A nucleic acid multi-sequence re-alignment method based on star alignment results

Abstract

The application belongs to the field of biological information, and provides a nucleic acid multi-sequence re-alignment method based on star alignment results, which is used to improve the adaptability and alignment sensitivity of the star alignment tool on different data types. The application firstly performs similarity calculation on all sequences in the initial nucleic acid multi-sequence alignment matrix by using a sliding window, identifies low-similarity sequences, and extracts a low-similarity sequence set from all low-similarity sequences; then, the remaining sequences are summarized to form an initial sequence spectrum, the initial spectrum is obtained through column screening, and a local re-alignment operation in the vertical direction is performed to obtain a local re-alignment sequence spectrum; finally, each sequence in the low-similarity sequence set is subjected to sequence-spectrum alignment with the local re-alignment sequence spectrum. The application adopts a hybrid re-alignment strategy, which combines horizontal re-alignment and vertical re-alignment, effectively optimizes the initial alignment result generated by the star alignment tool, and significantly improves the accuracy and robustness of the alignment.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics, specifically providing a method for nucleic acid multiple sequence re-alignment based on star alignment results. Background Technology

[0002] The star alignment algorithm, proposed by Stephen F. Altschul, is a fast multi-sequence alignment algorithm. It first selects a center sequence, then aligns other sequences with it, and finally merges the alignment results to form the final alignment result. Although the star alignment algorithm does not require a guide tree and is fast, it also has some key limitations: sequences that differ significantly from the center sequence will reduce the alignment accuracy, and the alignment accuracy will decrease as the number of sequences increases. These problems stem from the insertion of gaps during the merging stage, which leads to a decrease in the overall alignment quality and thus limits its applicability.

[0003] In nucleic acid multiple sequence alignment, to improve the accuracy of alignment results, Higgins et al. proposed three iterative re-alignment optimization algorithms in the paper "Wallace, I.M. and D.G. Higgins, Evaluation of iterative alignment algorithms for multiple alignment. Bioinformatics, 2005, 21(8): p.1408-1414.": Remove First, Best First, and Random re-alignment. These methods optimize the overall alignment by removing or segmenting sequences in each iteration and re-aligning them, and decide whether to retain the new alignment results based on whether the score improves. To further improve efficiency and accuracy, the above algorithms are also embedded in the tree alignment framework, that is, iterative optimization steps are added when constructing the evolutionary tree or merging sub-alignments. In addition, the authors also proposed a tree-based segmentation method, which significantly reduces the alignment time by dividing the sequence into smaller subsets, aligning them separately, and then merging them, combined with the iterative optimization strategy. On the other hand, the literature "Chakrabarti, S., et al., Refining multiple sequence alignments with conserved core regions. Nucleic acids research, 2006, 34(9): p.2598-2606" proposes a re-alignment tool called REFINER. Its core idea is to locate alignment blocks that need optimization and iteratively refine them. This tool defines alignment blocks as continuous segments without gaps in all sequences and without overlap. It first extracts columns without gaps from the original alignment to construct the alignment block model, and then performs multiple rounds of iterative optimization. The entire process takes a maximum of five rounds, and convergence is considered achieved when the alignment score no longer improves. However, in star alignment, all sequences are aligned with a single center sequence. Low similarity sequences or a large number of sequences significantly reduce alignment accuracy. Therefore, during the re-alignment process using current methods, insertion errors may persist in the segmented spectra, thus limiting the potential of the initial alignments obtained by optimizing star alignment tools. Summary of the Invention

[0004] The purpose of this invention is to provide a better-star method for nucleic acid multiple sequence re-alignment based on star alignment results, which improves the adaptability and alignment sensitivity of the star alignment tool on different data types. This invention adopts a hybrid re-alignment strategy, which integrates horizontal and vertical re-alignment methods, effectively optimizes the initial alignment results generated by the star alignment tool, and significantly improves the accuracy and robustness of the alignment.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for nucleic acid multiple sequence re-alignment based on star alignment results, characterized by comprising the following steps:

[0007] Step 1. Use a sliding window to calculate the similarity of all sequences in the initial nucleic acid multiple sequence alignment matrix, and identify low similarity sequences based on the similarity.

[0008] Step 2. For all low-similarity sequences, remove the insertions in the low-similarity sequences and extract the set of low-similarity sequences;

[0009] Step 3. Summarize the remaining sequences except for the low similarity sequences to form an initial sequence list spectrum, and then filter the columns of the initial sequence list spectrum to obtain the initial spectrum that needs to be locally re-aligned.

[0010] Step 4. Perform a local re-alignment operation in the vertical direction on the initial spectrum to obtain the locally re-aligned sequence spectrum;

[0011] Step 5. Perform sequence-spectrum alignment on each sequence in the low similarity sequence set with the local re-aligned sequence spectrum to obtain the global re-alignment matrix of nucleic acid multiple sequences.

[0012] Furthermore, the specific process of step 1 is as follows:

[0013] Set the length of the sliding window and slide the sliding window in the initial multiple sequence alignment matrix from left to right. For each sliding window, identify low similarity sequences. When there is a sequence in the sliding window that consists entirely of bases within the window and all other sequences that consist entirely of empty symbols within the window, the sequence containing the only bases is determined to be a low similarity sequence. The sliding window slides column by column, and the above identification process is repeated until the end of the initial multiple sequence alignment matrix.

[0014] Furthermore, the specific process of step 2 is as follows: for each low similarity sequence, delete the empty symbols in the sequence and keep only the base sequence to obtain a set of low similarity sequences.

[0015] Furthermore, in step 3, column filtering specifically involves deleting columns in the initial sequence list spectrum where all sequences are empty symbols.

[0016] Furthermore, the specific process of step 4 is as follows:

[0017] Step 4-1. Select the center sequence;

[0018] Select a central sequence from the initial spectrum as the reference sequence for vertical segmentation;

[0019] Step 4-2. Determine the vertical cutting strategy based on the vacancy distribution of the central sequence;

[0020] In the central sequence, each column is scanned from left to right, the regions where vacancy symbols appear are recorded, the number of bases between any two adjacent vacancy regions is measured, and the average number of bases between all adjacent vacancy regions is calculated and denoted as n.

[0021] Step 4-3. Set the segmentation threshold COUNT;

[0022] If the number of sequences in the initial spectrum exceeds 1000 or n>10, then set COUNT to 10; otherwise, set COUNT to the integer part of n.

[0023] Step 4-4. Perform vertical segmentation and locate the re-comparison blocks;

[0024] The central sequence is scanned again from left to right. When a void symbol is detected, it is marked as the starting position of the current block. The number of consecutive bases z after the starting position is continuously counted. When z exceeds the segmentation threshold COUNT, the current block is considered closed. The length of the current block is further checked to see if it reaches the minimum length threshold. If it does (the length of the current block is greater than or equal to the minimum length threshold), the current block is recorded as a re-aligned block. Finally, multiple re-aligned blocks are formed, each corresponding to a specific region of the initial spectrum.

[0025] Steps 4-5. Perform local re-alignment separately for each re-alignment block.

[0026] Furthermore, in step 4-1, the sequence with the most bases in the initial spectrum is selected as the default center sequence.

[0027] Furthermore, in steps 4-5, the specific process of local re-alignment is as follows: remove all empty symbols from the re-alignment block, retaining only bases; then call the specified multiple sequence alignment tool (such as MUSCLE or MAFFT) to perform re-alignment on the remaining sequence; after the multiple sequence alignment tool completes the re-alignment, calculate the SP (Sum-of-Pairs) score between the new alignment block and the original block, and select the alignment version with the higher SP score to replace the content of the re-alignment block.

[0028] Furthermore, the specific calculation method for the SP score is as follows:

[0029]

[0030] Where l represents the total number of columns in the current re-comparison block, and f(m) is the score of the m-th column, calculated as follows:

[0031]

[0032] Where M represents the total number of sequences in the current re-aligned block, and score(i,j) represents the alignment score of sequence i and sequence j in column m. If the two sequences have the same base in this column (perfect match), the score is 1; if the two sequences have different bases in this column (mismatch), the score is -1; if at least one of the two sequences has a void symbol in this column, the score is -2; otherwise, the score is 0.

[0033] Furthermore, the specific process of step 5 is as follows: for each sequence in the set of low similarity sequences, perform sequence-spectrum alignment with the local re-alignment sequence list, iterate until all low similarity sequences have been aligned, and then integrate the alignment results to obtain the global re-alignment matrix of nucleic acid multiple sequences.

[0034] Based on the above technical solution, the beneficial effects of the present invention are as follows:

[0035] 1) This invention proposes a dedicated re-alignment method for star alignment tools, denoted as Better-Star, which can effectively improve the accuracy and alignment quality of the initial alignment results of star alignment tools while maintaining the original alignment structure.

[0036] 2) This invention introduces a sliding window-based screening mechanism, which can accurately identify sequences with low similarity to the central sequence in the initial alignment results, significantly improving the sensitivity and reliability of low-quality sequence identification.

[0037] 3) This invention proposes a vertical cutting method based on the distribution of gaps in the central sequence for star alignment tools, and performs local re-alignment within the cutting area, which can effectively improve the alignment accuracy of densely interpolated regions in the initial spectrum.

[0038] 4) This invention further incorporates a horizontal re-alignment strategy, which integrates low-similarity sequences one by one into the optimized spectrum through sequence-spectrum alignment, effectively avoiding a large number of gaps caused by merging low-similarity sequences as a whole, thereby significantly improving the overall quality of the final alignment results.

[0039] 5) This invention comprehensively adopts both vertical and horizontal re-alignment strategies, which improves alignment accuracy while taking into account efficiency and stability, providing a new technical path and implementation idea for the development of multi-sequence alignment post-processing methods. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating the nucleic acid multiple sequence re-alignment method based on star alignment results in this invention.

[0041] Figure 2 This is a schematic diagram of the local re-alignment process of the nucleic acid multiple sequence re-alignment method based on star alignment results in this invention.

[0042] Figure 3 This is an example diagram illustrating the implementation of the local re-alignment of the nucleic acid multiple sequence re-alignment method based on star alignment results in this invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0044] like Figure 1 As shown, this embodiment provides a method for nucleic acid multiple sequence re-alignment based on star alignment results, specifically including the following steps:

[0045] Step 1. Use a sliding window to calculate the similarity of all sequences in the initial nucleic acid multiple sequence alignment matrix, and identify low similarity sequences based on the similarity.

[0046] Receive the initial multiple sequence alignment results generated by any star alignment tool. Assume that the alignment matrix contains N sequences, each sequence having a uniform length of L (including bases and inserted empty space symbols "-"). Use the initial alignment results as input for subsequent re-alignment processing.

[0047] For the initial multiple sequence alignment matrix, a sliding window mechanism is introduced to identify low-similarity sequences. The specific process is as follows: Set the sliding window length to w, which is an empirical value and can be set to 10 by default; slide the sliding window in the initial multiple sequence alignment matrix from left to right. For each sliding window, identify low-similarity sequences. When there is a sequence in the sliding window that consists entirely of bases within the window and the remaining N-1 sequences that consist entirely of empty symbols within the window, the sequence containing the unique bases is determined to be a low-similarity sequence. The sliding window slides column by column, and the above identification process is repeated until the end of the initial multiple sequence alignment matrix.

[0048] Step 2. For all low-similarity sequences, remove the insertions in the low-similarity sequences and extract the set of low-similarity sequences; specifically: for each low-similarity sequence, remove the empty symbols in the sequence and keep only the base sequence as the set of low-similarity sequences.

[0049] Step 3. Summarize the remaining sequences except for the low similarity sequences to form an initial sequence list spectrum, and perform column filtering on the initial sequence list spectrum to obtain the initial spectrum that needs to be locally re-aligned; the column filtering is specifically: delete the columns in the initial sequence list spectrum where all sequences are empty symbols;

[0050] Step 4. Perform a vertical local re-alignment operation on the initial spectrum to optimize the alignment relationship between the sequences within the spectrum, thereby improving the alignment accuracy of the initial spectrum and obtaining a locally re-aligned sequence list spectrum;

[0051] like Figure 2 As shown, the local re-alignment operation is specifically as follows:

[0052] Step 4-1. Select the center sequence;

[0053] A central sequence is selected from the initial spectrum as a reference sequence for vertical segmentation; the central sequence can be specified in advance, or if not specified, the sequence with the most bases in the initial spectrum will be automatically selected as the default central sequence.

[0054] Step 4-2. Determine the vertical cutting strategy based on the vacancy distribution of the central sequence;

[0055] In the central sequence, each column is scanned from left to right, the regions where vacancy symbols appear are recorded, the number of bases between any two adjacent vacancy regions is measured, and the average number of bases between all adjacent vacancy regions is calculated and denoted as n.

[0056] Step 4-3. Set the segmentation threshold COUNT;

[0057] The threshold COUNT represents the maximum allowed non-empty span. If the span exceeds this value, the current empty region is considered closed. The specific rule is: if the number of sequences in the initial spectrum exceeds 1000 or n>10, then COUNT is set to 10; otherwise, COUNT is set to the integer part of n.

[0058] Step 4-4. Perform vertical segmentation and locate the re-comparison blocks;

[0059] The central sequence is scanned again from left to right. When a void symbol is detected, it is marked as the starting position of the current block. The number of consecutive bases z after the starting position is continuously counted. When z exceeds the segmentation threshold COUNT, the current block is considered closed. The length of the current block is further checked to see if it reaches the minimum length threshold (specifically 5). If the length of the current block is greater than or equal to the minimum length threshold, the current block is recorded as a re-aligned block. Finally, multiple re-aligned blocks are formed, each corresponding to a specific region of the initial spectrum.

[0060] Steps 4-5: Perform local re-matching for each re-matching block separately;

[0061] like Figure 3 As shown, local re-alignment specifically involves: removing all empty symbols from the re-alignment block, retaining only bases; and then calling a specified multiple sequence alignment tool (such as MUSCLE or MAFFT) to re-align the remaining sequence.

[0062] After the multi-sequence alignment tool completes the re-alignment, it calculates the Sum-of-Pairs (SP) score between the new aligned block and the original block, and selects the aligned version with the higher SP score to replace the re-aligned block content. The SP score is calculated as follows:

[0063]

[0064] Where l represents the total number of columns in the current re-comparison block, and f(m) is the score of the m-th column, calculated as follows:

[0065]

[0066] Where M represents the total number of sequences in the current re-aligned block, and score(i,j) represents the alignment score of sequence i and sequence j in column m. If the two sequences have the same base in this column (perfect match), the score is 1; if the two sequences have different bases in this column (mismatch), the score is -1; if at least one of the two sequences has a void symbol in this column, the score is -2; otherwise, the score is 0.

[0067] Step 5. Perform sequence-spectrum alignment on each sequence in the low similarity sequence set with the local re-aligned sequence spectrum to obtain the global re-alignment matrix of nucleic acid multiple sequences; specifically: for each sequence in the low similarity sequence set, perform sequence-spectrum alignment on it with the local re-aligned sequence spectrum, iterate until all low similarity sequences have been aligned, and then integrate the alignment results to obtain the global re-alignment matrix of nucleic acid multiple sequences.

[0068] The above description is merely a specific embodiment of the present invention. Any feature disclosed in this specification may be replaced by other equivalent or similar features unless otherwise specified. All disclosed features, or steps in all methods or processes, may be combined in any way except for mutually exclusive features and / or steps.

Claims

1. A method for nucleic acid multiple sequence re-alignment based on star alignment results, characterized in that, Includes the following steps: Step 1. Use a sliding window to calculate the similarity of all sequences in the initial nucleic acid multiple sequence alignment matrix, and identify low similarity sequences based on the similarity. Step 2. For all low-similarity sequences, remove the insertions in the low-similarity sequences to extract the set of low-similarity sequences; Step 3. Summarize the remaining sequences except for the low similarity sequences to form an initial sequence list spectrum, and perform column filtering on the initial sequence list spectrum to obtain the initial spectrum of local re-alignment; Step 4. Perform a local re-alignment operation in the vertical direction on the initial spectrum to obtain the locally re-aligned sequence spectrum; the specific process is as follows: Step 4-1. Select the central sequence; Select a central sequence from the initial spectrum as the reference sequence for vertical segmentation; Step 4-2. Determine the vertical cutting strategy based on the vacancy distribution of the central sequence; In the central sequence, each column is scanned from left to right, recording the regions where vacancy symbols appear. The number of bases between any two adjacent vacancy regions is measured, and the average number of bases between all adjacent vacancy regions is calculated, denoted as . ; Step 4-3. Set the segmentation threshold ; If the number of sequences in the initial spectrum exceeds 1000 or Then set Set to 10, otherwise, set to for The integer part; Step 4-4. Perform vertical segmentation and locate the re-comparison blocks; Scan the center sequence again from left to right. When a void symbol is detected, mark it as the start position of the current block. Continuously count the number of consecutive bases z after the start position. When z exceeds the segmentation threshold... When the current block is closed, it is further determined whether the length of the current block has reached the minimum length threshold. If it does, the current block is recorded as a re-aligned block. Finally, multiple re-aligned blocks are formed, and each re-aligned block corresponds to a specific region of the initial spectrum. Steps 4-5. Perform local re-matching for each re-matching block separately; Step 5. Perform sequence-spectrum alignment on each sequence in the low similarity sequence set with the local re-aligned sequence spectrum to obtain the global re-alignment matrix of nucleic acid multiple sequences.

2. The nucleic acid multiple sequence re-alignment method based on star alignment results according to claim 1, characterized in that, The specific process of step 1 is as follows: Set the length of the sliding window and slide the sliding window in the initial multiple sequence alignment matrix from left to right. For each sliding window, identify low similarity sequences. When there is a sequence in the sliding window that consists entirely of bases within the window and all other sequences that consist entirely of empty symbols within the window, the sequence containing the only bases is determined to be a low similarity sequence. The sliding window slides column by column, and the above identification process is repeated until the end of the initial multiple sequence alignment matrix.

3. The nucleic acid multiple sequence re-alignment method based on star alignment results according to claim 1, characterized in that, The specific process of step 2 is as follows: For each low similarity sequence, delete the empty symbols in the sequence and keep only the base sequence to obtain a set of low similarity sequences.

4. The nucleic acid multiple sequence re-alignment method based on star alignment results according to claim 1, characterized in that, In step 3, column filtering specifically involves deleting columns in the initial sequence list spectrum where all sequences are empty symbols.

5. The nucleic acid multiple sequence re-alignment method based on star alignment results according to claim 1, characterized in that, In step 4-1, the sequence with the most bases in the initial spectrum is selected as the default center sequence.

6. The nucleic acid multiple sequence re-alignment method based on star alignment results according to claim 1, characterized in that, In steps 4-5, the specific process of local re-alignment is as follows: remove all empty symbols from the re-alignment block, retaining only bases; then call the specified multiple sequence alignment tool to perform re-alignment on the remaining sequence; after the multiple sequence alignment tool completes the re-alignment, calculate the SP score of the new alignment block and the original block, and select the alignment version with the higher SP score to replace the content of the re-alignment block.

7. The nucleic acid multiple sequence re-alignment method based on star alignment results according to claim 6, characterized in that, The SP score is calculated as follows: in, This indicates the total number of columns in the currently re-compared block. For the first The column score is calculated as follows: in, This indicates the total number of sequences in the currently re-aligned block. Represents a sequence and sequence In the The alignment score is calculated as follows: if the two sequences have the same base in that column, the score is 1; if the two sequences have different bases in that column, the score is -1; if at least one of the two sequences has a void symbol in that column, the score is -2; otherwise, the score is 0.

8. The nucleic acid multiple sequence re-alignment method based on star alignment results according to claim 1, characterized in that, The specific process of step 5 is as follows: For each sequence in the set of low similarity sequences, perform sequence-spectrum alignment with the local re-alignment sequence list, and perform iterative execution until all low similarity sequences have been aligned. After integrating the alignment results, the global re-alignment matrix of nucleic acid multiple sequences is obtained.

Images