KR101943053B1 - System and method for detecting copy number variation - Google Patents

Abstract

We propose a gene unit repeat detection system and method that extracts CNV region with minimum noise characteristics of data through transforming exome sequencing data using 4 step normalization transformation method. The proposed normalization processing unit of the gene unit repeat variation detection system divides ratio data or test data calculated on the basis of test data and control data into a plurality of segments having the same size with respect to the target region, The coverage data of each segment is transformed so as to have a normalized distribution based on the normalized average value of the segments and the ratio data of each segment to extract the effective ratio data.

Description

TECHNICAL FIELD The present invention relates to a system and a method for detecting a gene unit repeat variation,

TECHNICAL FIELD The present invention relates to a technique for detecting a copy number variation (CNV) which is one of structural variations (SV) of a human genome.

Copy Number Variation (CNV) is defined as a variation in the number of copies of DNA segments of 50 bp or more by comparing two different DNA sequences.

Genetic repeat variants are very important, associated with human diseases such as autism, intellectual disability, epilepsy, schizophrenia, obesity, and cancer. It is one of the variation types.

In particular, in recent studies, studies on the identification of causative genes based on rare and de novo CNV extraction using exome sequencing data have been continuously reported.

However, WES (Whole Exome Sequencing) data has various noise characteristics such as small exome area size, discontinuity of exoskeleton, probe dependency.

Existing CNV extraction algorithms based on WES data include ExomeCNV, Contra, CoNIFER, XHMM, and Excavator. Each algorithm adopts the following method to correct noise characteristics of WES data.

ExomeCNV uses Geary-Hinkley transformation to derive the normalized distribution of lead coverage data.

Contra adopts a normalization method that performs operations such as base-level log-ratio and library size correction.

CoNIFER and XHMM remove singular value decomposition (SVD) and principal component analysis (PCA) to remove noise in lead coverage data.

In addition, the excavator adopts the median normalization method to remove the noise generated from GC content, mappability, and exon size.

However, since CoNIFER and XHMM use very large amount of sample data to apply SVD or PCA normalization method, they are difficult to apply in a limited environment without many samples.

There is also a restriction in Contra that many samples must be used to generate the base control.

In some algorithms, a circular binary segmentation (CBS) algorithm is applied as a partitioning method for distinguishing mutation regions.

However, when the CBS algorithm is applied to a discontinuous data area, such as exo-coverage data, the CNV area is likely to be missed when the area size is very small or large .

Korean Patent Publication No. 10-2017-0000744 (name: method and apparatus for analyzing copy number variation (CNV) of a gene)

In order to solve the above-described problems of the present invention, there is provided a gene unit repeat mutation detection system and method for accurately extracting a CNV region by minimizing noise characteristics of data through transformation of exome sequencing data using a four-step normalization transformation technique The purpose is to provide.

According to another aspect of the present invention, there is provided a normalization processing apparatus for normalizing input data in a gene unit repeat variation detection system, comprising: A coverage calculation module that calculates coverage data, extracts lead coverage data of only each exon target area based on each coverage data, calculates ratio data of lead coverage data, and a ratio calculation module that calculates ratio data based on ratio data calculated by the coverage calculation module A segmentation processing module for dividing ratio data of a target area into a plurality of segments having the same size or dividing test data into a plurality of segments having the same size with respect to a target area; of And a normalization conversion module that converts the coverage data of each segment to have a normalization distribution based on the standardized average value and the ratio data of each segment adjusted by the average value adjustment module and the average value adjustment module adjusted by the standardization average value to extract the effective ratio data .

The coverage calculation module calculates

)을 이용하여 비율 데이터를 산출하고, 보정계수(ω)를 곱하여 비율 데이터를 보정하되, 보정계수(ω)는 라이브러리 사이즈의 불균형성을 보정하기 위한 방법으로, 컨트롤 데이터에 존재하는 모든 타깃 영역의 커버리지 값을 모두 더한 값을 테스트 데이터에 존재하는 모든 타깃 영역의 커버리지 값을 모두 더한 값으로 나눈 값일 수 있다. ^{_{(R T e (i, j}} ) is the lead data of the test data coverage, R ^e _C (i, j) is the lead data of the control data coverage, ε is

) Is used to calculate the ratio data, and the ratio data is corrected by multiplying the correction coefficient by the correction coefficient (?). The correction coefficient (?) Is a method for correcting the unbalance of the library size, The value obtained by adding all of the coverage values may be a value obtained by adding all the coverage values of all the target areas existing in the test data.

(1≤i≤N_e, 1≤k≤n_s(i), 1≤j_k≤b_s))를 산출하되, 세그먼트 분할 시 b_f(i)=[(mod(n_b(i),b_s)+b_s)/2]+1에서 타깃 영역의 첫 번째 세그먼트를 시작하고, 양끝의 남는 영역(R_{T| C} ^e(i,j) (1≤j≤b_s(i)-1)과 R_{T| C} ^e(i,j) (b_j(i)≤j≤N_b(i)))의 데이터는 제외할 수 있다.Segmentation processing module is configured to rate data for each target region _{(e i (1≤i≤N e))} | a ^{_{(R T C e (i,}} j), b f (i) ≤j≤b l (i)) Ns (i) = with the same size _{(b s) [N b (} i) / b s] is divided by -1 segments, data rate, each segment having the same size (

_{(1≤i≤N e, 1≤k≤n s (i} ), 1≤j k ≤b s)) a, b _f (i) = [(mod _(b n (i during segmentation but calculated), _{b s) + b s) /} 2] to start the first segment of the target region from the +1 of the areas remains at both ends ^{_{(R T | C e (i}} , j) (1≤j≤b s (i) -1 ) And R _{T | C} ^e (i, j) (b _j (i) ≤j ≤N _b (i)).

The average value adjustment module calculates

(1? K? N _s (i)) to calculate the average value of each segment,

, (1≤k≤n_s(i))

, (1? K? N _s (i))

(Where, M (mR _s ^e) is a full segment _{S i, k (1≤i≤N e,} 1≤k≤n s ( the average of the average value, which straddle _{i)), SD (mR s} e) is a total segment _{S i, k (1≤i≤N e,} 1≤k≤n s (i)) calculates a standardized mean value by the standard deviation) of the mean value which straddle and, when the calculated normalized average value is negative at the standardized average value The standardized average value can be corrected by subtracting the minimum value of the standardized average value.

The normalization conversion module can extract the effective ratio data by converting the data of each position in each segment to have a normal distribution based on the standardized average value and the standard deviation of the ratio data of each segment.

According to an aspect of the present invention, there is provided a gene unit repeat mutation detection system, comprising: a normalization processing unit for calculating ratio data using two input data including test data and control data, And a CNV extraction device for extracting a gene unit repeat variation based on the input data normalized by the normalization processing device, wherein the normalization processing device comprises: a normalization processing device for normalizing the data, The lead coverage is calculated and segmented based on the base unit lead coverage to divide the target region into a plurality of segments having the same size, adjust the average value of the read coverage data of each of the plurality of segments to the standardized average value, Each segment On the basis of the rate data and converts the coverage data of each segment to have a normalized distribution.

According to an aspect of the present invention, there is provided a method of normalizing input data using a normalization processing apparatus for detecting a gene unit repeat variation, the method comprising: inputting data including test data and control data; , Extracting the lead coverage data of each target area based on the coverage data, calculating ratio data of the lead coverage data, calculating the ratio of the target coverage area to the target area based on the calculated ratio data, Dividing the ratio data into a plurality of segments having the same size, or dividing the test data into a plurality of segments having the same size, adjusting the average value of the plurality of segmented segments to the standardized average value, The ratio data for each segment Georo converts the coverage data of each segment to have a normalized distribution comprises the steps of extracting an effective data rate.

In the step of calculating ratio data,

)을 이용하여 비율 데이터를 산출하는 단계 및 보정계수(ω)를 곱하여 비율 데이터를 보정하는 단계를 포함하되, 보정계수(ω)는 컨트롤 데이터에 존재하는 모든 타깃 영역의 커버리지 값을 모두 더한 값을 테스트 데이터에 존재하는 모든 타깃 영역의 커버리지 값을 모두 더한 값으로 나눈 값일 수 있다. ^{_{(R T e (i, j}} ) is the lead data of the test data coverage, R ^e _C (i, j) is the lead data of the control data coverage, ε is

), And correcting the ratio data by multiplying the correction coefficient (?), Wherein the correction coefficient (?) Is a value obtained by adding all the coverage values of all the target areas existing in the control data May be a value obtained by dividing the coverage value of all the target areas existing in the test data by the sum of all the coverage values.

(1≤i≤N_e, 1≤k≤n_s(i), 1≤j_k≤b_s))를 산출하되, 세그먼트 분할 시 b_f(i)=[(mod(n_b(i),b_s)+b_s)/2]+1에서 타깃 영역의 첫 번째 세그먼트를 시작하고, 양끝의 남는 영역(R_{T| C} ^e(i,j) (1≤j≤b_s(i)-1))과 R_{T| C} ^e(i,j) (b_j(i)≤j≤N_b(i)))의 데이터를 제외할 수 있다.Dividing a plurality of segment data rate for each target region _{(e i (1≤i≤N e))} (R T | C e (i, j), b f (i) ≤j≤b l ( i)) of Ns (i) = with the same size _{(b s) [N b (} i) / b s] is divided by -1 segments, data rate, each segment having the same size (

_{(1≤i≤N e, 1≤k≤n s (i} ), 1≤j k ≤b s)) a, b _f (i) = [(mod _(b n (i during segmentation but calculated), _{b s) + b s) /} 2] to start the first segment of the target region from the +1 of the areas remains at both ends ^{_{(R T | C e (i}} , j) (1≤j≤b s (i) -1 )) And R _{T | The data of C} ^e (i, j) (b _j (i) ≤j ≤N _b (i)) may be excluded.

The step of adjusting to the standardized average value may be performed by:

, (1≤k≤n_s(i))

, (1? K? N _s (i))

Calculating an average value of each segment using the equation

, (1≤k≤n_s(i))

, (1? K? N _s (i))

(Where, M (mR _s ^e) is a full segment _{S i, k (1≤i≤N e,} 1≤k≤n s ( the average of the average value, which straddle _{i)), SD (mR s} e) is a total segment _{s i, k (1≤i≤N e,} 1≤k≤n s (i)) to straddle when calculating a normalized mean value by the standard deviation) of the mean value and the calculated normalized average value is negative standardized mean value in And a step of correcting the standardized average value by subtracting the minimum value of the standardized average value.

In the step of extracting the effective ratio data, the effective ratio data can be extracted by converting the data of each basic position in each segment to have a normal distribution based on the standardized average value and the standard deviation of the ratio data of each segment.

According to another aspect of the present invention, there is provided a method for detecting a repeated unit of gene units, comprising the steps of: normalizing input data including test data, control data, and date data by a normalization processing apparatus; Wherein the step of normalizing the input data comprises the step of extracting a gene unit repeat variation based on the normalized input data in the step of normalizing the input data, Dividing the target area into a plurality of segments having the same size by segmenting based on the base unit lead coverage, adjusting the average value of the read coverage data of each of the plurality of segments to a standardized average value, The average value and each segment Based on the data rate of the agent can comprise a step of converting the coverage data of each segment to have a normalized distribution.

According to the present invention, the system and method for detecting a gene unit repeating mutation can minimize the noise characteristic of data by transforming exome sequencing data by applying a four-step normalization conversion technique.

In addition, the system and method for detecting a gene unit repeat variation can extract CNV regions of various shapes and sizes through scale-space filtering.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram for explaining a gene unit repeated variation detection system according to an embodiment of the present invention. FIG.
2 is a view for explaining the normalization processing apparatus of FIG.
FIG. 3 and FIG. 4 are views for explaining the CNV extracting apparatus of FIG. 1;
5 is a flow chart for explaining a method of detecting repeated gene unit variation according to an embodiment of the present invention.
6 is a flowchart for explaining the normalization processing step of FIG. 5;
FIG. 7 is a flowchart for explaining the CNV detection step of FIG. 5; FIG.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings in order to facilitate a person skilled in the art to easily carry out the technical idea of the present invention. . In the drawings, the same reference numerals are used to designate the same or similar components throughout the drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

The system and method for detecting a gene unit repeat variation according to an embodiment of the present invention minimizes noise characteristics of data by normalizing and transforming exome sequencing data and performs a genetic unit repeat displacement of various shapes and sizes CNV (Copy Number Variation) area.

Referring to FIG. 1, a gene unit repeat mutation detection system 100 according to an embodiment of the present invention includes a normalization processing unit 120 and a CNV extraction unit 140.

The normalization processing unit 120 normalizes the input data to minimize various noise effects due to GC content bias, library size effect, coverage bias in the exosome boundary region, and the like.

To this end, the normalization processing unit 120 receives at least one of test data and control data as input data.

At this time, if only the test data is input, the normalization processing unit 120 performs the normalization processing using only the read coverage data of the test data. When the test data and the control data are input, the normalization processing unit 120 performs normalization processing using ratio coverage data obtained from the coverage data of the test data and the control data. Hereinafter, the case where both test data and control data are used as input data will be described as an example.

The normalization processing unit 120 calculates a base-level lead coverage in each target area with respect to the input data, segments the calculated smoke-to-smoke lead coverage, adjusts the average value of the coverage data in each segment, Lt; RTI ID = 0.0 > normalization < / RTI > distribution.

2, the normalization processing apparatus 120 includes a coverage calculation module 122, a segmentation processing module 124, an average value adjustment module 126, and a normalization conversion module 128.

The coverage calculation module 122 calculates the coverage on a target basis using the input data. That is, the coverage calculation module 122 calculates coverage data (R _T , R _C ) using the result of alignment of test data and control data, which is exome sequencing data.

Coverage calculation module 122 may group all positions (j (1≤j≤N _b (i) from the coverage data (R _{T, C} R) calculated for each target region (e _i (1≤i≤N _e)) (R _T ^e (i, j) and R _C ^e (i, j)) in the current position Here, N _e denotes the total number of target areas, and N _b (i) denotes the length of the i-th target area.

,

)의 비율(ratio) 데이터(R_T|C ^e(i,j))를 산출한다.The coverage calculation module 122 calculates the coverage coverage data

,

(R _{T | C} ^e (i, j)).

At this time, the coverage calculation module 122 calculates ratio data of the lead coverage data (R _T ^e (i, j), R _C ^e (i, j)) using the following equation (1).

Where epsilon is 10 < ^{-3 &} gt ;.

The coverage calculation module 122 corrects the ratio data calculated for the library size correction of the two lead coverage data. That is, as shown in the following Equation 2, the coverage calculation module 122 corrects the ratio data by multiplying the ratio data R _{T | C} ^e (i, j) by the correction coefficient?

)이고, R_T ^e는 테스트 데이터에 존재하는 모든 타깃 영역의 커버리지 값을 모두 더한 값(

)이다.At this time, the correction coefficient ω = mR _C ^e / m R _T ^e , and R _C ^e is the sum of the coverage values of all the target areas existing in the control data (

), And R _T ^e is a value obtained by adding all the coverage values of all the target areas existing in the test data (

)to be.

The segmentation processing module 124 performs segmentation on the target area. That is, most of the actual exoskeleton boundaries have a lower coverage than the central portion of the exosomal region. Thus, a segmentation processing module 124 performs the segmentation for each target area eksom (e _i (1≤i≤N _e)).

(1≤i≤N_e, 1≤k≤n_s(i), 1≤j_k≤b_s)를 산출한다.The segmentation processing module 124 divides ratio data for each target area into a plurality of segments having the same size. That is, the segmentation-processing module 124, the ratio data for each target region (e _i (1≤i≤N _e)) (R _{T | C} ^e (i, j), (b _f (i) ≤j≤b by dividing the _l (i)) with _{Ns (i) = [N b} (i) / b s] -1 segments having the same size (b _s), data rate, each segment having the same size (

(1? I? N _e , 1? _K ? N _s (i), 1? J _{k? B s} ).

At this time, the segmentation processing module 124 starts the first segment of the target exosmotic region at b _f (i) = [(mod (n _b (i), b _s ) + b _s ) / 2] , and the remaining area of both ends ^{_{(R T | C e (i}} , j) (1≤j≤b s (i) -1) and ^{_{R T | C e (i,}} j) (b j (i) ≤j≤ N _b (i))) is excluded.

The average value adjustment module 126 adjusts the average value of each segment segmented by the segmentation processing module 124 to a normalized average value (hereinafter referred to as a standardized average value) for the entire target exome using T-Score.

At this time, the average value adjustment module 126 calculates an average value of each segment using the following equation (3).

(1? K? N _s (i))

The average value adjustment module 126 calculates the standardized average value of each segment using the following equation (4).

(1? K? N _s (i))

Here, M (mR _s ^e) is a full segment _{S i, k (1≤i≤N e,} 1≤k≤n s (i)) and the average of the average value, which straddle, SD (mR _s ^e) is a full segment Means a standard deviation of an average value over S _{i, k} (1? I? N _e , 1? _K ? N _s (i)).

The average value adjustment module 126 corrects to have a positive value when the calculated standardized average value is negative. That is, if the calculated standardized average value is negative, the average value adjustment module 126 corrects the standardized average value by subtracting the minimum value of the standardized average value from the standardized average value.

In this case, the minimum value of the normalized average value refers to the ^{_{min (tR s e (i,}} j)) (1≤i≤N e, 1≤k≤n s (i)).

The normalization conversion module 128 converts the coverage data of each segment to have a normalization distribution based on the standardized average value and ratio data of each segment.

(1≤i≤N_e, 1≤k≤n_s(i), 1≤j_k≤b_s))의 표준 편차를 기반으로 각 세그먼트 S_i,k(1≤i≤N_e, 1≤k≤n_s(i)) 내의 각 기본 포지션 (j(b_f(i)+(k-1)*b_s≤j≤b_f(i)+k*b_s))의 데이터가 정규 분포를 갖도록 변환한다. That is, the normalized conversion module 128 is normalized average value (tR _s ^e (i, k) and the ratio of each data segment ((

Based on the standard deviation _{(1≤i≤N e, 1≤k≤n s (i} ), 1≤j k ≤b s)) for each segment _{S i, k (1≤i≤N e,} 1≤k data of ≤n _s (i) each base position in _{a) (j (b f (i} ) + (k-1) * b s ≤j≤b f (i) + k * b s)) to have a normal distribution Conversion.

)를 추출한다.Thereby, the normalization conversion module 128 converts the effective ratio data (

).

The CNV extracting unit 140 calculates the CNV area based on the effective ratio data extracted by the normalization processing unit 120. At this time, the CNV extracting apparatus 140 calculates a CNV region using a scale-space filtering technique. Here, the scale-space filtering technique uses a fingerprint map obtained by extracting inflection point information from input data and a scale-space image obtained by converting input data into various resolution (multi-resolution) data, And extracts embedded events.

The CNV extracting apparatus 140 extracts the CNV shape included in the coverage data as an event using a scale space filtering technique. That is, the CNV extracting unit 140 tracks the value change of the continuous coverage data, and detects a region having a shape in which the coverage value is lower or higher than the average value in the sequence of the test data, ), And calculates these areas as CNV areas.

는 해상도 조절을 위한 스케일 파라미터를 의미한다.3 shows an example of extracting an event area using a scale space filtering technique. The lower part shows a scale space image for a given signal (the lowest part signal at? =? ₀ ), and the upper part shows a fingerprint of a scale space image (Finger Print Map). here,

Means a scale parameter for adjusting the resolution.

The vertical dotted line represents the fingerprinting process using the inflection point of the scale space image at σ (= σ ₀ ), and the horizontal arrow segment represents the process of extracting event information consisting of two neighboring inflection points at σ (= σ _k ) .

4, the CNV extracting apparatus 140 includes a hierarchy separating module 142, a fingerprint map generating module 144, a reference line adjusting module 146, and a CNV extracting module 148 .

)에 가우시안 컨볼루션(Gaussian Convolution)을 적용시켜 복수의 스케일 스페이스 이미지(scale-space image) 계층으로 분리한다. 이때, 가우시안 컨볼루션은 하기 수학식 5와 같다. 이후 표기법의 간략화를 위하여 유효 비율 데이터(

)를 c[i]로 표기하여 나타낸다.The hierarchy separating module 142 separates the validity ratio data extracted from the normalization processing device 120

) Into a plurality of scale-space image layers by applying a Gaussian convolution. At this time, the Gaussian convolution is expressed by Equation (5). In order to simplify the notation, effective ratio data (

) Is denoted by c [i].

Where k [i, k] is a scale space image, g [j, k] is a Gaussian kernel, and k ≪ / RTI > m is the window size of the Gaussian kernel, and σk can be experimentally specified as m = 3σk, σk = 10 ³ × (1.1) ^k as a scale parameter. σk = 10 ³ × (1.1) The ^k value may be determined in consideration of the size and time complexity of the CNV to be detected.

However, since the convolution in the time domain is the same as the product in the frequency domain, it is possible to obtain each scale-space image layer c [i, k] by applying Discrete Fourier Transform to reduce the complexity of the calculation have.

The fingerprint map generation module 144 may obtain zero crossings of the second derivatives for a plurality of layers separated by the layer separation module 142. The second derivative c "[i, k] of the kth-order scale space image c [i, k] can be approximated as shown in Equation 6. The zero crossing signal z [i, k] of the second derivative is Can be defined as shown in Equation (7).

That is, z [i, k] has a value of 1 when the value of the second derivative changes from minus to plus, with the zero crossing point as its starting point, and conversely, If it changes to negative, it has a value of -1. Of course, if i is not a zero crossing, then z [i, k] has a value of zero.

The fingerprint map generation module 144 may generate a single fingerprint map by displaying each zero crossing obtained for a plurality of layers. That is, when z [i, k] has a value of +1 or -1, the zero crossing point is displayed in the fingerprint map. At this time, when z [i, k] has a value of 0, no indication is made in the fingerprint map.

The reference line adjustment module 146 may acquire a reference line of each of a plurality of layers using values of a scale space image included in a predetermined range for each of a plurality of layers. The predetermined range may be an average value and a standard deviation value As shown in FIG. Here, if the average value of the scale space image is m (k) and the standard deviation value of the scale space image is delta (k), the predetermined range can be set to m (k) ± wδ (k). w included in the coefficient of? (k) can be set to an arbitrary value as a weight, and the values of m (k) and? (k) can be obtained from equations (8) and (9), respectively.

As an embodiment of the present invention, the weight w will be set to 2. The baseline adjustment module 146 again obtains an average value and a standard deviation value for the values included in m (k) ± 2δ (k), except for values outside the range of m (k) ± 2δ . The mean value and the standard deviation value at this time are referred to as the effective mean value m * (k) and the effective standard deviation value δ * (k) in order to distinguish the above-mentioned mean value m (k) from the standard deviation value δ (k).

The reference line adjustment module 146 calculates the reference lines m * (k), m * (k) + dδ * of each of the plurality of layers using the effective average value and the effective standard deviation value for the values of the scale space image included in the predetermined range, (k), m * (k) - d? * (k).

Here, d included as a coefficient of? * (K) is a weight, and may be various values depending on the case.

Further, the reference line adjustment module 146 may obtain a reference line for all of the hierarchy of the plurality of hierarchy of scale space image layers, but at least two of the non-zero zero crossing signals z [i, k] The reference line may be obtained only for the layer having the value. In other words, in some cases, the baseline adjustment module 146 may obtain a baseline only for a layer having two or more zero crossings.

The CNV extraction module 148 may determine an area for detecting CNV from the respective baselines obtained at the baseline adjustment module 146. [ The CNV extraction module 148 may determine the CNV using the effective mean value and the effective standard deviation value according to one embodiment of the present invention. The region for detecting CNV can be determined using the reference line m * (k) ± 3δ * (k) to which 3 is applied to the d value. There may be a plurality of regions for detecting the CNV at this time, or there may be no region. Let [m, k, um, k] denote the mth region of kth layer for detecting CNV. Where lm, k and um, k are the ranges of i of the zero crossing signal z [i, k] indicated in the fingerprint map and can be expressed as lm, k ≤ i ≤ um, k. It is an embodiment of the present invention that 3 is applied to the d value of the reference line to determine the area for detecting CNV, and the d value is not limited to 3.

However, the area for detecting CNV can be determined based on the following three conditions.

First, i corresponding to a region pre-designated as a region for detecting CNV at k-th layer, that is, at k + 1 to L-1 layers, should not be included in [lm, k, um, k].

Second, the product of z [lm, k, k] and z [um, k, k] must be less than 0 and z [i, k ] Should all be zero. That is, one of z [lm, k, k] and z [um, k, k] is +1 and the other is -1, and both values must be zero.

In the same expression, a value when i = lm, k in the fingerprint map and a value when i = um, k are represented by "o" and the other is represented by "+" There should be no indication.

Finally, the mean value of the scale space image contained in [lm, k, um, k] should deviate from the given reference line m * (k) ± 3δ * (k). Here, the average value of the scale space images included in [lm, k, um, k] can be expressed as shown in Equation (10).

Referring to FIG. 5, the method of detecting a gene unit repeat variation according to an embodiment of the present invention includes a normalization processing step S100 and a CNV detection step S200.

In the normalization processing step S100, at least one of test data and control data is input as input data. At this time, if only the test data is input, the normalization processing unit 120 performs the normalization processing using only the read coverage data of the test data.

In the normalization processing step S100, when test data and control data are input, normalization processing is performed using ratio coverage data obtained from the coverage data of the test data and control data.

The normalization processing step S100 calculates base-level lead coverage in each target area for the input data, segments the calculated delay unit lead coverage, adjusts the average value of the coverage data in each segment, Lt; RTI ID = 0.0 > normalization < / RTI > distribution.

Referring to FIG. 6, the normalization process S100 includes a lead coverage calculation step S120, a segmentation step S140, a coverage data average value adjustment step S160, (S180).

In the lead coverage calculation step (S120), coverage is calculated on a target basis using input data. That is, in the lead coverage calculation step (S120), the coverage data (R _T , R _C ) is calculated by using the result of alignment of test data and control data, which is exome sequencing data.

In the lead coverage calculation step S120, lead coverage data at all positions of each target area is extracted from the calculated coverage data, and ratio data of the extracted lead coverage data is calculated. Here, the ratio data means the ratio of test data and control data.

In the read coverage calculation step (S120), the ratio data calculated for correcting the library size of the read coverage data is corrected. At this time, in the lead coverage calculation step (S120), the ratio data is multiplied by the correction coefficient to correct the ratio data

In the segmentation step S140, segmentation is performed on the target area. That is, since most of the exoskeleton boundaries have a lower coverage than the center of the exosomal region, segmentation is performed on each of the target exosomal regions in the segmentation step S140.

In the segmentation step S140, the ratio data for each target area is divided into a plurality of segments having the same size. That is, in the segmentation step S140, the ratio data for each target area is divided into a plurality of segments having the same size, and the ratio data having the same size is calculated. At this time, in the segmentation step (S140), data of regions remaining at both ends of the target exosmotic region are excluded when the segment is segmented.

In the average value adjustment step S160, the average value of each segment segmented in the segmentation step S140 is adjusted to a normalized average value (hereinafter referred to as a standardized average value) for the entire target exome using T-Score. At this time, in the average value adjusting step (S160), when the standardized average value is negative, it is corrected to have a positive value. That is, in the average value adjusting step S160, if the standardized average value is negative, the minimum value of the standardized average value is subtracted from the standardized average value, and the standardized average value is corrected to have a positive value.

In the converting step S180, the coverage data of each segment is converted to have a normalized distribution based on the normalized average value and the ratio data of each segment adjusted in the average value adjusting step S160. That is, in step S180, the basic position data in each segment is transformed to have a normal distribution based on the standardized average value and the standard deviation of the ratio data of each segment, and effective ratio data is extracted.

The CNV detection step S200 calculates a CNV area based on the effective ratio data extracted in the normalization processing step S100. At this time, the CNV detection step (S200) extracts the CNV shape included in the coverage data as an event using a scale space filtering technique. That is, the CNV detection step (S200) tracks the value change of the continuous coverage data, and detects an area having a shape in which the coverage value is lower than or higher than the average value in the sequence of the test data, ), And calculates these areas as CNV areas.

Referring to FIG. 7, the CNV detection step S200 includes a layer separation step S220, a fingerprint map generation step S240, a reference line adjustment step S260, and a CNV extraction step S280.

In the hierarchical separation step S220, Gaussian convolution is applied to the effective ratio data extracted in the normalization processing step S100 to separate into a plurality of scale-space image layers.

In the fingerprint map generation step S240, the zero-crossing point of the second derivative may be acquired for each of the plurality of layers separated in the layer separation step S220. In the fingerprint map generation step S240, a single fingerprint map may be generated by displaying the respective zero crossings obtained for a plurality of layers. That is, when the zero crossing signal has a value of +1 or -1, the zero crossing point is displayed on the fingerprint map. At this time, when the zero-crossing signal has a value of 0, no indication is made on the fingerprint map.

In the baseline adjustment step S260, the baseline of each of the plurality of layers can be obtained by using the values of the scale space image included in the predetermined range for each of the plurality of layers, and the preset range includes the average value and the standard deviation value As shown in FIG.

In the baseline adjustment step S260, the baseline of each of the plurality of layers is obtained using the effective average value and the effective standard deviation value of the values of the scale space image included in the predetermined range.

In the baseline adjustment step S260, the baseline may be obtained for all of the plurality of hierarchy-separated scale space image layers. However, the baseline may be obtained for only a hierarchy having at least two zero- It can also be obtained. In other words, in some cases, the baseline may be obtained only for the layer having two or more zero crossings in the baseline adjustment step (S260).

In the CNV extracting step (S280), an area for detecting CNV from each of the reference lines obtained in the reference line adjusting step (S260) can be determined. The CNV detection region can be determined using the effective average value and the effective standard deviation value. In the CNV extraction step S280, an area for detecting CNV is determined only for a scale space image layer having at least two zero crossings . There may be a plurality of regions for detecting the CNV at this time, or there may be no region.

While the present invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but many variations and modifications may be made without departing from the scope of the present invention. It will be understood that the invention may be practiced.

100: gene unit repeat mutation detection system
120: Normalization processing unit 122: Coverage calculation module
124: Segmentation processing module 126: Average value adjustment module
128: normalization conversion module 140: CNV extraction device
142: hierarchical separation module 144: fingerprint map generation module
146: Reference line adjustment module 148: CNV extraction module

Claims (12)

A normalization processing apparatus for normalizing input data of a gene unit repeat variation detection system,
The coverage data is calculated based on two pieces of input data including the result of lead sorting of test data and control data, and lead coverage data of each exon target area is extracted based on each coverage data, and ratio data of the lead coverage data A coverage calculation module for calculating a coverage;
A segmentation process module for dividing the ratio data for the target area into a plurality of segments having the same size based on the ratio data calculated by the coverage calculation module or dividing the test data into a plurality of segments having the same size with respect to the target area, ;
An average value adjustment module for adjusting an average value of a plurality of segments divided by the segmentation processing module to a standardized average value; And
And a normalization conversion module that converts the coverage data of each segment to have a normalization distribution based on the normalized average value and ratio data of each segment adjusted by the average value adjustment module to extract effective ratio data,
Wherein the normalization conversion module comprises:
And converting the data of each position in each segment to have a normal distribution based on the standardized average value and the standard deviation of ratio data of each segment to extract effective ratio data. delete delete delete delete delete A method for normalizing input data using a normalization processing apparatus for detecting a gene unit repeat variation,
Calculating coverage of a target unit based on input data including test data and control data;
Extracting lead coverage data of each target area based on the coverage data;
Calculating ratio data of the read coverage data;
Dividing the ratio data for the target area into a plurality of segments having the same size based on the ratio data or dividing the test data into a plurality of segments having the same size;
Adjusting an average value of the segmented plurality of segments to a standardized average value; And
And converting the coverage data of each segment to have a normalization distribution based on the standardized average value and ratio data of each segment to extract effective ratio data,
In the step of extracting the effective ratio data,
Wherein the data of each base position in each segment is transformed to have a normal distribution based on the standardized mean value and the standard deviation of the ratio data of each segment to extract effective ratio data. delete delete delete delete delete

Images