cosine similarity can be used to measure the similarity between data points. At the same time, it is necessary to select appropriate clustering methods and stop conditions to achieve the best clustering results. The advantages of hierarchical clustering include: it can generate a tree structure, which is convenient for the visualization and interpretation of the data set; it can automatically identify the groups in the data set without presetting the number of clusters; it is suitable for small-scale data sets and data with hierarchical structures. In addition, hierarchical clustering can be combined with other clustering methods, such as K-means, DBSCAN, etc., to obtain better clustering results.

[0090] Each category obtained by classification and clustering can be used as a possible candidate phosphorylated protein group to generate a candidate phosphorylated protein list.

[0091] <Method for screening phosphorylated proteins as biomarkers / drug targets>

[0092] Another embodiment of the present invention relates to a method for screening phosphorylated proteins as biomarkers / drug targets, which comprises comparing the three-dimensional structure of the candidate phosphorylated protein obtained by the phosphorylated protein analysis method of the present invention with the three-dimensional structure of the protein that is not phosphorylated, and selecting phosphorylated proteins with large structural differences as biomarkers or drug targets.

[0093] When necessary, the selection can be assisted by combining existing research results related to the disease.

[0094] In the following examples, Tau protein is taken as an example to specifically illustrate the analysis method of phosphorylated protein of the present invention.

[0095] The microtubule system is a component of the neural cell skeleton and can participate in a variety of cell functions. Microtubules are composed of tubulin and microtubule-associated proteins, and Tau protein is the microtubule-associated protein with the highest content. The cellular function of Tau protein in the normal brain is to bind to tubulin to promote its polymerization to form microtubules; bind to the formed microtubules, maintain the stability of microtubules, reduce the dissociation of tubulin molecules, and induce microtubule bundling. In normal people, due to different editing methods of Tau protein mRNA, 6 isomers can be expressed. Tau protein is a phosphate-containing protein, and the Tau protein molecule in the normal mature brain contains 2 to 3 phosphate groups. However, the Tau protein in the brain of patients with Alzheimer's disease (AD) is abnormally overphosphorylated, and each molecule of Tau protein can contain 5 to 9 phosphate groups and lose its normal biological function.

[0096] After abnormal hyperphosphorylation, the binding force of Tau protein to tubulin in the brain of AD patients is only 1 / 10 of that of normal Tau protein, and it also loses its biological function of promoting microtubule assembly and formation and loses its role in maintaining microtubule stability. PHF-Tau (paired helical filament-Tau) competes with tubulin to bind to normal Tau protein and other macromolecular microtubule-related proteins, and seizes these proteins from microtubules, causing microtubule depolymerization and destroying the normal microtubule system. Abnormally phosphorylated Tau protein aggregates itself into PHF / NFT (nerve fiber entanglement) structure. The microtubule structure of affected neurons in the brain of AD patients is widely destroyed, normal axonal transport is impaired, synaptic loss and neuronal function damage occur, and brain neurodegeneration occurs.

[0097] It should be noted that based on the inventive concept of clustering analysis of the structure prediction of modified proteins to discover phosphorylated proteins with biological significance, those skilled in the art can further expand the method of the present invention to other types of post-translational modifications of proteins, such as glycosylation modification and acetylation modification of proteins. For example, a method for predicting the three-dimensional structure of glycosylated proteins can be provided, and different glycosylated proteins can be classified and clustered according to the predicted structure, and then each clustering result can be used as a candidate biomarker, thereby discovering new biomarkers.

[0098] Above, the analytical method of phosphorylated protein of the present invention has been described based on the embodiments, but the present invention is not limited to these. As long as it does not depart from the gist of the present invention, the various deformations that can be thought of by those skilled in the art are implemented in the embodiments, or other methods constructed by combining a part of the constituent elements in the embodiments are also included in the scope of the present invention.

[0099] Example

[0100] Below, the analysis method of phosphorylated protein of the present invention is specifically described by taking a specific protein (Tau protein) sequence as an example. It should be noted that the present invention is not limited to these embodiments, and can also be used for the analysis of post-translational modification of other proteins.

[0101] Example 1: Prediction of phosphorylated Tau protein 1

[0102] Figure 3 Schematic diagram of the steps of analyzing phosphorylated Tau protein using the method of the present invention in Example 1.

[0103] like Figure 3 As shown, the analysis method of phosphorylated Tau protein specifically includes the following steps: a step of selecting Tau protein (100), a step of predicting a phosphorylated Tau protein sequence with a high probability of existence (200), a step of predicting the three-dimensional structure of the phosphorylated Tau protein based on the phosphorylated Tau protein sequence with a high probability of existence (300), and a step of classifying and clustering the phosphorylated Tau protein based on the three-dimensional structure (400).

[0104] Figure 4 The figure is a schematic diagram of the steps of predicting the phosphorylated Tau protein sequence with a high probability of existence and the prediction results of Example 1, wherein: Figure 4 (a) is a schematic diagram of the prediction steps. Figure 4 (b) is a schematic diagram of the prediction results.

[0105] like Figure 4 As shown in (a), in the step (200) of predicting the phosphorylated Tau protein sequence with a high probability of existence, firstly, all possible phosphorylation combinations of less than 2 serine and threonine phosphorylation sites are listed (210), and then the probability of existence of each phosphorylated protein is predicted by the DeepPhos algorithm (220), and the phosphorylated Tau protein with an existence probability greater than or equal to 90% is selected (230). It should be noted that the above "less than 2" is only to illustrate the examples of phosphorylation combinations cited by the method of the present invention. As mentioned above, for Tau protein, all combinations of phosphorylated proteins containing 2, 3, 4, 5, 6, 7, 8 and 9 serine and threonine residues that are phosphorylated can be used as research objects to analyze the probability of existence of each phosphorylated protein.

[0106] like Figure 4As shown in (b), all possible phosphorylation combinations of one phosphorylation site and two phosphorylation sites in Tau protein are listed, and then phosphorylated Tau proteins with a probability less than a specified probability (e.g., 90%) such as pTau-205 are excluded, and all phosphorylated Tau proteins with a probability greater than or equal to 90% such as pTau-195, pTau-202, and pTau-202 / 205 are retained. Among them, pTau-195, pTau-202, and pTau-202 / 205 represent proteins phosphorylated at position 195 of Tau protein, proteins phosphorylated at position 202 of Tau protein, and proteins phosphorylated at positions 202 and 205 of Tau protein, respectively.

[0107] Figure 5 The schematic diagram of the prediction steps and prediction results of the three-dimensional structure of the phosphorylated Tau protein in Example 1, wherein: Figure 5 (a) is a schematic diagram of the prediction steps. Figure 5 (b) is a schematic diagram of the prediction results.

[0108] like Figure 5 As shown in (a), in the step (300) of predicting the three-dimensional structure of phosphorylated Tau protein, phosphorylation simulation is first performed to replace phosphorylated serine and phosphorylated threonine with aspartic acid (D) and glutamic acid (E), respectively (310), and then the RosseTTAFold or AlphaFold2 algorithm is used to predict the three-dimensional structure of Tau protein (320).

[0109] like Figure 5 As shown in (b), the phosphorylated serine (pS) in the phosphorylated protein is replaced by aspartic acid (D), and the phosphorylated threonine (pT) is replaced by glutamic acid (E), and then the AlphaFold2 algorithm is used to predict the three-dimensional structure of the replaced protein.

[0110] In order to verify whether the three-dimensional structure predicted by the method of the present invention is accurate, it is compared with the three-dimensional structure obtained by experiment and provided in RCSB PDB (Worldwide Protein Data Bank, wwPDB, www.rcsb.org).

[0111] Figure 6 A diagram showing the experimental results of the three-dimensional structure of the Tau protein fragment of Example 1 (https: / / www.rcsb.org / structure / 5e2v) and a visual comparison of the AlphaFold2 prediction results before and after amino acid replacement using the method of the present invention.

[0112] from Figure 6It can be seen that the predicted structures are significantly different before and after amino acid substitution, and the predicted structure after substitution is closer to the experimental three-dimensional structure.

[0113] It should be noted that, since there is a coordinate diagram on the 3D viewer, in order to more accurately compare the three-dimensional structure, the protein fragment structures at the xy plane, xz plane, yz plane and another random angle were selected for comparison. From the results, at some angles, the predicted results after replacement are closer to the experimental results, and the predicted results before and after replacement are significantly different; at some angles, the structures of the three are not very similar, but overall, the predicted structure after replacement is closer to the experimental results than before replacement.

[0114] Figure 7 The schematic diagram of the clustering steps and clustering results of classifying and clustering the three-dimensional structure of phosphorylated Tau protein in Example 1, wherein: Figure 7 (a) is a schematic diagram of clustering steps. Figure 7 (b) is a schematic diagram of the clustering results.

[0115] like Figure 7 As shown in (a), in the classification and clustering step (400), the three-dimensional structural features of the phosphorylated Tau protein are first extracted (410), and the phosphorylated Tau protein is clustered according to the predicted three-dimensional structural features of the phosphorylated Tau protein. Specifically, the phosphorylated Tau protein is classified and clustered according to the similarity of the structural features using the Hierarchical K-means algorithm (420), and then each category is used as a possible candidate biomarker to generate a candidate biomarker list (430).

[0116] like Figure 7 As shown in (b), pTau-195 and pTau-202 / 205 have similar structures and can be divided into one category, and pTau-202 can be divided into another category. Their structures are significantly different from those of the Tau proteins before phosphorylation. By analyzing the structural differences with the phosphorylated Tau proteins before phosphorylation with known structures, the phosphorylated proteins with large structural differences among pTau-195, pTau-202 / 205, and pTau-202 are calculated as candidate biomarkers for the diagnosis or treatment of Alzheimer's disease.

[0117] Example 2: Prediction of phosphorylated proteins 2

[0118] In order to further verify the reliability of the method of the present invention, 11 phosphorylated protein sequences (sequence numbers 1 to 11) were randomly selected from RCSB PDB for further research. The phosphorylation-related information of the 11 phosphorylated protein sequences is shown in Table 1.

[0119] Table 1

[0120]

[0121] In Table 1, “SEP” indicates phosphorylated serine, “TPO” indicates phosphorylated threonine, and “ptm” indicates post-translational modifications, including phosphorylation, glycosylation, ubiquitination, nitrosylation, methylation, acetylation, lipidation, and proteolysis.

[0122] The phosphorylated serine and phosphorylated threonine in the 11 phosphorylated protein sequences were replaced with aspartic acid and glutamate, respectively, and the three-dimensional structure of the amino acid sequence obtained after replacement was predicted by AphaFold2 (hereinafter also referred to as the "prediction result") and compared with the three-dimensional structure of the phosphorylated protein obtained by experiment provided in the database (hereinafter also referred to as the "experimental result").

[0123] Figure 8 Figure 8-1 to Figure 8-4 ) is an excel sheet showing the relevant data of the comparison between the prediction results and the experimental results of the 11 phosphorylated Tau protein sequences before and after the amino acid substitution in Example 2. Among them, "u" represents the calculation result of the predicted structure after the amino acid substitution, and "o" represents the calculation result of the predicted structure before the amino acid substitution, which is to predict the phosphorylated residues as unphosphorylated residues, for example, phosphorylated serine and phosphorylated threonine are directly regarded as serine and threonine for structural prediction.

[0124] In AphaFold2 prediction, for each sequence, the results of 5 models (rank 0 to rank 4) are output. The experimental results are used as the gold standard. After alignment, RMSD (Root Mean Square Deviation) is calculated. The smaller the RMSD value, the closer the prediction result is to the experimental result. "Minimum RMSD, u≤o" means judging by the minimum RMSD whether the prediction result after replacement is closer to the experimental result than the prediction result before replacement. "1" means "yes" and "0" means "no". Similarly, "Average RMSD, u≤o" means judging by the average RMSD whether the prediction result after replacement is closer to the experimental result than the prediction result before replacement. "1" means "yes" and "0" means "no".

[0125] As can be seen from Figure 8, judging by the minimum RMSD, the prediction results of 8 out of 11 proteins after replacement are closer to the experimental results than the prediction results before replacement; judging by the average RMSD, the prediction results of 7 out of 11 proteins after replacement are closer to the experimental results than the prediction results before replacement.

[0126] As shown in Table 8, the method of the present invention is a feasible and reliable prediction method for the three-dimensional structure of phosphorylated proteins, which can provide useful reference information for research on the structure and function of proteins and can greatly reduce the workload of experiments.

[0127] Industrial Applicability

[0128] As described above, the present invention can comprehensively, quickly and inexpensively predict the three-dimensional structure of phosphorylated proteins without relying on experiments but only relying on the calculation of AI models, and then screen phosphorylated proteins that can serve as biomarkers or drug targets based on the obtained three-dimensional structure.

Claims

1. A method for analyzing phosphorylated proteins based on calculation, comprising: Phosphorylated protein sequence prediction step: using a computational model to predict the probability of existence of each phosphorylation combination of the target protein, and selecting phosphorylated proteins whose probability of existence is greater than or equal to the specified probability; Phosphorylated protein three-dimensional structure prediction step: replacing the phosphorylated serine and phosphorylated threonine in the phosphorylated protein sequence obtained by the phosphorylated protein sequence prediction step with aspartic acid and glutamic acid, respectively, and using a computational model to predict the three-dimensional structure of the replaced protein; and Classification and clustering steps: classify and cluster the phosphorylated proteins according to the three-dimensional structure obtained in the phosphorylated protein three-dimensional structure prediction step.

2. The analysis method according to claim 1, wherein The phosphorylated protein sequence prediction step comprises: List all possible phosphorylation combinations containing phosphorylation sites; Predicting the probability of the presence of a phosphorylated protein by calculation; and Select phosphorylated proteins whose presence probability is greater than or equal to the specified probability.

3. The analysis method according to claim 1 or 2, wherein: The computational model used in the phosphorylated protein sequence prediction step is the DeepPhos algorithm.

4. The analysis method according to any one of claims 1 to 3, wherein The computational model used in the phosphorylated protein three-dimensional structure prediction step is the RosseTTAFold algorithm or the AlphaFold2 algorithm.

5. The analysis method according to any one of claims 1 to 3, wherein The computational model used in the phosphorylated protein three-dimensional structure prediction step is a combination of the RosseTTAFold algorithm and the AlphaFold2 algorithm.

6. The analysis method according to any one of claims 1 to 5, wherein The method for classifying and clustering phosphorylated proteins in the classification and clustering steps is a hierarchical clustering method.

7. The analysis method according to any one of claims 1 to 6, wherein Each category obtained in the classification and clustering steps is used as a possible candidate phosphorylated protein group to generate a candidate phosphorylated protein list.

8. The analysis method according to any one of claims 1 to 7, wherein The specified probability is 90%.

9. A method for screening phosphorylated proteins as biomarkers, comprising: The three-dimensional structure of the candidate phosphorylated protein obtained by the analysis method according to any one of claims 1 to 7 is compared with the three-dimensional structure of the non-phosphorylated protein, and the phosphorylated protein with a large structural difference is selected as a biomarker.

10. A method for screening phosphorylated proteins as drug targets, comprising: The three-dimensional structure of the candidate phosphorylated protein obtained by the analysis method according to any one of claims 1 to 7 is compared with the three-dimensional structure of the non-phosphorylated protein, and the phosphorylated protein with a large structural difference is selected as the drug target.

Images