KR100537636B1 - Apparatus for predicting transcription factor binding sites based on similar sequences and method thereof - Google Patents

Abstract

The present invention relates to an apparatus and method for predicting transcription factor binding sites by extracting similar sequences, in particular, constitute a suffix sequence after pre-processing the input sequence, and the LCP information for the generated suffix sequence is present in common in various sequences. Apparatus and method for predicting transcription factor binding site by extracting sequence from the candidate sequence by extracting the sequence and extracting candidate sequences for transcription factor binding site, which can reduce time and cost according to prediction of transcription factor binding site It is about.

Transcription factor binding site prediction apparatus through the extraction of the similar sequence, the transcription factor binding site predictor for outputting the transcription factor binding sites predicted in the target sequence by pre-processing, suffix arrangement and local alignment of the input sequence; And a binding site database for storing the transcription factor binding site sequences outputted from the transcription factor binding site predictor.

Description

Apparatus for predicting transcription factor binding sites based on similar sequences and method etc.

The present invention relates to an apparatus and method for predicting transcription factor binding sites by extracting similar sequences, in particular, constitute a suffix sequence after pre-processing the input sequence, and the LCP information for the generated suffix sequence is present in common in various sequences. Apparatus and method for predicting transcription factor binding site by extracting sequence from the candidate sequence by extracting the sequence and extracting candidate sequences for transcription factor binding site, which can reduce time and cost according to prediction of transcription factor binding site It is about.

Recently, as the genome sequence is revealed after the human genome project, interest in the field of transcriptional regulators involved in gene expression is increasing. By studying the regulation of transcription in detail the location and function of the gene and by looking at the expression level of the gene according to the in vivo, it is possible to study the various expression possibilities of the gene. The transcription factor binding site is a very important research field for predicting gene function along with the completed human chromosome map and expression information obtained from a large-scale experiment DNA chip.

Despite this importance, the transcription factor binding site in the laboratory is time-consuming and expensive and difficult to predict for unknown sites because the binding site of the transcription factor is relatively short and the location is not constant. There was this.

Various approaches have been taken on in silico to compensate for this. These include the 'search by signal' method for creating a mathematical and statistical model using known data from a certain field and searching for similarity, and the 'search by content' method for predicting the characteristics of a sequence class.

However, the above-described conventional methods through biological experiments and curation have the disadvantage that the approach is very limited and difficult to predict unknown sites. In addition, statistical methods or feature-based methods require a large amount of statistical data for learning a system, or require a lot of information in advance, such as having to be aware of the characteristics of the sequence in advance.

The present invention has been made to solve the above problems, an object of the present invention is to pre-process, suffix sequence generation and local alignment when the base sequence is input to predict the binding site in a short time without a separate learning process The present invention provides an apparatus for predicting a transcription factor binding site by extracting similar sequences configured to be stored in a database.

Another object of the present invention is to construct a suffix sequence after pre-processing the input sequence and extract the sequences that are commonly present in the various sequences by LCP information on the generated suffix sequence, and then use a local alignment technique to select transcription factor binding site candidate sequences. The present invention provides a method of predicting a transcription factor binding site by extracting a similar sequence that can efficiently reduce time and cost according to the prediction of the transcription factor binding site.

In order to achieve the above object, a transcription factor binding site predicting apparatus through the extraction of a similar sequence of the present invention is a transcription for outputting the transcription factor binding sites predicted from a target sequence by pre-processing the input sequences, generating a suffix array and local alignment. Factor binding site predictor; And a binding site database for storing the transcription factor binding site sequences outputted from the transcription factor binding site predictor.

In the above-described configuration, the transcription factor binding site predictor may include a sequence preprocessor for performing preprocessing to connect the input sequences into one long sequence; A suffix array generation processing unit for sorting the suffixes for the entire sequence preprocessed from the sequence preprocessor in ascending order and maintaining the sorted result in an array form; And generating each LCP sequence by comparing each term of the suffix array generated from the suffix array generating processor, filtering the LCP sequences in the generated LCP sequence according to a predetermined length and ratio, and filtering each of the filtered LCP sequences. It is preferable to include a local alignment processor for calculating an alignment score after performing local alignment on all input sequences and outputting predicted transcription factor binding sites according to the result.

In addition, a method for predicting transcription factor binding sites by extracting similar sequences, the method comprising: (a) performing pre-processing of input sequences; (b) generating a suffix sequence for the entire pretreated sequence; (c) comparing each term of the generated suffix array to produce an LCP array; (d) performing approximate matching using local alignment when the length and ratio of each LCP sequence of the generated LCP sequence is greater than or equal to a predetermined reference value; And (e) calculating the alignment scores of the locally aligned sequences and predicting the corresponding sequences as candidates for transcription factor binding sites when the alignment scores are greater than or equal to a predetermined threshold value, and storing the sequences in a previously prepared database.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. This embodiment is not intended to limit the scope of the invention, but is presented by way of example only.

FIG. 1 is a block diagram illustrating an apparatus for predicting a transcription factor binding site by extracting similar sequences according to an embodiment of the present invention.

As shown in FIG. 1, the apparatus for predicting transcription factor binding sites through the extraction of similar sequences according to the present invention includes pre-treatment of the input sequence 100, generation of a suffix sequence, and local alignment to predict transcription factor binding sites from a target sequence. And a binding site database 300 for storing the transcription factor binding site predictor 200 and the transcription factor binding site sequences output from the transcription factor binding site predictor 200.

In the above configuration, the input sequence 100 is a nucleotide sequence that is assumed to have the same transcription factor binding site to find a transcription factor binding site.

The transcription factor binding site predictor 200 includes a sequence preprocessor 210 for performing preprocessing to connect the input sequences into one long sequence; A suffix array generation processor 220 for sorting the suffixes for the entire sequence preprocessed from the sequence preprocessor 210 in ascending order and maintaining the sorted results in an array form; Comparing the terms of the suffix array generated from the suffix array generation processing unit 220 to generate the LCP sequence, and compares the length and ratio of the LCP sequence in the generated LCP sequence with a predetermined reference value to determine the common sequence according to the result. A local alignment processor for extracting and performing local alignment on the extracted common sequence and the entire input sequence, calculating the alignment scores of the locally aligned sequences, and outputting predicted transcription factor binding sites according to the results 230.

The transcription factor binding site stored in the binding site database 300 may be used to search for transcription factor binding sites in the sequence when an unknown sequence is input later.

Hereinafter, a method of predicting a transcription factor binding site by extracting a similar sequence of the present invention having the above-described configuration will be described in detail.

2 is an overall flowchart illustrating a method of predicting a transcription factor binding site by extracting similar sequences according to an embodiment of the present invention.

As shown in FIG. 2, first, in step S100, a preprocessing operation of connecting input sequences into one long sequence is performed. That is, the preprocessing operation is a necessary operation for finding a common pattern existing in several sequences since the suffix array described later is a data structure for finding a pattern existing in a single sequence.

Meanwhile, a special character (eg, # 1, # 2, ..., #n, see FIG. 3) consisting of ASCII characters is inserted between the linked sequence and the sequence. Here, the special character serves to identify the frequency of the sequence in which each pattern appears by distinguishing the position of the input sequence to which the pattern found in the linked sequence belongs.

Next, in step S200, a suffix array is generated for the entire preprocessed sequence. Here, the suffix arrangement sorts the suffixes for all linked sequences in ascending order to maintain the sorted result in an array form (see FIG. 4B).

When the suffix array is generated, the process proceeds to step S300 to compare each term of the generated suffix array to generate an LCP array. In this case, the Longest Common Prefix (LCP) is the longest common prefix that exists in a specific section of the suffix array and is used for comparing sequences for common pattern discovery. Therefore, LCP according to an embodiment of the present invention is limited to meaning the longest common prefix of two adjacent items in the suffix arrangement.

Once the LCP sequence is generated, the length of each LCP sequence and the number of input sequences in which the LCP appears can be found. These two factors are used as a factor to verify whether the LCP sequence is likely to be a binding site candidate.

Next, in step S400, it is determined whether the length of each LCP sequence of the generated LCP sequence is greater than or equal to a predetermined reference length (hereinafter, defined as “LEN”). If LEN is greater than or equal to LEN, the flow proceeds to step S500 to determine whether the ratio of each LCP sequence of the generated LCP sequence is greater than or equal to a predetermined reference ratio (hereinafter, referred to as “RTO”).

Here, the length of the sequence is set as an evaluation factor to reflect the general binding site length is less than 20bp, and the ratio of appearance to the input sequences is more likely as the repeating pattern appears in the input sequences. This is to reflect the high.

Experiments with varying RTO and LEN for the verification of the present invention, the experimental results can be obtained in Table 1 as follows.

Positive probability value (PPV) bp% 100 95 90 85 80 75 70 65 60 55 4 30.9 30.9 30.0 35.3 29.4 29.6 30.0 30.2 27.5 27.6 5 18.2 18.2 14.6 17.5 13.0 16.9 7.6 18.1 15.9 16.7 6 0 0 0 0 15.2 14.3 13.3 13.3 10.9 12.1 7 0 0 0 0 25.0 25.0 25.0 25.0 10.4 10.0 8 0 0 0 0 0 0 0 0 8.6 8.2 9 0 0 0 0 0 0 0 0 7.3 7.3

Herein, in order to verify the transcription factor coupling site prediction result, a positive probability value (PPV) representing the correct prediction ratio is obtained from the predicted result. , TP: True positive, FP: False positive).

In fact, even if the length of the transcription factor binding site LEN is 4bp or more and the RTO is 85% or more, the PPV value is 30% or more. However, when the LEN is set to 4bp, verification through actual experiments becomes very difficult due to an increase in false positives. As a result of analyzing the program, LEN showed significant performance at 5 ~ 7bp, and RTO showed appropriate performance at 65 ~ 85%.

However, there is a possibility that the optimized value of the factor may be changed according to the type of object, and the optimized factor value may be determined through various tests for each input data.

On the other hand, as a result of the determination in step S500, if the ratio of each LCP sequence of the generated LCP array is not a predetermined RTO or more terminated as it is, otherwise the ratio of each LCP sequence of the generated LCP sequence is a predetermined RTO or more In the case of step S600, approximate matching using local alignment is performed.

That is, the approximate matching technique performs local alignment on each filtered LCP sequence and the entire input sequence, which is local to a sequence that is not found by loss that may occur due to insertion, deletion, or the like, of bases in the sequence. We want to increase the probability of discovery by performing approximate matching using.

Next, in step S700, the alignment score of the locally aligned sequences is calculated to determine whether or not the threshold is greater than or equal to a preset threshold, and if not abnormally terminated, otherwise, the alignment score of the locally aligned sequences is calculated to calculate the alignment score of the locally aligned sequences. In case of abnormality, the process proceeds to step S800, the sequence is predicted as a candidate for transcription factor binding site, and the binding sites that are likely to exist for each input sequence are listed as a result of the prediction, and are stored in the binding site database 300 for later use. .

3 is a view for explaining the overall method of predicting the transcription factor binding site through the extraction of similar sequences in accordance with an embodiment of the present invention.

As shown in Fig. 3, the input sequence is n base sequences (A), in which sequences are expected to have a common transcription factor binding site. Concatenation of input sequences after preprocessing yields one long sequence (B). Since a special character is inserted for each sequence, a total of n special characters are inserted.

Generating a suffix sequence for the linked sequence produces an array of the same type as (C). (C) shows an example of sequences having a common prefix of 'AGCTC' in the entire array. Sequences with prefixes can be identified by special characters to which input sequence they belong. That is, in this figure, the input sequence number is shown to the right of (C). LCPs like (D) can be obtained by comparing adjacent prefixes in an array.

In case (C), 'AGCTCG' and 'AGCTC' are obtained. The generated LCPs are then obtained with their lengths and the frequencies shown in the entire input sequence and compared with the predefined factors LEN, RTO (E) to perform local alignment if the conditions are met.

An example of (F) shows the local alignment of 'AGCTC' among LCPs satisfying the condition. Each filtered LCP is subjected to local alignment with the entire input sequence, respectively. If the alignment result score value is greater than or equal to the threshold (G), the corresponding part is selected as a candidate for joining sites. In (H), the predicted binding site candidates are shown in bold type after local alignment of LCP 'AGCTC'. In addition to 'AGCTC', 'AGCC' and 'AACTC' can be found. Multiple binding site sequences can be predicted from several input sequences and they are stored in binding site database 300.

4A and 4B are diagrams for describing a suffix array generation process in a method of predicting transcription factor coupling sites through similar sequence extraction according to an embodiment of the present invention, and FIG. 4A illustrates a suffix array generation process applied to the present invention. 4B is a flowchart illustrating a result of generating a suffix array for the example sequence applied to the present invention.

As shown in FIGS. 4A and 4B, the algorithm for generating the suffix array in step S200 uses an algorithm introduced in 2003 by Karkkainen and Sanders.

In other words, the algorithm generates an array of suffixes for a given string in a linear time using a divide and conquer technique. Whereas the previous algorithm divides the string into two parts and creates a suffix array for each of them, merges them, whereas the algorithm of the present invention takes a mod operation on the string and divides it into three parts and performs a sort on the two parts. Sorting and merging them has the advantage that the merging step is very simple compared to the previous method.

Karkkainen and Sanders' suffix array generation process is shown in Figure 4a. First, given the input sequence as shown in (I), each of the suffixes in the sequence divided by 3 (mod 3 result) is 2 and 0 Group three bases for the phosphorus positions to form a new sequence (J). If the last part is not a multiple of 3, the random symbol '0' is inputted according to the number and the operation is performed. This symbol has no effect on the operation. The sequences thus bundled are numbered in alphabetical order.

The same sequence is given the same number and the alignment is repeated until all of the numbers in the aligned sequence have a unique number. The result of an array with a unique number is shown in (L). The suffixes at positions 0 and 2 of the Mod 3 operation are easily ordered via (L).

Next, sort the suffixes where the result of mod 3 operation is 1 (M). This can be easily obtained by taking advantage of the fact that the mod 3 result is 1 in front of the mod 3 result. Now we can merge the two sorted arrays into one to get the full suffix array (N). The result of generating the suffix array by the present algorithm for the sequence given in (I) is shown in FIG. 4B.

FIG. 5 is a conceptual diagram illustrating a local alignment for approximation matching in a method of predicting transcription factor coupling sites through similar sequence extraction according to an embodiment of the present invention.

As shown in FIG. 5, the local alignment in step S600 is an alignment method used to examine the homology of partial sequences existing within two sequences, and a threshold or maximum value for a score on a matrix using a dynamic programming technique. Refers to a technique for searching for homology by reversing the trace generated based on In the present invention, the widely used Smith Waterman algorithm is basically used.

If only simple matching method is applied for pattern discovery, accurate prediction about insertion, deletion, etc. cannot be made. Although the (O) sequence was predicted to be a binding site in the six input sequences, the actual binding site may be inserted into or deleted from this sequence, and thus may not necessarily be identical to this sequence. If the sequence (P) is a binding site sequence, the simple matching method will yield meaningless prediction results.

It is an effect that can be obtained using an approximate matching method by performing local alignment on the (O) sequence and the input sequences so that the highly homologous (P) sequence can be predicted as the binding site.

Although a preferred embodiment of the apparatus and method for predicting transcription factor binding sites by extracting similar sequences according to the present invention has been described above, the present invention is not limited thereto, and the claims and the detailed description of the invention and the accompanying drawings. It is possible to carry out various modifications within the scope of this also belongs to the present invention.

According to the apparatus and method for predicting transcription factor binding sites by extracting similar sequences of the present invention as described above, the suffix array is constructed after pre-processing the input sequence and is common in various sequences by LCP information on the generated suffix sequence. By extracting the existing sequences and extracting candidate sequences of transcription factor binding sites using local alignment techniques, less data is compared to the conventional statistical method, which undergoes a learning step with a large amount of learning data and predicts the learned contents. It is possible to obtain fast performance results and to reduce the time and cost of predicting the transcription factor binding site.

In addition, the binding site sequences predicted by the present invention has an advantage that can be useful in the case where it is necessary to predict the binding site for the unknown sequence by establishing a database for each transcription factor.

1 is an overall block diagram showing an apparatus for predicting a transcription factor binding site by extracting similar sequences according to an embodiment of the present invention;

2 is an overall flowchart illustrating a method of predicting a transcription factor binding site by extracting a similar sequence according to an embodiment of the present invention;

3 is a view for explaining the overall transcription factor binding site prediction method through the extraction of similar sequences in accordance with an embodiment of the present invention,

4A and 4B are diagrams for describing a suffix array generation process in a method of predicting transcription factor coupling sites through similar sequence extraction according to an embodiment of the present invention, and FIG. 4A illustrates a suffix array generation process applied to the present invention. 4B is a flowchart illustrating a result of generating a suffix array for the example sequence applied to the present invention.

FIG. 5 is a conceptual diagram illustrating a local alignment for approximation matching in a method of predicting transcription factor coupling sites through similar sequence extraction according to an embodiment of the present invention.

*** Explanation of symbols on main parts of drawing ***

100: input sequence, 200: transcription factor binding site predictor,

210: sequence preprocessor, 220: suffix array generation processor,

230: local sort processing unit, 300: coupling site database (DB)

Claims (3)

A transcription factor binding site predictor for pre-processing the input sequences, generating a suffix array, and locally aligning the predicted transcription factor binding sites in the target sequence; And Comprising a binding site database for storing the transcription factor binding site sequences outputted from the transcription factor binding site predictor, apparatus for predicting transcription factor binding sites through the extraction of similar sequences. The method of claim 1, wherein the transcription factor binding site predictor, A sequence preprocessor for performing preprocessing to link the input sequences into one long sequence; A suffix array generation processing unit for sorting the suffixes for the entire sequence preprocessed from the sequence preprocessor in ascending order and maintaining the sorted result in an array form; And Comparing each term of the suffix array generated from the suffix array generation processing unit to generate the longest common prefix (LCP) array, and to determine the length and ratio of the longest common prefix (LCP) sequence in the generated longest common prefix (LCP) array. A common sequence is extracted according to the result compared with a predetermined reference value, local alignment is performed on the extracted common sequence and the entire input sequence, and then the alignment scores of the locally aligned sequences are calculated according to the result. An apparatus for predicting transcription factor binding sites by extracting similar sequences, comprising a local alignment processor for outputting predicted transcription factor coupling sites. In the method of predicting transcription factor binding site by extracting similar sequence, (a) preprocessing the input sequences; (b) generating a suffix sequence for the entire pretreated sequence; (c) comparing each term of the generated suffix array to produce a longest common prefix (LCP) array; (d) performing approximate matching using local alignment when the length and ratio of each longest common prefix (LCP) sequence of the generated longest common prefix (LCP) sequence is greater than or equal to a predetermined reference value; And (e) calculating an alignment score of the locally aligned sequences, and predicting the sequence as a candidate for a transcription factor binding site when the alignment score is greater than or equal to a predetermined threshold, and storing the sequence in a database prepared in advance. Method of predicting transcription factor binding site by extraction.