A protein pocket matching method, system and application based on protein structure fingerprints
By constructing protein structure fingerprints at the amino acid scale, the problem of insufficient protein pocket similarity calculation in traditional methods is solved, achieving more accurate protein pocket matching and improving the efficiency of novel drug design and development.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GALIXIR BIOTECHNOLOGY (SHANGHAI) LTD
- Filing Date
- 2022-12-27
- Publication Date
- 2026-07-14
AI Technical Summary
Existing technologies struggle to effectively construct underlying connections based on protein structures, resulting in insufficient model generalization ability and an inability to fully utilize existing data. This is particularly true in the discovery of lead compounds and skeletal transitions in FIC proteins, where traditional methods cannot accurately match the interactions between protein pockets and small molecules.
By constructing protein structural fingerprints at the amino acid scale, calculating the structural fingerprint of each amino acid, and using a hash algorithm to generate fingerprint matching for protein pockets, the underlying connections of protein structures are established, and new lead compounds or skeletal transition molecular fragments are discovered.
It improves the efficiency of novel drug design and development, shortens the R&D cycle, enhances the model screening and target protein binding ability, and improves the accuracy of protein pocket similarity calculation.
Smart Images
Figure CN116246695B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of biogenetic technology, and in particular relates to a protein pocket matching method, system and application based on protein structure fingerprinting. Background Technology
[0002] In structure-based drug discovery and design, a crucial aspect is identifying or designing small molecules that strongly interact with target proteins. These small molecules can influence the function of target proteins through binding, ultimately leading to therapeutic effects. However, the number of known small molecules interacting with target proteins is typically very limited, especially for FIC proteins, where there may be no known active compounds. This necessitates building deeper connections by leveraging information from other proteins related to the target protein to construct a generalized model. The collected results of interactions between other proteins and small molecules can then be used to optimize and design molecular drugs targeting new protein groups. Existing techniques typically rely on protein sequence similarity to establish connections between proteins. However, the connection between protein sequence and function is far less robust than that between protein structure and function. This is evident in the conclusions of some studies: many proteins with different sequences have similar three-dimensional structures and functions; and in some cases, proteins with similar sequences exhibit significantly different protein-small molecule interaction patterns. Existing structure-based similarity comparison methods treat proteins as a whole to calculate features, but typically, it is the protein's substructures (often called protein pockets) that interact with small molecules. Some proteins have dissimilar sequences and overall structures, but share similar protein pockets. Existing similarity calculation methods struggle to establish underlying connections between these proteins, leading to underutilization of existing data and weak model generalization ability.
[0003] Therefore, there is indeed a need to propose better solutions for the aforementioned existing technologies. Summary of the Invention
[0004] The purpose of this invention is to provide a protein pocket matching method, system, and application based on protein structural fingerprinting. By constructing protein structural fingerprints, this method identifies protein pockets similar to the binding pockets of target proteins, enhancing the model's ability to screen molecules capable of binding to the target protein. Specifically, by constructing amino acid-scale fingerprints and calculating the structural fingerprint of each amino acid, similar pockets are found through pooling protein pocket fingerprint matching. This achieves the goal of matching similar protein pockets, establishing underlying connections in protein structures, and discovering new lead compounds or replaceable molecular fragments for skeletal transitions. In particular, the molecular fragments crystallized from these matched pockets can be applied to the discovery of lead compounds and skeletal transitions in FIC proteins.
[0005] This invention provides a protein pocket matching method based on protein structure fingerprinting, comprising:
[0006] S1, Generate the protein pocket structure fingerprint corresponding to the target site;
[0007] S2, Establish protein structure fingerprint and fragment database;
[0008] S3, similarity calculation based on the protein pocket structure fingerprint of the target point and the protein pocket set in the protein structure fingerprint and fragment database;
[0009] S4, protein pocket matching is performed based on the similarity score.
[0010] Preferably, S1 includes:
[0011] S11, determine the pocket amino acid structure fingerprint of each amino acid in the protein pocket corresponding to the target site;
[0012] S12, obtaining protein pocket structure fingerprint based on pocket amino acid structure fingerprint, including: pooling all pocket amino acid structure fingerprints in the protein pocket to obtain protein pocket structure fingerprint.
[0013] Preferably, the method for determining the pocket amino acid structure fingerprint in S11 includes:
[0014] (1) Set the threshold distance between amino acid neighbors;
[0015] (2) Calculate the pocket amino acid fingerprint based on the threshold distance, the third-order neighbor features of each amino acid, and the hash algorithm;
[0016] The hash algorithm includes two methods:
[0017] The first method is to use the perceptual hash algorithm to automatically identify similar amino acids by comparing the similarity of images based on the different characteristics of amino acids. The perceptual hash algorithm extracts the hash code of each image to be tested and compares it with the hash codes of the labeled amino acids. The label represented by the hash code with the highest similarity is found as the pocket amino acid structure fingerprint.
[0018] The second approach is to establish an empirical hash table, using the number of hash tables L, the number of bits in the hash table k, and the number of global alignment vectors Q to determine the hash code of a certain amino acid as a pocket amino acid structure fingerprint.
[0019] Preferably, S2 includes:
[0020] S21, Obtain the protein-molecule cocrystal structure database and extract all protein pockets from the protein-molecule cocrystal structure database;
[0021] S22, for each protein pocket p, take multiple spheres of radius n and extract small molecule fragments m from the spheres, where n is a first variable threshold set according to the similarity calculation requirements; if none of the small molecule fragments m contain the specified molecular fragment, discard the corresponding protein pocket; otherwise, retain the corresponding protein pocket, and finally obtain the protein pocket set P = {p1, p2, ... p}. l .} and the corresponding set of molecular fragments M = {m1, m2, ... m l .}, where l is the number of database protein pockets or the number of effective molecular fragments.
[0022] Preferably, the similarity calculation method in S3 includes:
[0023] The similarity calculation for the protein pocket structure fingerprint of the target site is shown below.
[0024] Similarity=Jaccard(FP q FP t (1)
[0025] Among them, FP q ,FP t These are the protein pocket fingerprints to be queried, the protein structure fingerprints, and the protein pocket fingerprints in the fragment database, respectively. The similarity is calculated using the Jaccard vector distance.
[0026] Preferably, S4 includes:
[0027] S41, sort the l similarity values from largest to smallest;
[0028] S42, according to the rule of similarity value from large to small, obtain x protein pockets and corresponding small molecule fragment sets corresponding to the similarity values in the protein structure fingerprint and fragment database as the matching set C; where x is the second variable threshold defined by the application.
[0029] A second aspect of the present invention provides a protein pocket matching system based on protein structure fingerprinting, comprising:
[0030] The protein pocket structure fingerprint generation module (101) is used to generate the protein pocket structure fingerprint corresponding to the target site.
[0031] The database establishment module (102) is used to establish a protein structure fingerprint and fragment database;
[0032] The similarity calculation module (103) is used to calculate the similarity between the protein pocket structure fingerprint of the target point and the protein pocket set in the protein structure fingerprint and fragment database;
[0033] Protein pocket matching module (104) is used to perform protein pocket matching based on the magnitude of similarity values.
[0034] A third aspect of the present invention is to provide an application of a protein pocket matching method based on protein structure fingerprinting in the discovery of seed compounds based on FIC proteins.
[0035] A fourth aspect of the present invention is to provide a protein pocket matching method based on protein structure fingerprinting for the discovery of replaceable molecular fragments for backbone transitions.
[0036] A fourth aspect of the present invention is to provide an application of a protein pocket matching method based on protein structure fingerprinting in molecular drugs.
[0037] The system, method, and application provided by this invention have the following beneficial technical effects:
[0038] This invention innovatively proposes a hash fingerprint generation method based on protein structure and amino acid scale, thereby solving the problem of protein pocket similarity calculation and pocket-based lead compound discovery and scaffold transition work. It overcomes the limitations of traditional protein similarity calculation methods based on sequence or whole protein structure, which have weak functional correlations. By calculating protein similarity at the protein pocket scale, which is closer to function, to construct underlying protein connections, it improves the efficiency of novel drug design and development, shortens the R&D cycle, and has high practical value. Attached Figure Description
[0039] Figure 1 This is a schematic flowchart illustrating a protein pocket matching method based on protein structure fingerprinting according to a preferred embodiment of the present invention;
[0040] Figure 2 This is a schematic diagram illustrating the protein pocket similarity calculation and application process according to a preferred embodiment of the present invention, wherein:
[0041] Figure 2 (a) is a schematic diagram of a protein structure according to a preferred embodiment of the present invention;
[0042] Figure 2 (b) is a schematic diagram of a pocket amino acid fingerprint according to a preferred embodiment of the present invention;
[0043] Figure 2 (c) is a schematic diagram of a pocket pooled fingerprint according to a preferred embodiment of the present invention;
[0044] Figure 2 (d) is a schematic diagram of a target pocket fingerprint according to a preferred embodiment of the present invention;
[0045] Figure 2(e) is a schematic diagram of a fingerprint database according to a preferred embodiment of the present invention;
[0046] Figure 2 (f) is a schematic diagram of the matched corresponding molecular fragments according to a preferred embodiment of the present invention.
[0047] Figure 3 This is a schematic diagram illustrating the pocket fingerprinting process according to a preferred embodiment of the present invention;
[0048] Figure 4 This is a schematic diagram of a protein pocket matching system architecture based on protein structure fingerprinting according to a preferred embodiment of the present invention. Detailed Implementation
[0049] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.
[0050] Example 1
[0051] See Figure 1 This embodiment provides a reference. Figure 1 This invention provides a protein pocket matching method based on protein structure fingerprinting, comprising:
[0052] S1, Generate the protein pocket structure fingerprint corresponding to the target site, including:
[0053] See Figure 2 (a) and Figure 2 (b), S11, determine the pocket amino acid structure fingerprint of each amino acid in the protein pocket corresponding to the target site;
[0054] In this embodiment, the method for determining the pocket amino acid structure fingerprint includes:
[0055] (1) Set the threshold distance between amino acid neighbors;
[0056] (2) Figure 3 As shown, the pocket amino acid fingerprint is calculated based on the threshold distance, the third-order neighbor features of each amino acid, and a hash algorithm. The pocket amino acid fingerprint includes a first-layer fingerprint and a second-layer fingerprint. Specifically, amino acids, as the basic building blocks of proteins, are characterized by each amino acid molecule containing at least one amino group (-NH2) and one carboxyl group (-COOH), and both amino and carboxyl groups are attached to the same carbon atom. Different R groups result in different amino acid types. Therefore, R groups, amino groups, and carboxyl groups constitute the third-order neighbor features. The principle of the hash algorithm is to map a message M of arbitrary length to a shorter and fixed-length value H(M). In this embodiment, the hash algorithm includes two methods:
[0057] The first method is to use the perceptual hash algorithm to automatically identify similar amino acids by comparing the similarity of images based on the different characteristics of amino acids. The perceptual hash algorithm extracts the hash code of each image to be tested and compares it with the hash codes of the labeled amino acids. The label represented by the hash code with the highest similarity is found as the pocket amino acid structure fingerprint.
[0058] The second approach is to establish an empirical hash table, using the number of hash tables (L), the number of bits in the hash table (k), and the number of global alignment vectors (Q) to determine the hash code of a certain amino acid as a pocket amino acid structure fingerprint.
[0059] S12, Obtaining protein pocket structure fingerprints based on pocket amino acid structure fingerprints;
[0060] In this embodiment, S12 is implemented as follows: Figure 2 (b) and Figure 2 As shown in (c), it includes:
[0061] The protein pocket structure fingerprint is obtained by pooling all the pocket amino acid structure fingerprints in the protein pocket.
[0062] S2, Establish a protein structure fingerprint and fragment database, including:
[0063] S21, Obtain the protein-molecule cocrystal structure database and extract all protein pockets from the protein-molecule cocrystal structure database;
[0064] S22, for each protein pocket p, take multiple spheres of radius n and extract small molecule fragments m from the spheres, where n is a first variable threshold set according to the similarity calculation requirements; if none of the small molecule fragments m contain the specified molecular fragment, discard the corresponding protein pocket; otherwise, retain the corresponding protein pocket, and finally obtain the protein pocket set P = {p1, p2, ... p}. l .} and the corresponding set of molecular fragments M = {m1, m2, ... m l .}, where l is the number of database protein pockets or the number of effective molecular fragments.
[0065] See Figure 2 (d) and 2(e), S3, similarity calculation between protein pocket structure fingerprints of target points and protein pocket sets in protein structure fingerprint and fragment databases;
[0066] In this embodiment, the similarity calculation method includes:
[0067] The similarity calculation for the protein pocket structure fingerprint of the target site is shown below.
[0068] Similarity=Jaccard(FPq FP t (1)
[0069] Among them, FP q ,FP t These are the protein pocket fingerprints to be queried, the protein structure fingerprints, and the protein pocket fingerprints in the fragment database, respectively. The similarity is calculated using the Jaccard vector distance. In this embodiment, a total of l similarity values are obtained.
[0070] See Figure 2 (f), S4, protein pocket matching based on similarity scores, including:
[0071] S41, sort the l similarity values from largest to smallest;
[0072] S42, according to the rule of similarity value from large to small, obtain x protein pockets and corresponding small molecule fragment sets corresponding to the similarity values from the protein structure fingerprint and fragment database as the matching set C; where x is a second variable threshold defined by the application, and in this embodiment, the value is between 1 and 150. C serves as a supplementary set for the target protein for molecular design and optimization.
[0073] See Figure 4 Example 2
[0074] A protein pocket matching system based on protein structure fingerprinting is provided, comprising:
[0075] The protein pocket structure fingerprint generation module 101 is used to generate the protein pocket structure fingerprint corresponding to the target site.
[0076] Database creation module 102 is used to create a protein structure fingerprint and fragment database;
[0077] The similarity calculation module 103 is used to calculate the similarity between the protein pocket structure fingerprint of the target point and the protein pocket set in the protein structure fingerprint and fragment database.
[0078] Protein pocket matching module 104 is used to perform protein pocket matching based on the magnitude of similarity values.
[0079] Example 3
[0080] Application of a protein pocket matching method based on protein structure fingerprinting in the discovery of FIC-based protein-based lead compounds.
[0081] FIC is an abbreviation for FIRST-IN-CLASS, meaning first-time application of a particular type.
[0082] Specifically, for FIC proteins, their pockets can be fingerprinted first, thereby finding similar associated protein pockets and corresponding molecular fragments and small molecule data in existing databases, and carrying out the work of discovering and optimizing lead compounds.
[0083] Example 4
[0084] This paper presents an application of a protein pocket matching method based on protein structure fingerprinting in the discovery of replaceable molecular fragments for backbone transitions.
[0085] Example 5
[0086] This paper presents an application of a protein pocket matching method based on protein structure fingerprinting in molecular drugs.
[0087] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.
Claims
1. A protein pocket matching method based on protein structure fingerprinting, characterized in that, include: S1, Generate the protein pocket structure fingerprint corresponding to the target site; S2, Establish protein structure fingerprint and fragment database; S3, similarity calculation based on the protein pocket structure fingerprint of the target point and the protein pocket set in the protein structure fingerprint and fragment database; S4, protein pocket matching is performed based on the similarity scores; S1 includes: S11, determine the pocket amino acid structure fingerprint of each amino acid in the protein pocket corresponding to the target site; S12, obtaining protein pocket structure fingerprint based on pocket amino acid structure fingerprint, including: pooling all pocket amino acid structure fingerprints in the protein pocket to obtain protein pocket structure fingerprint; The method for determining the pocket amino acid structure fingerprint in S11 includes: (1) Set the threshold distance for amino acid neighbors; (2) Calculate the pocket amino acid fingerprint based on the threshold distance, the third-order neighbor features of each amino acid, and the hash algorithm; The hash algorithm includes two methods: The first method is to use the perceptual hash algorithm to automatically identify similar amino acids by comparing the similarity of images based on the different characteristics of amino acids. The perceptual hash algorithm extracts the hash code of each image to be tested and compares it with the hash codes of the labeled amino acids. The label represented by the hash code with the highest similarity is found as the pocket amino acid structure fingerprint. The second approach is to establish an empirical hash table and use the number of hash tables L, the number of bits in the hash table k, and the number of global alignment vectors Q to determine the hash code of a certain amino acid as a pocket amino acid structure fingerprint. S2 includes: S21, Obtain the protein-molecule cocrystal structure database and extract all protein pockets from the protein-molecule cocrystal structure database; S22, for each protein pocket p Take multiple spheres with radius n and extract small molecular fragments from inside the spheres. m Where n is the first variable threshold set according to the similarity calculation requirements; if all small molecule fragments m If none of the specified molecular fragments are found, the corresponding protein pocket is discarded; otherwise, the corresponding protein pocket is retained, resulting in a final protein pocket set P = {p1, p2, ... p...}. l .} and the corresponding set of molecular fragments M = {m1, m2, ... m l .},in l For the number of protein pockets or effective molecular fragments in the database and The similarity calculation method of S3 includes: The similarity calculation for the protein pocket structure fingerprint of the target site is shown below. （1） in, These are the protein pocket fingerprints to be queried, the protein structure fingerprints, and the protein pocket fingerprints in the fragment database, respectively. The similarity is calculated using the Jaccard vector distance. S4 includes: S41, l Sort the similarity scores from largest to smallest; S42, according to the rule of similarity value from large to small, obtain x protein pockets and corresponding small molecule fragment sets corresponding to the similarity values in the protein structure fingerprint and fragment database as the matching set C; where x is the second variable threshold defined by the application.
2. A protein pocket matching system based on protein structure fingerprinting, used to implement the method of claim 1, characterized in that, include: A protein pocket structure fingerprint generation module (101) is used to generate protein pocket structure fingerprints corresponding to the target site. The database creation module (102) is used to create a protein structure fingerprint and fragment database; The similarity calculation module (103) is used to calculate the similarity between the protein pocket structure fingerprint of the target point and the protein pocket set in the protein structure fingerprint and fragment database; The protein pocket matching module (104) is used to perform protein pocket matching based on the magnitude of the similarity value.
3. The application of a protein pocket matching method based on the protein structure fingerprint of claim 1 in the discovery of seed compounds based on FIC proteins.
4. An application of a protein pocket matching method based on the protein structure fingerprint of claim 1 in the discovery of replaceable molecular fragments for backbone transitions.
5. The application of a protein pocket matching method based on the protein structure fingerprint of claim 1 in the optimization of molecular drugs.