A method of automatically constructing a DFA graph for nucleic acid sequence design

Abstract

The present application relates to the field of computer-aided drug research and bioengineering technology, and particularly relates to a method for automatically constructing DFA graph for nucleic acid sequence design, comprising the following steps: A. reading a codon table of a given species to obtain codon sequences of each amino acid; B. constructing an initial DFA subgraph of amino acid residues; C. connecting the initial DFA subgraph of amino acid residues to form a DFA graph of RNA or DNA; D. compressing and optimizing the DFA graph of RNA or DNA, and merging and deleting redundant edges and vertices, so that the DFA graph has the least vertices and edges; the operation sequence of steps C and D can be interchanged. The method can support sequences of any length and any legal codon table, thereby expanding and applying the nucleic acid sequence design method for drugs and protein expression systems to different biological and protein expression systems, avoiding redundant calculation in the design process, and providing an effective and optimized solution in terms of calculation time for nucleic acid sequence design.

Description

Technical Field

[0001] This invention relates to the fields of computer-aided drug development and bioengineering, specifically to a method for automatically constructing DFA diagrams for nucleic acid sequence design (including RNA drugs and DNA protein expression systems). Background Technology

[0002] According to the central dogma, DNA is transcribed into messenger RNA (mRNA), which is then translated onto ribosomes to produce corresponding proteins, thereby performing biological functions. Each codon, consisting of three consecutive residues from the start codon to the stop codon, corresponds to one natural amino acid. mRNA drugs, as an emerging drug form, have become a research hotspot in drug development. In the mRNA sequence, the codon-amino acid correspondence is one-to-many, and different codon choices affect properties such as mRNA stability and translation efficiency. Therefore, drug design must simultaneously satisfy the requirements of both the chemical stability and translation efficiency of the mRNA molecule.

[0003] Furthermore, in bioengineering experiments and production processes, preparing high-quality, high-purity, and highly natural proteins is a crucial first step in many processes. Over the years, significant advancements in biotechnology have enabled the large-scale expression and isolation of recombinant proteins. However, for large-scale applications such as enzyme, antibody, or vaccine production, the required protein quantities are considerably high. Therefore, optimizing the sequence design of DNA vectors such as plasmids to improve protein expression efficiency is an important research and development step.

[0004] The DNA vectors and mRNA vaccines mentioned above share similar biological foundations, therefore they can use the same technical solutions for sequence design. For example, the invention patent with publication number CN114974428A, entitled "System and Method for Sequence Design," provides the technical cornerstone of sequence design, offering a method for linear complexity design based on the DFA diagram of RNA or DNA. Preliminary animal trials of the designed SARS-CoV-2 spike protein vaccine using this method yielded enhanced antigen protein expression levels and up to 20-fold increases in related neutralizing immunoglobulins, resulting in better efficacy. Besides designing human mRNA vaccines, this method can also be applied to the optimization of nucleic acid sequences on other biological protein expression system vectors. Different organisms, and even different protein expression systems within the same biological system, have different codon tables. For instance, the nuclear genome uses a different codon table than the mitochondrial and chloroplast genomes. In the standard codon table, AGA and AGG encode arginine, but in vertebrate mitochondria, these two codons are stop codons. The stop codon UGA in the standard codon table encodes tryptophan in vertebrate mitochondria (Note: The mitochondrial genome not only uses ribosomes and tRNA in the cytoplasm to translate proteins, but it also encodes its own ribosomes and tRNA, using a unique codon table to translate proteins). The codon differences between mitochondrial and nuclear genomes also vary in different organisms.

[0005] However, the aforementioned patented technical solutions, which manually encode the DFA subgraphs corresponding to the amino acids (including stop codons) of biological codons, cannot efficiently cover all possible situations when faced with such diverse codon tables across different species and the customized translation systems in synthetic biology, or the generated DFA graphs may not be in optimal form. Clearly, manual encoding is insufficient to address the problem of different codon tables in nature. Therefore, there is an urgent need to optimize existing nucleic acid sequence design methods to extend their application to different organisms and protein expression systems, in order to handle the diverse codon tables in different species and protein expression systems. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a method for automatically constructing DFA diagrams for nucleic acid sequence design. This method is based on the most primitive biological codon table and automatically generates an optimized DFA diagram. It can support sequences of arbitrary length and any valid codon table, thereby extending the nucleic acid sequence design method to different organisms and protein expression systems.

[0007] The objective of this invention is achieved through the following technical solution:

[0008] A method for automatically constructing DFA diagrams for nucleic acid sequence design is provided, including the following steps:

[0009] Step A: Read the codon table of a given species to obtain the sequence of all possible codons, including the stop codon, for each amino acid;

[0010] Step B: Constructing the initial DFA subgraph of amino acid residues. This process involves converting the codon sequence into multiple linear DFA subgraphs and connecting them in parallel.

[0011] Step C: Connect the initial DFA subgraphs of amino acid residues obtained in Step B to form a DFA graph of RNA or DNA.

[0012] The RNA or DNA DFA graphs obtained in steps D and C are compressed and optimized by merging and deleting redundant edges and vertices to make the DFA graph have the fewest vertices and edges.

[0013] In this process, the order of operations in steps C and D can be interchanged. That is, the initial DFA subgraph of amino acid residues obtained in step B is first compressed and optimized, and then the compressed and optimized DFA subgraph is linked to form a DFA graph of RNA or DNA.

[0014] In the above technical solution, step B involves constructing an initial DFA subgraph of amino acid residues, which is constructed as follows:

[0015] Step B1: Iterate through the codons for each amino acid in sequence, including the stop codon;

[0016] Step B2: For each codon, construct a linear DFA subgraph, where the attributes of each edge of the linear DFA subgraph are, in turn, the bases of the codon.

[0017] Step B3: Merge the starting vertices of all linear DFA subgraphs (e.g., ... Figure 1 The vertex numbered 0 in the middle) and the ending vertex (such as Figure 1 (Vertex numbered 3 in the middle), that is, the starting vertex is merged with the starting vertex, and the ending vertex is merged with the ending vertex.

[0018] In the above technical solution, the method for compressing and optimizing the DFA diagram of RNA or DNA or the initial DFA sub-diagram of amino acid residues in step C or step D is as follows:

[0019] Step D1: Starting from each vertex in the DFA graph or DFA subgraph, check if there are any pairs of identical attributes among the outgoing edges of that vertex. That is, iterate through the outgoing edges of each vertex and determine if there are identical attributes. If there are identical attributes, the following merging operation needs to be performed:

[0020] Delete the second edge with the same attribute, and change all subsequent edges of that edge to end the vertex connection with the first edge that has the same attribute;

[0021] Step D2: Repeat step D1 until no situation arises where the DFA diagram or DFA subgraph needs to be modified.

[0022] The beneficial effects of this invention are:

[0023] This invention discloses a method for automatically constructing DFA diagrams for nucleic acid sequence design. This method automatically constructs an important data structure (RNA or DNA → DFA diagram) in the nucleic acid sequence design of RNA drugs and DNA protein expression systems. The method includes step A: reading the codon table of a given species to obtain all optional codon sequences, including stop codons, for each amino acid; step B: constructing an initial DFA subgraph of amino acid residues, which involves converting the codon sequences into multiple linear DFA subgraphs and concatenating them in parallel; step C: connecting the initial DFA subgraphs of amino acid residues obtained in step B to form an RNA or DNA DFA diagram; and step D: compressing and optimizing the RNA or DNA DFA diagram obtained in step C, merging and deleting redundant edges and vertices to minimize the number of vertices and edges. The order of steps C and D can be interchanged; that is, the initial DFA subgraphs of amino acid residues obtained in step B can be compressed and optimized first, and then the compressed and optimized DFA subgraphs can be concatenated to form the RNA or DNA DFA diagram. Since the above process occurs before the computation of the codon-optimized dynamic programming algorithm, the compression and optimization of the DFA diagram can significantly accelerate the computation and design speed, thereby improving R&D efficiency. Therefore, compared to existing technologies that manually construct optimized RNA or DNA DFA diagrams, the method of this invention supports sequences of arbitrary length and any valid codon table, thus extending nucleic acid sequence design methods to different biological and protein expression systems. It is more flexible, avoids human-introduced errors, and eliminates redundant computations during the design process, providing a computationally efficient and optimized solution for RNA and DNA sequence design. Attached Figure Description

[0024] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0025] Figure 1 This refers to step B1 of a method for automatically constructing a DFA graph for nucleic acid sequence design, as described in Examples 1 and 2, where a linear DFA subgraph is constructed for each codon.

[0026] Figure 2 This is a comparison of the DFA diagrams of the polypeptide RNA or DNA in Example 3 before and after compression optimization. Detailed Implementation

[0027] Example 1.

[0028] This embodiment provides a method for automatically constructing DFA diagrams for nucleic acid sequence design, comprising the following steps:

[0029] Step A: Read the codon table of a given species to obtain the sequence of all possible codons, including the stop codon, for each amino acid.

[0030] Step B: Constructing the initial DFA subgraph of amino acid residues. This process involves converting the codon sequence into multiple linear DFA subgraphs and connecting them in parallel.

[0031] In this step, an initial DFA subgraph of amino acid residues is constructed, specifically as follows:

[0032] Step B1: Iterate through the codons for each amino acid in sequence, including the stop codon;

[0033] Step B2: For each codon, construct as follows Figure 1 The linear DFA subgraph shown has each edge whose attribute is a base of the codon.

[0034] Step B3: Merge the starting vertices of all linear DFA subgraphs (e.g., ... Figure 1 The vertex numbered 0 in the middle) and the ending vertex (such as Figure 1 (Vertex numbered 3 in the middle), that is, the starting vertex is merged with the starting vertex, and the ending vertex is merged with the ending vertex.

[0035] Step C: Connect the initial DFA subgraphs of amino acid residues obtained in step B to form a DFA graph of RNA or DNA.

[0036] Step D: Compress and optimize the DFA graph of RNA or DNA obtained in step C, merging and deleting redundant edges and vertices to make the DFA graph have the fewest vertices and edges.

[0037] In this step, the specific method for compressing and optimizing the obtained RNA or DNA DFA diagram is as follows:

[0038] Step D1: Starting from each vertex in the DFA graph, check if there are any pairs of identical attributes among the outgoing edges of that vertex. That is, iterate through the outgoing edges of each vertex and determine if there are identical attributes. If there are identical attributes, the following merge operation needs to be performed:

[0039] Delete the second edge with the same attribute, and change all subsequent edges of that edge to end the vertex connection with the first edge that has the same attribute;

[0040] Step D2: Repeat step D1 until no situation requiring modification of the DFA diagram is found.

[0041] Since the above process involves compression and optimization before the computation of the dynamic programming algorithm for codon optimization, it can greatly accelerate the computation and design speed, thereby improving R&D efficiency. The method of this invention can support sequences of arbitrary length and any valid codon table, thus extending sequence design methods to different biological and protein expression systems. It is more flexible, avoids human-introduced errors, and eliminates redundant computations during the design process, providing a computationally efficient and optimized solution for RNA and DNA design.

[0042] Example 2.

[0043] This embodiment provides a method for automatically constructing DFA diagrams for nucleic acid sequence design. The main technical solution is the same as that in Embodiment 1, except that the order of operations in steps C and D is interchanged, but both have the same effect.

[0044] The method includes the following steps:

[0045] Step A: Read the codon table of a given species to obtain the sequence of all possible codons, including the stop codon, for each amino acid.

[0046] Step B: Constructing the initial DFA subgraph of amino acid residues. This process involves converting the codon sequence into multiple linear DFA subgraphs and connecting them in parallel.

[0047] In this step, an initial DFA subgraph of amino acid residues is constructed, specifically as follows:

[0048] Step B1: Iterate through the codons for each amino acid in sequence, including the stop codon;

[0049] Step B2: For each codon, construct as follows Figure 1 The linear DFA subgraph shown has each edge whose attribute is a base of the codon.

[0050] Step B3: Merge the starting vertices of all linear DFA subgraphs (e.g., ... Figure 1 The vertex numbered 0 in the middle) and the ending vertex (such as Figure 1 (Vertex numbered 3 in the middle), that is, the starting vertex is merged with the starting vertex, and the ending vertex is merged with the ending vertex.

[0051] Step C: Compress and optimize the initial DFA subgraph of amino acid residues obtained in Step B, merging and deleting redundant edges and vertices to make the DFA subgraph have the fewest vertices and edges.

[0052] In this step, the specific method for compressing and optimizing the initial DFA subgraph of amino acid residues is as follows:

[0053] Step C1: Starting from each vertex in the DFA subgraph, check if there are any pairs of identical attributes among the outgoing edges of that vertex. That is, iterate through the outgoing edges of each vertex and determine if the attributes are the same. If there are identical attributes, the following merging operation needs to be performed:

[0054] Delete the second edge with the same attribute, and change all subsequent edges of that edge to end the vertex connection with the first edge that has the same attribute;

[0055] Step C2: Repeat step C1 until no situation requiring modification of the DFA subgraph is found.

[0056] Step D: Connect the DFA subgraphs of the amino acid residues compressed and optimized in step C to form a DFA graph of RNA or DNA.

[0057] The above steps are simply swaps of steps C and D in Example 1. Instead of first linking the DFA subgraphs of basic amino acid residues and then compressing and optimizing the RNA or DNA DFA diagrams, the process is changed to first compressing and optimizing the DFA subgraphs of basic amino acid residues and then linking them. Both methods have similar effects throughout the sequence optimization process; therefore, the technical solution of this invention does not limit the specific product to which method is used.

[0058] Example 3.

[0059] The specific implementation of this embodiment takes the polypeptide MLP as an example. The M-methionine codon can be AUG (ATG for DNA) and is also the start codon. The L-leucine codon can be UUA, UUG, CUU, CUC, CUA, CUG (TTA, TTG, CTT, CTC, CTA, CTG for DNA). The P-proline codon can be CCU, CCC, CCA, CCG (CCT, CCC, CCA, CCG for DNA). Finally, there are three stop codon options: UAA, UGA, UAG (TAA, TGA, TAG for DNA).

[0060] The DFA diagram of this polypeptide was constructed using the method described in Example 1 or Example 2. The DFA diagrams of the mRNA before and after compression optimization are shown below. Figure 2As shown (for DNA DFA, simply replace the U bases with T bases):

[0061] The (0,0) node is the starting node, and the (12,0) node is the ending node. Figure 2 The left-hand graph in the image is an uncompressed DFA graph with 33 vertices and 42 edges, where the attributes (i.e., base symbols) on the edges have a lot of redundancy. Figure 2 The right-hand graph in the diagram is a compressed and optimized DFA graph, with the number of vertices reduced to 16 and the number of edges reduced to 25, nearly halving the size of the data structure. This significantly reduces the complexity of subsequent parsing based on context-free syntax.

[0062] Example 4.

[0063] The specific implementation of this embodiment takes the current SARS-CoV-2 spike protein sequence design as an example. This protein contains 1273 amino acids, therefore, the translation region has a sequence length of 3*(1273+1) = 3822 nucleotides. The RNA or DNA sequence design process for this protein first uses the method of Example 1 or Example 2 to automatically construct a DFA diagram (and then compresses and optimizes the DFA diagram), and then uses existing technical methods to test the running time on a computer.

[0064] To demonstrate the improvements of the present invention in terms of DFA graph properties and computational efficiency, Table 1 below lists two existing RNA or DNA sequence design techniques:

[0065] 1. Bottom-up dynamic methodology on Nussinov model;

[0066] 2. The left-to-right dynamic methodology with beam pruning, which is the method used in the invention patent with publication number CN114974428A and patent title "System and Method for Sequence Design" mentioned in the background art;

[0067] Under the two different computational techniques described above, the changes in the properties and computational performance of the DFA graphs constructed using the method of this invention and those constructed without this invention are compared. The results are as follows: Figure 1 As shown.

[0068] The test was run on a machine with an AMD EPYC 7742 CPU, therefore the time taken is the test value under that hardware condition, but this technical method is not limited to the hardware used in this embodiment.

[0069] Table 1. Computational efficiency of SARS-CoV-2 spike protein sequence design using different methods (3822 nt)

[0070]

[0071]

[0072] 1 Bottom-up dynamic methodology on Nussinov model;

[0073] 2 A left-to-right dynamic methodology with beam pruning;

[0074] 3 Beam size: The size setting for the pruning process.

[0075] As shown in Table 1, the overall data size of the RNA or DNA DFA diagram is reduced by nearly half after the compression optimization steps using the method of this invention. Furthermore, the method of this invention automatically constructs the DFA diagram, and subsequent calculations using different computational techniques (Bottom-Up in Table 1) are also possible. 1 and LinearDesign 2 The peak main memory usage and computation time are significantly reduced. For example, under the conditions of LinearDesign method and beam size = 500, after the compression optimization process of this invention, the computation time is significantly reduced from 101 minutes to 16 minutes, and the main memory usage is significantly reduced from 7.1GB to 3.6GB. Therefore, the automatic DFA diagram construction method of this invention has a significant efficiency advantage in the entire technical solution of RNA or DNA sequence design, and can provide greater throughput under the same computing power conditions. Compared with the existing technology of manually constructing optimized RNA or DNA DFA diagrams, this invention can handle the diverse codon tables in different species and protein expression systems, is more flexible and avoids human-introduced errors, directly avoids redundant computation in the design process, and provides an effective and optimized solution for RNA and DNA sequence design in terms of computation time.

[0076] The above-described embodiments are preferred embodiments of the present invention and are only used to facilitate the illustration of the present invention. They are not intended to limit the present invention in any way. Any person skilled in the art who makes local modifications or alterations to the technical content disclosed in the present invention without departing from the scope of the technical features of the present invention shall still fall within the scope of the technical features of the present invention.

Claims

1. A method for automatically constructing DFA diagrams for nucleic acid sequence design, characterized in that: Includes the following steps: Step A: Read the codon table of a given species to obtain the sequence of all possible codons, including the stop codon, for each amino acid; Step B: Constructing the initial DFA subgraph of amino acid residues. This process involves converting the codon sequence into multiple linear DFA subgraphs and connecting them in parallel. Step C: Connect the initial DFA subgraphs of amino acid residues obtained in Step B to form a DFA graph of RNA or DNA. The RNA or DNA DFA graphs obtained in steps D and C are compressed and optimized by merging and deleting redundant edges and vertices to make the DFA graph have the fewest vertices and edges. In this case, the order of operations in steps C and D can be interchanged. That is, the initial DFA subgraph of amino acid residues obtained in step B is first compressed and optimized, and then the compressed and optimized DFA subgraph is connected to form a DFA graph of RNA or DNA. In step B, the initial DFA subgraph of amino acid residues is constructed as follows: Step B1: Iterate through the codons for each amino acid in sequence, including the stop codon; Step B2: For each codon, construct a linear DFA subgraph, where the attributes of each edge of the linear DFA subgraph are, in turn, the bases of the codon. Step B3: Merge the starting and ending vertices of all linear DFA subgraphs, that is, merge the starting vertices with each other and merge the ending vertices with each other. In step C or step D, the method for compressing and optimizing the DFA diagram of RNA or DNA or the initial DFA sub-diagram of amino acid residues is as follows: Step D1: Starting from each vertex in the DFA graph or DFA subgraph, check if there are any pairs of identical attributes among the outgoing edges of that vertex. That is, iterate through the outgoing edges of each vertex and determine if there are identical attributes. If there are identical attributes, the following merging operation needs to be performed: Delete the second edge with the same attribute, and change all subsequent edges of that edge to end the vertex connection with the first edge that has the same attribute; Step D2: Repeat step D1 until no situation arises where the DFA diagram or DFA subgraph needs to be modified.

Images