A hybrid DNA evidence analysis method based on STR genotyping profiles

Abstract

The invention discloses a hybrid DNA evidence analysis method based on STR genotyping map, comprising: step 1: obtaining data feature information; step 2: fitting the feature information; step 3: determining whether the number of contributors K and the suspect's STR map are obtained, and setting mutually exclusive propositions; step 4: calculating the likelihood ratio LR under the parameter conditions corresponding to the maximum likelihood value, and sorting the genotype combinations; step 5: judging the supported propositions according to the LR; step 6: considering whether the contributor's relatives participate in the map; step 7: database retrieval, sorting the likelihood ratio to obtain candidate individuals, and repeating steps 3-4 to determine the screening individual; the method of the invention comprehensively considers various complex situations that may be involved in the case, provides individual genotype splitting and accurate likelihood ratio LR values ​​for diversified scenarios, objectively evaluates the reliability and scientificity of the LR value, and provides a detailed strategy for how to efficiently use the DNA database for suspect retrieval.

Description

Technical Field

[0001] The invention relates to the technical field of forensic genetics, and in particular to a mixed DNA evidence analysis method based on STR genotyping maps. Background Art

[0002] The mixed DNA pattern evidence analysis method based on short tandem repeats (STR) refers to the technology of detecting mixed STR genotyping patterns of multiple individual DNA in forensic biological samples by using STR genetic markers on human genomic DNA sequences, and analyzing and interpreting the patterns. Its purposes include: 1) separating individual genotypes that may participate in the composition of the DNA mixture from the mixed genotyping pattern, a process called "deconvolution" or "genotype splitting". 2) Under a certain assumption Hp, calculate the likelihood value of a certain individual genotype participating in the composition of the pattern, and compare it with the likelihood value calculated under the mutually exclusive assumption Hd of Hp to obtain the likelihood ratio LR, and evaluate the probative force of the individual's participation in the DNA mixture evidence, that is, "evidence strength evaluation". 3) In the absence of a reference individual, the individual genotype obtained by deconvolution is put into the relevant DNA data for retrieval and comparison, find matching individuals, and calculate the likelihood rate of the matching individual as a contributor to the DNA mixture, and evaluate the possibility that the individual is a real contributor to the DNA mixture, thereby providing DNA evidence and corresponding evidence strength evaluation for case detection and case trial.

[0003] The existing mixed DNA profile analysis method is to model the peak height of the STR profile, and to achieve deconvolution of the STR mixed profile and calculation of the strength of evidence by modeling the parameters such as the template amount, degradation level, locus-specific amplification efficiency and parallel amplification efficiency of each contributor to the DNA mixture. At the same time, the current common search method is to use the deconvolution results as query conditions, perform homotypic comparison or comparison of a few loci in the DNA database to obtain the search results.

[0004] This method has the following problems: 1) There is a priori parameter dependence, that is, it is necessary to pre-set parameters such as peak height saturation value, shadow peak ratio, insertion rate and peak height fluctuation variance. The acquisition of these parameters not only requires a large amount of experimental data support, but is also easily affected by factors such as experimental conditions and instrument errors. The generalization ability of the mathematical model is limited to a certain extent; 2) There are differences in genetic backgrounds between people in different regions and ethnic groups in my country. This difference will affect the accuracy of likelihood calculation. The existing calculation model does not use Chinese localized data for likelihood correction. 3) In the current DNA mixture analysis technology scheme, there is a lack of effective and targeted biotechnology system application method for DNA mixture analysis, that is, whether the genetic markers and technical methods used meet the needs of mixed DNA analysis. 4) The mixed DNA evidence analysis method for pathological tissue STR profiles that may have pathological gene mutations can only process DNA profile data from tumor tissues. It is impossible to provide evidence strength evaluation for the STR profile of DNA mixtures at the forensic evidence scene. At the same time, the former method cannot consider the impact of interference phenomena such as allele insertion and shadow peaks on profile analysis. 5) The existing mixed DNA kinship analysis model has limited analytical capabilities and can only analyze kinship within the second level, and can only be analyzed when there are reference pairs of kinship individuals. It is impossible to infer whether there are pairs of kinship individuals participating in the composition of mixed DNA without reference individuals, nor can it evaluate the impact of kinship on the evaluation of evidence strength. 6) The current modeling and analysis methods have limited capabilities for joint analysis of spectra, and cannot perform joint analysis and data interpretation on STR spectra of DNA mixtures of biological samples from different times, places, and cases, and cannot meet the needs of DNA string case analysis for roving crimes. 7) Existing database retrieval methods are inefficient and often cannot be directly applied to the retrieval of mixed DNA spectra, resulting in a low utilization rate of mixed DNA spectra and an inability to give full play to their role in case investigation. Summary of the invention

[0005] The present invention aims at solving the problems in the prior art and provides a mixed DNA evidence analysis method based on STR genotyping map.

[0006] The technical solution adopted by the present invention is: a hybrid DNA evidence analysis method based on STR genotyping map, comprising the following steps:

[0007] Step 1: Obtain STR profiles of N mixed DNA samples to obtain data feature information, N ≥ 1;

[0008] Step 2: Fit the feature information in step 1. If the fitting result error reaches the set threshold, go to step 3, otherwise exit;

[0009] Step 3: Determine the number of contributors K and determine whether the STR profile of the target individual is obtained. If yes, go to step 4; otherwise, set the proposition Hd, where Hd is N mixed profiles composed of K unrelated individuals. According to the assumptions, calculate the corresponding parameters and maximum likelihood values ​​when the likelihood value is the largest under the given assumptions, and sort the genotype combinations; perform DNA database search based on the sorting results;

[0010] Step 4: Set mutually exclusive propositions Hp and Hd; Hp is N mixed graphs consisting of target individuals and K-1 irrelevant individuals; Hd is N mixed graphs consisting of K irrelevant individuals;

[0011] According to the likelihood function under the given assumptions Hp and Hd and the corresponding parameters, the corresponding parameter value and the maximum likelihood value when the likelihood value is the largest under the given assumptions Hp and Hd are calculated, the likelihood ratio LR under the parameter conditions is calculated, and the genotype combinations are sorted according to the LR value;

[0012] Step 5: If LR>1, then the proposition Hp is supported and the process goes to step 6. If LR<1, then the proposition Hd is supported and the process goes to step 7.

[0013] Step 6: Consider whether the contributor's relatives participate in the spectrum. If yes, set the mutually exclusive hypothesis, calculate the corresponding parameter and maximum likelihood value when the likelihood value is the largest, and calculate the likelihood ratio LR' and LR" under the parameter conditions, and perform a non-contributor test; otherwise, directly perform a non-contributor test;

[0014] Step 7: Traverse each individual typing in the DNA database, and take each individual as the target individual. Under the given assumptions Hp and Hd, calculate the likelihood ratio LR'' without considering the peak height, and sort the likelihood ratio LR'' from large to small; take the first d as candidate suspicious matching individuals, repeat steps 3-4, and the individuals with LR>1 are regarded as matching individuals.

[0015] Furthermore, the mutually exclusive assumptions in step 6 are as follows:

[0016] Set the propositions Hp' and Hd', where Hp' is N mixed graphs consisting of the target individual and K-1 unrelated individuals, and Hd' is N groups of mixed graphs consisting of the relatives of the target individual and K-1 unrelated individuals;

[0017] Under this assumption, find parameters for maximum likelihood estimation, calculate the maximum likelihood value, and obtain the likelihood ratio LR' under this assumption and parameter conditions; if LR'>1, then the proposition Hp' is supported, and if LR'<1, then the proposition Hd' is supported;

[0018] Set propositions Hp” and Hd”, where Hp” is N sets of mixed graphs consisting of target individuals, relatives of target individuals, and K-2 unrelated individuals; Hd” is N sets of mixed graphs consisting of random individuals, relatives of random individuals, and K-2 unrelated individuals;

[0019] Under this assumption, we search for parameters to perform maximum likelihood estimation, calculate the maximum likelihood value, and obtain the likelihood ratio LR” under this assumption and parameter conditions; if LR”>1, then the proposition Hp” is supported; if LR”<1, then the proposition Hd” is supported.

[0020] Furthermore, the data feature information in step 1 includes locus vectors, allele vectors and peak height vectors of the mixed DNA map.

[0021] Furthermore, in step 2, the data feature information is fitted using normal distribution, lognormal distribution and gamma distribution in turn.

[0022] Furthermore, before calculating the likelihood ratio LR in step 4, the genotype combination vector is first obtained according to the N groups of map allele vectors, and the probability of each genotype combination in the genotype combination vector appearing under the given assumptions Hp and Hd, that is, the probability of the corresponding peak height, is calculated.

[0023] Furthermore, step 4 also includes: calculating the posterior probability of all genotype combinations at a certain locus according to the parameter value corresponding to the maximum likelihood value under the given assumptions Hp and Hd, and sorting them to obtain the desired sorting result.

[0024] Furthermore, in step 7, the calculation method of LR'' without considering the peak height is as follows:

[0025] In the atlas On the basis of the allele vector at locus m and genotype combination vector Computation exists in But not in The probability p of the occurrence of allele Q in dropout ;

[0026] Exists in But not in The probability of occurrence of an allele in drop-in ;

[0027] Existing in Also exists in The probability of an allele appearing in is 1-p dropout ;

[0028] Combine the N STR profile data from all loci and calculate the conditional probability of observing the profile under a given hypothesis;

[0029]

[0030] Where: Given the assumption Hp, p drop-in , p dropout Spectra observed under the conditions The conditional probability of Given the assumption Hd, p drop-in , p dropout Spectra observed under the conditions The conditional probability of .

[0031] Furthermore, the non-contributor testing process in step 6 is as follows:

[0032] According to the national population STR frequency table, genotypes were simulated at each locus in turn, and the genotypes at different loci were combined to generate a simulated random individual typing data set. Each individual was regarded as a target individual, and the parameter value and maximum likelihood value corresponding to the maximum likelihood value under the given assumptions of Hp and Hd were calculated, and the likelihood ratio LR under the parameter conditions was calculated;

[0033] The beneficial effects of the present invention are:

[0034] (1) The method of the present invention comprehensively considers various complex situations that may be involved in the case, and provides individual genotype splitting results and likelihood ratio (LR) values ​​for various scenarios, such as the need for joint analysis of multiple maps and the joint participation of relatives in map contributions;

[0035] (2) The method of the present invention does not need to rely on preliminary experiments on single-source samples of different concentrations for parameter calibration, which makes it more adaptable and practical in the field of mixed DNA analysis;

[0036] (3) The present invention uses a non-contributor test to determine the possibility of random unrelated individuals participating in the graph and provide a case-by-case dynamic system performance evaluation to objectively evaluate the reliability and scientificity of the LR value.

[0037] (4) The present invention effectively calls the DNA database and uses LR calculation to combine the hybrid map with database retrieval, thereby improving the efficiency of analyzing and using DNA evidence in court. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0039] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0040] like Figure 1As shown, a hybrid DNA evidence analysis method based on STR genotyping profile includes the following steps:

[0041] Step 1: Obtain STR profiles of N mixed DNA samples and obtain data feature information;

[0042] Obtain STR profiles of N mixed DNA samples, extract DNA from N samples, perform DNA amplification using polymerase chain reaction, and perform STR typing using capillary electrophoresis. The data feature information includes the locus vector, allele vector, and peak height vector of the mixed DNA profile. The locus vector of the profile, is the Nth map locus vector; allele vector is the Nth map allele vector; peak height vector is the Nth map allele peak height vector.

[0043] Step 2: Fit the feature information in step 1. If the fitting result error reaches the set threshold, go to step 3, otherwise exit;

[0044] The Nth map locus vector is represented as Where L N,m is the mth locus on the Nth map. The Nth map peak height vector in For L N,m The vector of allele peak heights on . Y N,m,q For L N,m The peak height of the qth allele.

[0045] Get the Y at the mth locus of N maps N,m,q The data sets are formed by normal distribution, lognormal distribution and gamma distribution. If the error sum of the fitting results is less than 0.001, the likelihood ratio can be calculated and the genotype can be split by the following method.

[0046] Step 3: Determine the number of contributors K and determine whether the STR profile of the target individual is obtained; if so, proceed to step 4; if not, set the proposition Hd, where Hd is N mixed profiles composed of K unrelated individuals; according to the assumptions, calculate the corresponding parameters and maximum likelihood values ​​when the likelihood value is the largest under the given assumptions, and sort the genotype combinations; perform DNA database search based on the sorting results;

[0047] Based on N STR profile data, the number of contributors K is determined using the maximum allele counting method or the maximum likelihood estimation method.

[0048] Step 4: Set mutually exclusive propositions Hp and Hd; Hp is N mixed graphs consisting of target individuals and K-1 irrelevant individuals; Hd is N mixed graphs consisting of K irrelevant individuals;

[0049] According to the likelihood function under the given assumptions Hp, Hd and corresponding parameters, the parameter value and the maximum likelihood value corresponding to the maximum likelihood value under the given assumptions Hp, Hd are calculated, the likelihood ratio LR under the parameter conditions is calculated, and the genotype combinations are sorted according to the LR value.

[0050] Before calculating the likelihood ratio LR, we first obtain the genotype combination vector based on the N map allele vectors, and calculate the probability of each genotype combination in the genotype combination vector under the given assumptions Hp and Hd, that is, the probability of the corresponding peak height.

[0051] The calculation process is as follows:

[0052] Nth map allele vector in For L N,m The allele vector on . where a N,m,q For L N,m Get all the qth alleles. N,m,q , remove duplicate values, and add the possible missing allele Q to obtain the allele vector at locus m a m,q is the qth allele a at the mth locus. Repeat twice to select The alleles in the genotype vector are repeated K times to generate a genotype combination vector. g m,p is the pth genotype combination at the mth locus, containing 2K alleles.

[0053] Calculate the appearance of g m,p The probability of the corresponding peak height is divided into the following cases:

[0054] (1) For g m,p Each allele a in can be defined as the allele a from contributor k at any autosomal locus m, with a peak height y m,a,k The probability f(y m,a,k ) is calculated as follows:

[0055] f(y m,a,k ) = gamma(y m,a,k |ω -2 ∑ k∈K mx k n a,k ,μω2 )

[0056] Where μ is the expected peak height, ω is the coefficient of variation of the peak height, mx k is the proportion of the kth contributor, n a,k is the number of allele a of the kth contributor.

[0057] (2) When considering the occurrence of stutter, given g m,p , where each allele a will produce stutter. Generate a stutter vector where g m,p,stutter,o The oth stutter produced by the pth genotype combination at the mth locus is generated by the corresponding allele a. m,p,stutter,o Peak height Probability The calculation method is as follows:

[0058]

[0059] Where δ is the stutter peak height ratio.

[0060] Peak height m,a,k The probability f(y m,a,k ) is calculated as follows:

[0061] f(y m,a,k ) = gamma(y m,a,k |(1-δ)ω -2 ∑ k∈K mx k n a,k ,μω 2 )

[0062] (3) Determine whether the peak height of the allele is linearly correlated with the length of the allele fragment. For each map, change Y N,m,q The logarithm of the mean and a N,m,q The average fragment length is linearly fitted. If the absolute value of the slope of the fitting equation is greater than 0.9995, there is no need to consider the addition of the degradation parameter ε; if the degradation coefficient is less than 0.9995, the addition of the degradation parameter ε is required.

[0063] When considering the addition of degradation parameter ε, y m,a,k Probability f(y m,a,k ) is calculated as follows:

[0064]

[0065] The probability is calculated as follows:

[0066]

[0067] Where: S m,a is the fragment length of allele a at locus m, S min It is the allele length with the shortest fragment length in the population for all loci in the STR profile. g m,p,stutter,o The length of the fragment.

[0068] (4) When g m,p The genotype includes the missing allele Q. Then, at any autosomal locus m, the allele Q from contributor k has a peak height y m,Q,k The probability of ignoring degradation parameters is calculated as follows:

[0069]

[0070] f(y m,Q,k ) = gamma(y m,Q,k |ω -2 ∑ k∈K mx k n Q,k ,μω 2 )

[0071] The calculation method when considering the degradation parameters is as follows:

[0072]

[0073] Among them, AT is the analysis threshold of STR spectrum, f(y m,Q,k <AT) is y m,Q,k <Conditional probability when AT. S m,Q is the length of the fragment corresponding to the longest allele at locus m. Q,k is the number of the kth contributor allele Q.

[0074] (5) When Alleles in g m,p If the allele appears in the range of y, the allele is a drop-in. Then the peak height y of the drop-in on any autosomal locus m is m,drop-in The probability calculation method is:

[0075] f(y m,drop-in )=p c *f(drop-in|λ)*f drop-in

[0076]

[0077] Among them, p cis the probability of drop-in, λ is the parameter of exponential distribution, which is a constant. f(drop-in|λ) is the probability of drop-in alleles appearing under the given parameter λ. drop-in is the frequency of the drop-in allele. e is a natural constant.

[0078] (6) Calculate the g under given assumptions Hp and Hd m,p The probability of a genotype combination. At any autosomal locus m, under a given hypothesis H, the observed genotype combination g m,p The probability p(g m,p |H) is the product of the genotype frequencies within the combination.

[0079] The probability of a heterozygous genotype is the product of the frequencies of the two alleles times 2. The probability of a homozygous genotype is the square of the frequency of the allele.

[0080] If there is a possible missing allele Q in the genotype, the frequency of allele Q is 1-the sum of the frequencies of all detected alleles at the locus.

[0081] (7) When considering the existence of subpopulation structure within the corresponding population, the genotype combination g m,p The probability p(g m,p |H) needs to be corrected with the Fst value, represented by η.

[0082] (8) Combine all loci and N STR profile data. Give the likelihood function of LR under the assumptions and corresponding parameters. Obtain the corresponding parameter values ​​and the maximum likelihood value when the likelihood value under the conditions of Hp and Hd is the largest, and calculate LR.

[0083] First, under the assumptions of H and parameters, we observe The conditional probability

[0084] in, is N graph vectors, E N represents the Nth spectrum; H is the general name of the assumptions Hp and Hd, is the parameter vector to be estimated, is the mixing ratio vector of N spectra, is the mixing ratio vector of the Nth spectrum, mx k A vector representing the proportion of the DNA template amount of k contributors in the Nth map. is the expected peak height vector of N spectra, μ N Represents the expected peak height vector of the Nth spectrum. is the peak height variation coefficient vector of N spectra, ω NRepresents the peak height variation coefficient vector of the Nth spectrum. is the degradation parameter vector of N spectra, ε N Represents the degradation parameter vector of the Nth profile. is the stutter peak height ratio vector of N spectra, δ N Represents the stutter peak height ratio vector of the Nth spectrum. p(g m,p |H,η) is the observed g under given H and η conditions. m,p The conditional probability of . For the genotype combination g m,p and parameters p c , λ, in the nth STR profile, the peak height Y of the qth allele is observed at locus m. N,m,q The conditional probability of . m represents any autosomal locus m, and there are M loci in total.

[0085] (9) The LR calculation method is as follows: (LR is calculated under the condition of uncertain parameters)

[0086]

[0087] Perform maximum likelihood estimation on the unknown parameters under the assumptions Hp and Hd, and obtain the maximum likelihood value LR MLE (The LR used in the final judgment condition in step 4 of the present invention is the maximum likelihood value), and the calculation method is as follows:

[0088]

[0089] in, For the assumption that Hp and given Under these conditions, the parameter θ p The likelihood function of ; For the hypothesis θ d and given Under these conditions, the parameter θ d The likelihood function of .

[0090] If the calculated LR>1 or Log 10 When LR>0, the hypothesis Hp is supported, that is, the N-group mixed graph consists of the target individual and K-1 irrelevant individuals. 10 When LR is less than 0, the hypothesis Hd is supported, that is, the N groups of mixed graphs are composed of K unrelated individuals.

[0091] (10) Optimization parameter search results max under fixed Hp and Hd θ , calculate g m,p The posterior probability p(g m,p|max θ ,H,y), and sort them according to the calculated probability values. The sorting results under the assumption that Hp can be used to analyze the typing of another contributor involved in DNA. The sorting results under Hd can be used to assist in determining the typing of the contributor.

[0092]

[0093] Among them, p(g m,p |max θ ,H,y) is the observed genotype combination g under the premise of hypothesis H, maximum likelihood estimated parameter θ value and peak height y m,p The posterior probability of N,m,q |g m,p ,max θ ) is the genotype combination g m,p And the given optimal parameter search result max θ Under the conditions of N,m,q The posterior probability of p(g m,p |H,η) is the observed genotype combination g under the assumption of proposition H and parameter η m,p The posterior probability of .

[0094] If the target individual's DNA and corresponding STR profile are not available, only the corresponding parameter value when the likelihood value under Hd is the largest is obtained. θ , calculate g m,p The posterior probability p(g m,p |max θ ,H,y), and sort them according to the calculated values. The sorting results can be used for database retrieval.

[0095] Step 5: If LR>1, then proposition Hp is supported and go to step 6. If LR<1, then proposition Hd is supported and go to step 7.

[0096] Step 6: Consider whether the contributor's relatives participate in the map. If yes, set the mutually exclusive hypothesis, calculate the corresponding parameter and maximum likelihood value when the likelihood value is the largest, calculate the likelihood ratio LR' and LR" under the parameter condition, and perform a non-contributor test; if no, directly perform a non-contributor test.

[0097] When the target individual's relatives need to be considered as contributors to the spectrum, mutually exclusive propositions Hp' and Hd' are set. When the target individual and the target individual's relatives need to be considered as common contributors to the spectrum, mutually exclusive hypotheses Hp", Hd" are set.

[0098] (1) Set the propositions Hp' and Hd', where Hp' is N mixed graphs consisting of the target individual and K-1 unrelated individuals, and Hd' is N groups of mixed graphs consisting of the relatives of the target individual and K-1 unrelated individuals;

[0099] At this time, under the Hd' proposition and parameter θ, the probability of observing the STR spectrum is for:

[0100]

[0101] Under this assumption, find parameters for maximum likelihood estimation, calculate the maximum likelihood value, and obtain the likelihood ratio LR' under this assumption and parameter conditions; if LR'>1, then the proposition Hp' is supported, and if LR'<1, then the proposition Hd' is supported;

[0102] (2) Set the propositions Hp” and Hd”. Hp” is N groups of mixed spectra consisting of the target individual and its relatives and K-2 unrelated individuals; Hd” is N mixed spectra consisting of random individuals and their relatives and K-2 unrelated individuals. At this time, under the condition of H” proposition (H” is a general term for Hp” and Hd”) and the parameter θ, the probability of observing the STR spectrum is:

[0103]

[0104] Among them, κ=[κ0,κ1,κ2] is the vector of adjusted genotype probability under different kinship relationships. When the kinship relationship is father and son, κ=[0,1,0]; when the kinship relationship is full siblings, κ=[1 / 4,1 / 2,1 / 4]; when the kinship relationship is uncle and nephew, grandparents and grandchildren, and half siblings, κ=[1 / 2,1 / 2,0]; when the kinship relationship is cousins, κ=[3 / 4,1 / 4,0].

[0105] Under this assumption, we search for parameters to perform maximum likelihood estimation, calculate the maximum likelihood value, and obtain the likelihood ratio LR” under this assumption and parameter conditions; if LR”>1, then the proposition Hp” is supported; if LR”<1, then the proposition Hd” is supported.

[0106] (3) The non-contributor test procedure is to use the population frequency table to simulate the genotype at each locus in turn, combine the genotypes at different loci to generate 1000 simulated random individual typings, set the Hp hypothesis for each individual typing as the target individual in turn, repeat steps 3-4, and calculate the likelihood value LR under each set of mutually exclusive hypotheses.

[0107] If LR>0, or Log 10 When LR>1, the Hp hypothesis is supported, and the N groups of mixed maps are composed of simulated random individuals and K-1 unrelated individuals; LR<1 or Log10 When LR<0, the hypothesis Hd is supported, that is, the N groups of mixed graphs are composed of K unrelated individuals.

[0108] Step 7: Traverse each individual typing in the DNA database, and take each individual as the target individual, set mutually exclusive propositions Hp and Hd, where Hp proposition is N groups of mixed spectra consisting of the target individual and K-1 unrelated individuals; Hd proposition is N groups of mixed spectra consisting of K unrelated individuals. For each group of mutually exclusive propositions, obtain and Exists in g m,p But not in The allele Q in the dropout , exists in But not in g m,p The probability of occurrence of the allele in drop-in , which exists in Also exists in g m,p The probability of an allele in the dropout .

[0109] At this time, all loci are combined with N STR profile data to calculate the given assumptions under the premise of H (including Hd and Hp). The conditional probability of:

[0110]

[0111] Among them, p(a|g m,p ,p drop-in ,p dropout ) is a given genotype combination g m,p , p drop-in , p dropout The probability of allele a appearing under the premise of m,p ,p dropout ) is a given genotype combination g m,p , p dropout The probability of allele Q appearing under the condition of . m represents any autosomal locus m, and there are M loci in total.

[0112] Calculate the likelihood ratio LR"' of each individual in the DNA database as the target individual:

[0113]

[0114] Where: Given the assumption Hp, p drop-in , p dropout Spectra observed under the conditions The conditional probability of Given the assumption Hd, p drop-in , p dropout Spectra observed under the conditions The conditional probability of .

[0115] Sort the LR's of all individuals from large to small, and take the first ten as candidate individuals. Repeat steps 3-4 with the ten candidate individuals as target individuals. Calculate the corresponding individuals with LR>1 as the final individuals to be screened.

[0116] Example

[0117] To illustrate the beneficial effects of the method of the present invention, the method of the present invention is further used for processing as follows:

[0118] Step 1: Now simulate the preparation of three mixed DNAs from different sites, use the EX22 kit combined with polymerase chain reaction for DNA amplification, capillary electrophoresis technology for STR typing, and obtain the locus vector of the map Allele vector Allele peak height vector

[0119] Step 2: Fit the feature information in step 1. If the fitting result error reaches the set threshold, go to step 3, otherwise exit;

[0120] The three maps each have 21 autosomal loci, and the locus vectors of the three maps are obtained. Spectral peak height vector

[0121] Taking the four alleles at the first locus of the first map as an example, the allele peak height vector is The peak height and data sets were obtained by summing the allele peak height vectors of each map at 21 loci in turn, and three data sets were obtained. The data sets were fitted using normal distribution, lognormal distribution and gamma distribution respectively.

[0122] The sum of squares of the fitting errors of the three maps under the three distributions are [0.000001, 0.000002, 0.000002], [0.000006, 0.000007, 0.000007], [0.000005, 0.000005, 0.000005] respectively. The above values ​​are all less than 0.001, and the likelihood calculation and genotype splitting can be continued.

[0123] Step 3: Three STR profile data, according to the maximum likelihood estimation method, the number of contributors K is 2. Determine whether the STR profile of the target individual is obtained; if so, go to step 4; if not, set the proposition Hd, Hd is 3 mixed profiles composed of 2 unrelated individuals.

[0124] According to the assumptions, the corresponding parameters and maximum likelihood values ​​are calculated when the likelihood value is the largest under the given assumptions, and the genotype combinations are sorted; the DNA database is searched according to the sorting results;

[0125] Simulate the preparation of the target individual's DNA and obtain the target individual's DNA profile according to the method in step 1;

[0126] Step 4: Set mutually exclusive propositions Hp and Hd; Hp is a mixture of 3 graphs consisting of target individuals and 1 irrelevant individual; Hd is a mixture of 3 graphs consisting of 2 irrelevant individuals.

[0127] According to the likelihood function under the given assumptions Hp and Hd and the corresponding parameters, calculate the corresponding parameter value and the maximum likelihood value when the likelihood value is the largest under the given assumptions Hp and Hd, calculate the likelihood ratio LR under the parameter conditions, and sort the genotype combinations according to the LR value. If LR>1, then the proposition Hp is supported, and if LR<1, then the proposition Hd is supported;

[0128] Before calculating the likelihood ratio LR, we first obtain the genotype combination vector based on the N groups of map allele vectors, and calculate the probability of each genotype combination in the genotype combination vector under the given assumptions Hp and Hd, that is, the probability of the corresponding peak height.

[0129] Get allele vector for each map Taking the first locus of the first map as an example, L 1,1 The allele vector on is Get the alleles of each map at 21 loci in turn, remove duplicate values, and increase the Q of alleles that may be lost. Get the allele vector at each locus Repeat the selection twice The alleles in the genotype vector are generated. Each genotype of the genotype vector is repeated twice each time to generate a genotype combination vector Contains 4 alleles.

[0130] Calculate the appearance of g m,p The probability of the corresponding peak height is divided into the following cases:

[0131] (1) For g m,p Each allele a in can be defined as the allele a from contributor k at any autosomal locus m, with a peak height y m,a,k The probability f(y m,a,k ) is calculated as follows:

[0132] f(y m,a,k ) = gamma(y m,a,k |ω-2 ∑ k∈K mx k n a,k ,μω 2 )

[0133] Where μ is the expected peak height, ω is the coefficient of variation of the peak height, mx k is the proportion of the kth contributor, n a,k is the number of allele a of the kth contributor.

[0134] (2) When considering the occurrence of stutter, given g m,p , where each allele a will produce stutter, generating a stutter vector where g m,p,stutter,o is the oth stutter produced by the pth genotype combination at the mth locus, generated by the corresponding allele a, g m,p,stutter,o Peak height Probability The calculation method is as follows:

[0135]

[0136] Where δ is the stutter peak height ratio.

[0137] Peak height m,a,k The probability f(y m,a,k ) is calculated as follows:

[0138] f(y m,a,k ) = gamma(y m,a,k |(1-δ)ω -2 ∑ k∈K mx k n a,k ,μω 2 )

[0139] (3) Determine whether the peak height of the allele is linearly correlated with the length of the allele fragment. N,m,q The logarithm of the mean and a N,m,q The average fragment length was linearly fitted, and the slopes of the fitting results of the three spectra were -0.9978, -0.9969, and -0.9977, respectively. The absolute values ​​were all <0.9995, and the addition of the degradation parameter ε should be considered.

[0140] Considering the addition of degradation parameter ε, y m,a,k Probability f(y m,a,k ) is calculated as follows:

[0141]

[0142] The probability is calculated as follows:

[0143]

[0144] Where: S m,a is the fragment length of allele a at locus m, S min It is the allele length with the shortest fragment length in the population for all loci in the STR profile. g m,p,stutter,o The length of the fragment.

[0145] (4) When g m,p The genotype includes the missing allele Q. Then, at any autosomal locus m, the allele Q from contributor k has a peak height y m,Q,k The probability of ignoring degradation parameters is calculated as follows:

[0146]

[0147] f(y m,Q,k ) = gamma(y m,Q,k |ω -2 ∑ k∈K mx k n Q,k ,μω 2 )

[0148] The calculation method when considering the degradation parameters is as follows:

[0149]

[0150] Among them, AT is the analysis threshold of STR spectrum, f(y m,Q,k <AT) is y m,Q,k <Conditional probability when AT. S m,Q is the length of the fragment corresponding to the longest allele at locus m. Q,k is the number of the kth contributor allele Q.

[0151] (5) When Alleles in g m,p If the allele appears in the range of y, the allele is a drop-in. Then the peak height y of the drop-in on any autosomal locus m is m,drop-in The probability calculation method is:

[0152] f(y m,drop-in )=p c *f(drop-in|λ)*f drop-in

[0153]

[0154] Among them, p c is the probability of drop-in, λ is the parameter of exponential distribution, which is a constant. f(drop-in|λ) is the probability of drop-in alleles appearing under the given parameter λ. drop-in is the frequency of the drop-in allele.

[0155] Calculate the g under given assumptions Hp and Hd m,p The probability of a genotype combination. At any autosomal locus m, under a given hypothesis H, the observed genotype combination g m,p The probability p(g m,p |H) is the product of the genotype frequencies within the combination.

[0156] (6) The probability of a heterozygous genotype is the product of the frequencies of the two alleles multiplied by 2. The probability of a homozygous genotype is the square of the frequency of the allele. If there is a possible missing allele Q in the genotype, the frequency of the allele is 1-the sum of the frequencies of the alleles at the locus. When considering the existence of substructure in the corresponding population, it is necessary to use F st The value is corrected, represented by η; the national population STR frequency used in the present invention, F st It is 0.0108.

[0157] (7) Combine all loci and the three STR profile data. Give the likelihood function of LR under the assumptions and corresponding parameters. Obtain the corresponding parameter values ​​and the maximum likelihood value when the likelihood value under the conditions of Hp and Hd is the largest, and calculate LR.

[0158] First, under the assumptions of H and parameters, we observe The conditional probability

[0159] in, are three spectral vectors, H is the collective name of the assumptions Hp and Hd, is the parameter vector to be estimated, is the mixing ratio vector of the third spectrum, where is a vector representing the proportion of the DNA template amounts of the k contributors in the Nth map. is the expected peak height vector of the three spectra, is the peak height variation coefficient vector of the three spectra, is the degradation parameter vector of the three spectra, is the ratio vector of the stutter peak heights of the three spectra. p(g m,p |H,η) is the observed g under given H and η conditions. m,p The conditional probability of .

[0160] For the genotype combination g m,p and parameters p c , λ, in the nth STR profile, the peak height Y of the qth allele is observed at locus m. N,m,q The conditional probability of .

[0161] (8) The LR calculation method is as follows: (LR here is the calculation method under the condition of uncertain parameters)

[0162]

[0163] Perform maximum likelihood estimation on the unknown parameters under the assumptions Hp and Hd, and obtain the maximum likelihood value LR MLE (The LR used in the final judgment condition in step 4 of the present invention is the maximum likelihood value), and the calculation method is as follows:

[0164]

[0165] in, For the assumption that Hp and given Under these conditions, the parameter θ p The likelihood function of ; For the hypothesis θ d and given Under these conditions, the parameter θ d The likelihood function of .

[0166] The maximum likelihood value under the hypothesis proposition Hp is -2080.8, and the parameter search result under Hp is

[0167] Assume that the maximum likelihood value under proposition Hd is -2149.3; the result of searching for parameters under Hd is

[0168] LR is 5.45E+29, Log 10 The LR is 29.74.

[0169] (9) Optimization parameter search results max under fixed Hp and Hd θ , calculate g m,p The posterior probability p(g m,p |max θ ,H,y), and sort them according to the calculated probability values.

[0170]

[0171] Among them, p(g m,p |max θ ,H,y) is the observed genotype combination g under the premise of hypothesis H, maximum likelihood estimated parameter θ value and peak height y m,p The posterior probability of N,m,q |g m,p ,max θ ) is the genotype combination g m,p And the given optimal parameter search result max θ Under the condition of N,m,q The posterior probability of p(g m,p |H,η) is the observed genotype combination g under the assumption of proposition H and parameter η m,p The posterior probability of .

[0172] The genotype combination with the maximum posterior probability at each locus under Hp is shown in Table 1, and the genotype combination with the maximum posterior probability at each locus under Hd is shown in Table 2.

[0173] Table 1. Genotype combinations at different loci corresponding to the maximum posterior probability under Hp

[0174]

[0175] Table 2. Genotype combinations at different loci corresponding to the maximum posterior probability under Hd

[0176]

[0177] Step 5: LR>1, Log 10 When LR>0, the hypothesis Hp is supported, that is, the three mixed graphs are composed of the target individual and one irrelevant individual. Go to step 6. If the LR result is <1 at this time, the hypothesis Hd is supported, and go to step 7;

[0178] Step 6: Consider the father of the target individual as the actual contributor to the spectrum, that is, set mutually exclusive propositions Hp' and Hd', where the Hp' proposition is that 3 groups of mixed spectra are composed of the target individual and 1 unrelated individual. The Hd' proposition is that 3 groups of mixed spectra are composed of the father of the target individual and 1 unrelated individual. At this time, κ = [0,1,0], under the Hd' proposition and parameter θ, the probability of observing the STR spectrum is:

[0179]

[0180] At this time, Hp' and Hp assume that the propositions are completely consistent, and the parameter search results are consistent with those of Hp. The maximum likelihood value under Hd' is -2117.6. The parameter search results under Hd' are

[0181] LR' is 9.345E+15, Log 10 LR' is 15.97.

[0182] LR'>1,Log 10 LR'>0, which supports the hypothesis Hp', that is, it supports that the three groups of mixed graphs are composed of the target individual and one unrelated individual, rather than the father of the target individual and one unrelated individual.

[0183] To conduct a non-contributor test, use the population frequency table to simulate the genotype at each locus in turn, combine the genotypes at different loci to generate 1,000 simulated random individual typings, and conduct a non-contributor test. That is, set the Hp hypothesis for each simulated individual typing as the target individual in turn, and repeat steps 3-4. Calculate the likelihood value LR under each set of mutually exclusive hypotheses.

[0184] 1000 sets of simulation logs 10 The maximum value of (LR) is -22.4229, the median is -42.9418, the mean is -42.9870, and the 99% quantile is -35.0335. All of them are <0, which supports the hypothesis Hd. The three mixed maps are composed of simulated random individuals and 1 unrelated individual.

[0185] Step 7: The simulated DNA database includes 2000 unrelated individuals, the father and son of the target individual, 2 full siblings of the target individual, and the target individual himself. Set mutually exclusive propositions Hp and Hd, treat each individual in the database as a known individual under the Hp proposition, and calculate the likelihood ratio LR''. The Hp proposition is composed of 3 groups of mixed spectra consisting of known individuals and 1 unrelated individual. The Hd proposition is composed of 3 groups of mixed spectra consisting of 2 unrelated individuals.

[0186] For each set of mutually exclusive propositions, obtain and Exists in g m,p But not in The allele Q in the dropout , calculated to be 0.05, exists in But not in g m,p The probability of occurrence of the allele in drop-in , calculated to be 0.05, exists in Also exists in g m,p The probability of an allele in the dropout , calculated to be 0.05.

[0187] At this time, all loci are combined with N STR profile data to calculate the given assumptions under the premise of H (including Hd and Hp). The conditional probability of:

[0188]

[0189] Among them, p(a|g m,p ,p drop-in ,p dropout ) is a given genotype combination g m,p , p drop-in , p dropout The probability of allele a appearing under the premise of m,p ,p dropout ) is a given genotype combination g m,p , p dropout The probability of allele Q appearing under the condition of . m represents any autosomal locus m, and there are M loci in total.

[0190] Calculate the likelihood ratio LR"' of each individual in the DNA database as the target individual:

[0191]

[0192] Sort all the individuals from large to small, and take the first ten as candidate individuals. Repeat 3-4 with the ten candidate individuals as target individuals. The LR'' and LR of the first ten are shown in Table 3. The corresponding individuals with LR>1 are calculated as individuals 1 and 2, among which individual 1 is the target individual himself, and combined with the actual case to facilitate the next step of investigation.

[0193] Table 3. LR”' and LR calculated for the top ten individuals

[0194]

[0195] The method of the present invention calculates and evaluates LR through an automated and process-based strategy method, which can more comprehensively examine the entire DNA mixture analysis process and enhance the reliability of the results. It not only improves the efficiency and quality of DNA mixture analysis, but also provides more scientific and powerful support for subsequent legal decisions. In this process, the generalization ability of the model is improved, and it is more widely adapted to different types of DNA samples and complex and changeable experimental conditions, which is convenient for the majority of DNA analysis-related practitioners who lack a mathematical foundation, and helps to improve the application level of DNA analysis technology in the field of forensic science.

[0196] The present invention expands the applicable scenarios of the LR calculation model and meets the actual needs of joint analysis of multiple spectra in serial and parallel cases. This breakthrough greatly expands the application scope of the LR calculation model and can meet the urgent needs of the forensic evidence application field for mixed spectra analysis. The method of the present invention addresses the problem that mixed spectra are difficult to use directly for database retrieval, provides strong technical support for case detection, and provides a comprehensive analysis strategy for DNA evidence as court evidence.

Claims

1. A hybrid DNA evidence analysis method based on STR genotyping profiles, characterized in that: The following steps are involved: Step 1: Obtain STR profiles of N mixed DNA samples and obtain data feature information; Step 2: Fit the feature information in step 1. If the fitting result error reaches the set threshold, go to step 3, otherwise exit; Step 3: Determine the number of contributors K and determine whether the STR profile of the target individual is obtained; if so, proceed to step 4; if not, set the proposition Hd, where Hd is N mixed profiles composed of K unrelated individuals; according to the assumptions, calculate the corresponding parameters and maximum likelihood values ​​when the likelihood value is the largest under the given assumptions, and sort the genotype combinations; perform DNA database search based on the sorting results; Step 4: Set mutually exclusive propositions Hp and Hd; Hp is N mixed spectra consisting of target individuals and K-1 unrelated individuals; Hd is N mixed spectra consisting of K unrelated individuals; according to the likelihood function under the given assumptions Hp and Hd conditions and corresponding parameters, calculate the corresponding parameter value and maximum likelihood value when the likelihood value is maximum under the given assumptions Hp and Hd conditions, calculate the likelihood ratio LR under the parameter conditions, and sort the genotype combinations; Step 5: If LR>1, then the proposition Hp is supported and the process goes to step 6. If LR<1, then the proposition Hd is supported and the process goes to step 7. Step 6: Consider whether the contributor's relatives participate in the spectrum. If so, set mutually exclusive hypotheses, calculate the corresponding parameters and maximum likelihood values ​​when the likelihood value is the largest, calculate the likelihood ratios LR' and LR" under the parameter conditions, and perform a non-contributor test; otherwise, directly perform a non-contributor test; The mutually exclusive assumptions are as follows: Set the propositions Hp' and Hd', where Hp' is N mixed graphs consisting of the target individual and K-1 unrelated individuals, and Hd' is N groups of mixed graphs consisting of the relatives of the target individual and K-1 unrelated individuals; Under the assumption, find the parameters to perform maximum likelihood estimation, calculate the maximum likelihood value, and obtain the likelihood ratio LR' under the assumption and the parameter conditions; If LR'>1, then the proposition Hp' is supported, and if LR'<1, then the proposition Hd' is supported; Set propositions Hp” and Hd”, where Hp” is N sets of mixed graphs consisting of target individuals and their relatives and K-2 unrelated individuals; Hd” is N sets of mixed graphs consisting of random individuals and their relatives and K-2 unrelated individuals; Under this assumption, find parameters for maximum likelihood estimation, calculate the maximum likelihood value, and obtain the likelihood ratio LR” under this assumption and parameter conditions; if LR”>1, then the proposition Hp” is supported; if LR”<1, then the proposition Hd” is supported; The non-contributor testing process is as follows: According to the national population STR frequency table, simulate the genotype at each locus in turn, combine the genotypes at different loci to generate a simulated random individual typing data set, regard each individual as a target individual, repeat steps 3-4, calculate the corresponding parameter value and maximum likelihood value when the likelihood value is maximum under the given assumptions Hp and Hd, and calculate the likelihood ratio LR under the parameter conditions; Step 7: Traverse each individual typing in the DNA database, and take each individual as the target individual. Under the given assumptions Hp and Hd, calculate the likelihood ratio LR'' without considering the peak height, and sort the likelihood ratio LR'' from large to small; take the first d as candidate individuals, repeat steps 3-4, and the individuals with LR>1 are used as screening individuals; The calculation method of LR'' without considering the peak height is as follows: According to the allele vector at locus m and genotype combination vector Computation exists in But not in The probability p of the occurrence of allele Q in dropout ; Exists in But not in The probability of occurrence of an allele in drop-in ; exists in Also exists in The probability of an allele appearing in is 1-p dropout ; Combine the N STR profile data from all loci and calculate the conditional probability of observing the profile under a given hypothesis; Where: Given the assumption Hp, p drop-in , p dropout Spectra observed under the conditions The conditional probability of Given the assumption Hd, p drop-in , p dropout Spectra observed under the conditions The conditional probability of .

2. A hybrid DNA evidence analysis method based on STR genotyping profile according to claim 1, characterized in that: The data feature information in step 1 includes the locus vector, allele vector and peak height vector of the mixed DNA map.

3. A hybrid DNA evidence analysis method based on STR genotyping profile according to claim 1, characterized in that: In step 2, the data feature information is fitted using normal distribution, lognormal distribution and gamma distribution in turn.

4. A hybrid DNA evidence analysis method based on STR genotyping profile according to claim 1, characterized in that: Before calculating the likelihood ratio LR in step 4, the genotype combination vector is first obtained according to the N groups of spectrum allele vectors, and the probability of each genotype combination in the genotype combination vector appearing under the given assumptions Hp and Hd, that is, the probability of the corresponding peak height, is calculated.

5. A hybrid DNA evidence analysis method based on STR genotyping profile according to claim 1, characterized in that: The step 4 also includes: calculating the posterior probability of all genotype combinations at a certain locus according to the parameter value corresponding to the maximum likelihood value under the given assumptions Hp and Hd, and sorting them to obtain the desired sorting result.

Images