Systems and methods for intelligent genotyping by allele combination deconvolution

Abstract

Methods and systems are disclosed for improving computer efficiency by intelligently selecting a subset of possible short tandem repeat (STR) allele combinations for further deconvolution analysis. In one embodiment, at each locus, a processor calculates an adjusted evidence spectrum for a currently analyzed contribution rate scene of a plurality of contribution rate scenes. For a first or next unidentified contributor having a predetermined highest residual contribution rate in a currently analyzed contribution rate scenario for the plurality of contributors, the processor calculates a first range of expected peak heights using at least the predetermined highest residual contribution rate, the selected degradation value, and the peak height ratio distribution. Methods and systems for intelligently estimating the number of contributors of a biological sample are also disclosed.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Provisional Application Serial No. 63 / 294,342, filed December 28, 2021, and U.S. Provisional Application Serial No. 63 / 330,287, filed April 12, 2022. The entire contents of these applications are incorporated herein by reference to the extent permitted by applicable jurisdiction. Background Art

[0003] The present disclosure relates generally to genotyping, and more particularly to systems, devices, and methods for performing deconvolution analysis on biological samples potentially having multiple contributors.

[0004] Genotyping is the process of determining from a biological sample what combination of alleles (specific DNA sequence variations) an individual has at a specific genetic locus. Human identification, for example for forensic purposes, is often performed by analyzing the alleles at several known short tandem repeat (STR) loci. The STR region has a short DNA repeating sequence. The most common is a repeating sequence of four bases ("repeat units"). But some STR loci have repeat units of different lengths, for example in the range of 2 to 7 bases. Different STR alleles at a specific locus have different numbers of repeat units at that locus. STR loci have significant variability between individuals. Although two different individuals may have the same genotype (i.e., the same combination of two alleles) at a given STR locus, when considering several STR loci, the probability that two individuals have exactly the same genotype at each STR locus is very small, i.e., very close to zero. Therefore, analyzing the genotypes at several STR loci is a reliable way to identify individuals from a biological sample.

[0005] In forensic investigations, thirteen core STR loci are routinely used for DNA typing. The 13 core STR loci include, for example, TPOX, VWA, AMEL, TH01, FGA, D3S1358, etc. These STR loci are targeted with sequence-specific primers, and PCR is used to amplify fragments corresponding to these loci. The resulting DNA fragments are then separated and detected using electrophoresis. There are two common separation and detection methods, capillary electrophoresis (CE) and gel electrophoresis, of which CE is more common in recent years. Next-generation sequencing technology has also recently been used for DNA typing. Summary of the invention

[0006] In forensic investigations, biological samples often contain DNA from multiple contributors. The identities of some or all of these contributors may be unknown. Furthermore, the proportion of the sample that is attributed to each contributor is not necessarily known. Often, the exact number of contributors to a given biological sample is also unknown. The process of determining which contributors have what genotype in a multi-contributor sample is called deconvolution analysis. Before performing a deconvolution analysis, it is often necessary to make assumptions or estimates about the number of contributors to a sample.

[0007] Existing deconvolution techniques usually calculate and analyze theoretical spectra for each and every conceivable combination of alleles at a given locus. However, these techniques are time-consuming, and there is no need to study every conceivable scenario, and they are not efficient. Therefore, a more intelligent deconvolution analysis system and method are needed.

[0008] Embodiments of the present invention improve computer efficiency by intelligently selecting a subset of possible short tandem repeat (STR) allele combinations for further deconvolution analysis using the evidential profile obtained from the biological sample and the expected signal range of unidentified contributors.

[0009] In some embodiments of the invention, at each locus, for a currently analyzed contribution scenario among a plurality of contribution scenarios, the processor calculates an adjusted evidence spectrum by subtracting the calculated expected contributions of all known contributors (if any) from the evidence spectrum. For a first or next unidentified contributor having a predetermined highest remaining contribution in a currently analyzed contribution scenario for a plurality of contributors, the processor calculates a first range of expected peak heights using at least the predetermined highest remaining contribution, the selected degradation value, and the peak height ratio distribution.

[0010] In some embodiments, for all other remaining unidentified contributors (if any), the processor uses at least the predetermined contribution rate, selected degradation value and peak height ratio distribution in the contribution rate scenario of the current analysis of all other remaining unidentified contributors to calculate the expected peak height second range. The processor also uses the adjusted evidence spectrum and one or more of the first range and the second range to select one or more selected genotypes corresponding to the first or next unidentified contributor for further deconvolution analysis. The one or more selected genotypes include a genotype less than the total number of genotypes potentially associated with the current locus in the general population.

[0011] In some embodiments of the present disclosure, an improved method and system for estimating the number of contributors (NOC) of a biological sample is provided. In some embodiments, the improved NOC analysis technique for estimating NOC can enhance deconvolution analysis, making it easier to handle, faster and more efficient to perform deconvolution analysis. The prior art for estimating the number of contributors (NOC) of a biological sample uses a single model to provide a prediction or simply assumes all possible numbers of contributors. However, these techniques may not be accurate in all cases, and / or may be very time-consuming if all possible numbers of contributors are considered. For example, by considering all possible numbers of contributors to a sample, deconvolution analysis may need to calculate many possible scenarios for each possible number of contributors. Therefore, if all possible numbers of contributors are considered, the computational work of deconvolution analysis may be prohibitively large or impractical, and time-consuming. It is also unnecessary and inefficient to consider every conceivable scenario of the number of contributors. Therefore, a more intelligent NOC prediction system and method are needed.

[0012] Embodiments of the present invention improve computer efficiency by intelligently predicting the number of contributors to a biological sample using a combination of multiple models, including one or more machine learning-based models and / or one or more non-machine learning-based models. The combination of models can be pre-selected based on past analysis results, user preferences, the accuracy of each model, etc. A weight is assigned to each of the selected models so that a more accurate prediction of NOC can be achieved.

[0013] In some embodiments of the present invention, a method is provided for using an evidence spectrum obtained from a biological sample to predict the number of contributors associated with the biological sample to improve the computer efficiency of genotyping at least one contributor, the evidence spectrum including genetic signal data. The method includes using one or more computer processors to perform processing, the processing including: receiving an indication of a combination of selected models for estimating the number of contributors associated with the biological sample, and generating a predetermined number of calculated sample spectra based on the evidence spectrum. The method also includes: for each selected model in the combination of selected models, using the selected model and the calculated sample spectrum to estimate the probability of multiple contributor number possibilities, and using the estimated probabilities of multiple NOC possibilities obtained based on the combination of selected models to predict the number of contributors.

[0014] In some embodiments, a computer-implemented method is provided for predicting the number of contributors associated with a biological sample using an evidence spectrum obtained from the biological sample to improve computer efficiency for genotyping at least one contributor, the evidence spectrum comprising genetic signal data. The method includes: using one or more computer processors to perform processing, the processing including: receiving an expected maximum number of contributors, and generating a predetermined number of calculated sample spectra based on the evidence spectrum and the expected maximum number of contributors. The method also includes: for each possible number of contributors less than or equal to the expected maximum number of contributors, determining the number of peaks for each calculated sample spectrum with the corresponding possible number of contributors; obtaining an expected peak count distribution based on the number of peaks for each calculated sample spectrum with the corresponding possible number of contributors; determining the total number of peaks in the evidence spectrum; and estimating the probability of the multiple NOC possibilities based on the expected peak count distribution and the total number of peaks in the evidence spectrum.

[0015] These and other embodiments are described more fully below. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 An example of a capillary electrophoresis sequencing system according to an embodiment of the present disclosure is illustrated;

[0017] Figure 2A illustrates an example of a biological sample with multiple contributors and a corresponding evidence profile generated by analyzing the biological sample according to some embodiments of the present disclosure;

[0018] Figure 2B An example of a biological sample and a corresponding evidence spectrum generated by analyzing the biological sample according to an embodiment of the present disclosure is illustrated;

[0019] Figure 3A is a high-level flow chart illustrating a method of performing NOC prediction according to some embodiments of the present disclosure;

[0020] Figure 3B is a high-level flow chart illustrating a method of performing deconvolution analysis according to an embodiment of the present disclosure;

[0021] Figure 4 illustrates an example scenario for use in a forensic investigation according to an embodiment of the present disclosure;

[0022] Figure 5 is a flow chart illustrating a method of deconvolution analysis according to an embodiment of the present disclosure;

[0023] Figure 6 is a flow chart illustrating a method of calculating an expected first range of peak heights and an expected second range of peak heights for computerized display of an adjusted evidence spectrum according to an embodiment of the present disclosure;

[0024] Fig. 7A is a flow chart illustrating a method of determining a selected subset of allele combinations of unknown contributors for current analysis according to an embodiment of the present disclosure;

[0025] FIG. 7B to FIG. 7G Various scenarios for selecting allele combinations according to embodiments of the present disclosure are illustrated;

[0026] Figure 8 is a flow chart illustrating a method for calculating theoretical contribution according to an embodiment of the present disclosure;

[0027] Fig. 9 is a flow chart illustrating a method for calculating a theoretical spectrum of allele combinations of the contribution rate of the current analysis at a locus according to an embodiment of the present disclosure;

[0028] Fig.10 is a flow chart illustrating a method for calculating scores for bins in a theoretical spectrum according to one embodiment of the present disclosure;

[0029] Fig.11 A scenario of comparing a theoretical spectrum with an evidence spectrum according to an embodiment of the present disclosure is illustrated; and

[0030] Fig.12 illustrates an example of a user interface for displaying scores of a plurality of theoretical spectra according to one embodiment of the present disclosure;

[0031] Fig.13 illustrates an example user interface for receiving user input and displaying deconvolution analysis parameters and results according to one embodiment of the present disclosure;

[0032] Fig.14 is a flow chart illustrating a method of predicting the number of contributors of a sample according to some embodiments of the present disclosure;

[0033] Fig.15 is a flow chart illustrating a method of generating a predetermined number of calculated sample spectra according to some embodiments of the present disclosure;

[0034] Fig.16A is a flow chart illustrating a method of determining a characteristic of a peak height distribution of each calculated sample spectrum according to some embodiments of the present disclosure;

[0035] Fig. 16B is an example calculated sample spectrum generated by using the methods described in this disclosure;

[0036] Fig.17A illustrates a histogram of all loci for a calculated sample profile with one contributor according to some embodiments of the present disclosure;

[0037] Fig. 17B illustrates a histogram of all loci for a calculated sample profile with two contributors according to some embodiments of the present disclosure;

[0038] Fig.18 is a flow chart illustrating a method for determining one or more other features of each calculated sample spectrum according to some embodiments of the present disclosure;

[0039] Fig.19 illustrates expected peak count distributions according to some embodiments of the present disclosure;

[0040] FIG. 20A to FIG. 20B Illustrated is a computer interface for visualizing the effects on peak count distribution of different selections of the minimum relative fluorescence unit (RFU) used to count peaks according to some embodiments of the present disclosure.

[0041] FIG. 21A to FIG. 21B A computer interface for visualizing the relationship between the minimum peak height (minimum RFU) and the maximum allele count (MAC) from analyzed samples according to some embodiments of the present disclosure is illustrated.

[0042] Fig. 22 A block diagram of an example computing device is illustrated that may incorporate embodiments of the present disclosure.

[0043] Although embodiments of the present disclosure are described with reference to the above drawings, the drawings are intended to be illustrative and other embodiments are consistent with the spirit of the present disclosure and are within the scope of the present disclosure. DETAILED DESCRIPTION

[0044] Various embodiments will now be described more fully below with reference to the accompanying drawings, which form a part of the present invention and are used as illustrations to show specific examples of practical embodiments. However, this specification can be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein; on the contrary, these embodiments are provided to make this specification more thorough and complete and to fully convey the scope of the invention to those skilled in the art. In addition, this specification can be embodied as a method or an apparatus. Therefore, any of the various embodiments herein may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Therefore, the following description should not be considered to be limiting.

[0045] The number of repeats at a short tandem repeat (STR) locus is called an allele. For example, an STR locus called D7S820 found on chromosome 7 contains five to sixteen GATA repeats. Therefore, there are twelve possible different alleles for the D7S820 STR locus. For example, an individual with D7S820 alleles 10 and 15 will inherit a copy of D7S820 with 10 GATA repeats from one parent and a copy of D7S820 with 15 GATA repeats from the other parent. Because there are 12 different alleles for this STR locus, there are 78 different possible genotypes. A genotype corresponds to a pair of alleles. Specifically, there are 12 homozygotes where the same allele is received from each parent (e.g., (5,5), (6,6), (7,7), etc.). And there are 66 heterozygotes in which the genotypes of the two alleles received from the parents are different (e.g., (5,6), (7,10), (11,15), etc.).

[0046] In genotyping for forensic investigations, biological samples collected from crime scenes typically include contributions from multiple contributors (e.g., suspects, victims, persons discovering the crime scene, etc.) Many variables potentially explain a particular spectrum of evidence, including, for example, the number of contributors, the genotypes of those contributors, the level of sample degradation, the relative proportions (ratios) of contributions from each contributor, laboratory equipment variations, etc.

[0047] In conventional methods, a large amount of simulations are usually required for genotyping mixed samples by considering potential variables, using, for example, Markov chain Monte Carlo techniques to generate millions of random spectra for comparison. This process may take a very long time (e.g., several days, weeks or even months) to complete, and is therefore very inefficient. In addition, if the Monte Carlo simulation does not converge, the simulation will not provide any meaningful results. Repeated simulations may then be required, which results in further delays, increases energy consumption, and reduces analytical efficiency. This delay is significant and intolerable sometimes. The method described herein intelligently selects allele combinations for deconvolution analysis, and therefore improves timeliness and efficiency relative to the prior art.

[0048] The disclosed embodiments discussed herein intelligently select only a limited subset of allele combinations for deconvolution analysis. Relative to the methods of the prior art, this significantly reduces the quantity of allele combinations that need to be fully deconvolution analyzed. Therefore, results can be obtained faster without sacrificing accuracy. Therefore, the embodiments described herein improve the efficiency of computer-implemented genotyping.

[0049] Figure 1A system 100 is illustrated according to an example embodiment of the present disclosure. The system 100 includes a capillary electrophoresis ("CE") instrument 101, one or more computers 103, and a user device 107. The system 100 can be used to provide sequencing data or an evidential profile from a biological sample.

[0050] refer to Figure 1 , a CE instrument 101 in one embodiment includes a source buffer 118 containing a buffer and receiving a fluorescently labeled sample 120, a capillary 122, a destination buffer 126, a power source 128, and a controller 112. The sample 120 can be a biological sample to which a fluorescent marker is added for genotyping. Therefore, the sample 120 can include biological materials from one or more known or unknown contributors. The source buffer 118 is in fluid communication with the destination buffer 126 through the capillary 122. The power source 128 applies a voltage to the source buffer 118 and the destination buffer 126, generating a bias voltage through an anode 130 in the source buffer 118 and a cathode 132 in the destination buffer 126. The voltage applied by the power source 128 is configured by the controller 112 operated by the computing device 103. The fluorescently labeled sample 120 near the source buffer 118 is pulled through the capillary 122 by the voltage gradient, and when the optically labeled nucleotides of the DNA fragments in the sample reach the destination buffer 126 through the optical sensor 124, they are detected. Different sized DNA fragments within the fluorescently labeled sample 120 are pulled through the capillary at different times due to their size.

[0051] The optical sensor 124 detects the fluorescent labels on the nucleotides as image signals and transmits the image signals to the computing device 103. The computing device 103 aggregates the image signals into sample data and generates an electropherogram that can be shown on the display 108 of the user device 107. The electropherogram includes a DNA spectrum with peaks and their corresponding allele numbers. The electropherogram may include, for example, an evidence spectrum of a biological sample collected from a crime scene.

[0052] Instructions for implementing the relevant processing reside on the computing device 103 in the computer program product 104 stored in the storage device 105, and these instructions are executable by the processor 106. When the processor 106 is executing the instructions of the computer program product 104, the instructions or a portion thereof are typically loaded into the working memory 109, from which the processor 106 can easily access the instructions. In one embodiment, the computer program product 104 is stored in the storage device 105 or another non-transitory computer-readable medium (which may include distribution of the medium across different devices and different locations). In an alternative embodiment, the storage medium is transient.

[0053] In one embodiment, processor 106 includes multiple processors, which may include additional working memory (additional processors and memory are not separately illustrated), including a graphics processing unit (GPU), which includes at least thousands of arithmetic logic units that support large-scale parallel computing. Other embodiments include one or more special processing units, which include systolic arrays and / or other hardware arrangements that support efficient parallel processing. In some embodiments, such special hardware works with the CPU and / or GPU to perform the various processes described herein. In some embodiments, such special hardware includes application-specific integrated circuits, etc. (which may refer to a part of an application-specific integrated circuit), field programmable gate arrays, etc., and combinations thereof. However, in some embodiments, processors such as processor 106 can be implemented as one or more general-purpose processors (preferably with multiple cores) without having to depart from the essence and scope of the present disclosure.

[0054] The user device 107 includes a display 108 for displaying the results of the processing performed on the computer 103 .

[0055] Figure 2A The biological sample 210 and the method of using e.g. Figure 1 An example of a corresponding evidence spectrum 200 generated by analyzing a biological sample 210 by the CE instrument 101 shown. As shown in FIG2 , the biological sample 210 includes genetic material from multiple contributors such as contributor 202. The identities of some contributors may be known. But the identities of other contributors may be unknown. Therefore, the total number of contributors 202 may be uncertain. For example, although one of the contributors 202 may be determined to be a victim, the number of other possible contributors and their identities may be unknown. In addition, the biological sample 210 has a mixture of genetic material contributed by each contributor 202, and the proportion of their contributions may also be unknown. Therefore, before any analysis is performed to interpret the evidence spectrum, it is necessary to predict the number of contributors to the biological sample.

[0056] exist Figure 2A, an evidence profile 200 obtained from analyzing a biological sample 210 includes a graph having a y-axis representing signal intensity in relative fluorescence units ("RFU") and an x-axis representing the number of alleles at different loci. As described above, in forensic analysis, analyzing the alleles at several known short tandem repeat (STR) loci facilitates identification of contributors. The number of repeats at a short tandem repeat (STR) locus is called an allele. For example, an STR locus called D7S820 found on chromosome 7 contains five to sixteen GATA repeats. Therefore, excluding partial repeats in the form of microvariants, the D7S820 STR locus may have twelve different alleles. For example, an individual with D7S820 alleles 10 and 15 will inherit a copy of D7S820 with 10 GATA repeats from one parent and a copy of D7S820 with 15 GATA repeats from the other parent. Because there are 12 different alleles for this STR locus, there are 78 different possible genotypes. A genotype corresponds to a pair of alleles. Specifically, there are 12 homozygotes in which the same allele is received from each parent (e.g., (5,5), (6,6), (7,7), etc.), and there are 66 heterozygotes in which the genotypes of the two alleles received from the parents are different (e.g., (5,6), (7,10), (11,15), etc.).

[0057] Evidence spectrum 200 represents the signal intensity of the fluorescent label detected on the nucleotides as a sequence of peaks corresponding to different alleles at different loci (e.g., 7-11, 12-15). The signals corresponding to the fluorescently labeled nucleotides can be displayed in different colors, grayscale, or as different variations of black and white hatching representing various colors. Figure 2AAs shown, the evidence spectrum 200 includes, for example, peaks associated with several loci, such as locus 1 (e.g., D8S1179), locus 2 (e.g., D21S11), locus 3 (D7S820), etc. Each peak has its peak height. In CE, the peak height of an allele peak refers to the signal intensity of a signal (e.g., 500RFU) of a specific allele at a specific locus. For next generation sequencing (NGS) techniques for detecting alleles at STR loci and / or detecting SNPs (single nucleotide polymorphisms), the peak height refers to the number of reads at a specific locus at a specific genomic position of a specific allele or SNP. For NGS techniques for detecting micro-haplotypes (MH), the peak height refers to the number of reads covering the entire phased alleles of a specific locus or target (e.g., the number of reads constituting an allele genome construction and allele definition conversion tool or GACT). Microhaplotype loci are molecular markers of less than 300 nucleotides defined by two or more closely linked SNPs that are associated in multiple allele combinations. These markers facilitate massively parallel sequencing analysis using NGS technology.

[0058] As described above, the evidence spectrum 200 can be obtained by analyzing the biological sample 210 using, for example, the CE instrument 101. Because the biological sample 210 contains biological materials contributed by multiple contributors, the evidence spectrum 200 is a mixed spectrum, and a deconvolution analysis needs to be performed to assign DNA profiles (genotypes at each locus) to different contributors, which explain the peaks in the evidence spectrum 200 with sufficiently high certainty. Typically, in order to perform such a deconvolution analysis, the number of contributors needs to be predicted in advance.

[0059] It should be understood that Figure 2A The evidence spectrum 200 shown is a simplified spectrum for illustrative purposes. An actual evidence spectrum may include more or fewer loci, different loci, more or fewer peaks, different peaks, various shadow peaks, noise, etc. A shadow peak is a small peak that usually appears one repeat unit before or after the real peak. During PCR amplification, when copying the DNA strand, the polymerase may lose its position, such as sliding forward or backward by multiple base pairs, thereby causing a shadow peak.

[0060] Figure 2B An example of a biological sample 220 and a corresponding evidence spectrum 230 generated by analyzing the biological sample 220 is illustrated. Figure 2B As shown, biological sample 220 includes genetic material from multiple contributors (including contributors 221, 222, and 223). Figure 2BIn the example shown, contributor 221 is a contributor whose identity is known and whose DNA profile is available. Contributors 222 and 223 are contributors whose identities and DNA profiles are unknown. Biological sample 220 has a mixture of biological materials contributed by each of contributors 221, 222, and 223. In some cases, the exact number of contributors to a biological sample can be predetermined.

[0061] In some cases, the number of contributors is determined or estimated, but the ratio or proportion that each contributor contributes to the biological material of biological sample 220 may be unknown. Therefore, different scenarios of contribution ratios can be created and used in deconvolution analysis. Deconvolution analysis refers to genotyping analysis that deconvolutes a mixed DNA profile to explain the peaks in the DNA profile and assign a single profile to each individual contributor.

[0062] In some cases, the deconvolution analysis is a partial deconvolution analysis, which, for example, does not take into account contribution rates. By taking into account the contributor ratios (and optionally the sample degradation levels), a complete deconvolution analysis can be performed on the mixed DNA evidence spectrum. A complete deconvolution analysis can determine, for example, the actual or most likely contribution rate, the actual or most likely sample degradation rate, and the actual or most likely DNA spectrum of each contributor. The intelligent genotyping method for complete deconvolution analysis is described in more detail below.

[0063] exist Figure 2B , an evidence spectrum 230 obtained from a biological sample 220 includes a graph having a y-axis representing signal intensity in RFU and an x-axis representing the number of alleles. The evidence spectrum 230 represents the signal intensity of the fluorescent markers detected on the nucleotides as a sequence of peaks corresponding to different alleles (e.g., 7-11). The signals corresponding to the fluorescently labeled nucleotides can be displayed in different colors, grayscale, or as different variations of black and white hatching representing various colors.

[0064] As described above, the evidence spectrum 230 can be obtained by analyzing the biological sample 210 using, for example, the CE instrument 101. Because the biological sample 220 contains biological materials contributed by different contributors, the evidence spectrum 230 is a mixed spectrum, and a deconvolution analysis needs to be performed to assign DNA profiles (genotypes at each locus) to the different contributors, which explain the peaks in the evidence spectrum 230 with a sufficiently high certainty.

[0065] It should be understood that Figure 2B The evidence spectrum 230 shown is a simplified spectrum for illustrative purposes. The actual evidence spectrum may include more or fewer peaks, various shadow peaks, noise, etc. Shadow peaks are small peaks that appear immediately before or after the real peak. During PCT amplification, when replicating the DNA chain, the polymerase may lose its position, such as sliding forward or backward by multiple base pairs, thereby causing shadow peaks.

[0066] Figure 3A 3 is a flowchart 300 illustrating a method for predicting the number of contributors. In step 320 shown in FIG. 3, various inputs are provided for estimating the number of contributors using multiple selected models. Such inputs include evidence spectrum (or its data representation) 312, statistical model 314, parameter 316 and selected NOC model 318. Evidence spectrum 312 is a mixed DNA spectrum obtained from a biological sample with multiple contributors. An example of evidence spectrum 312 is evidence spectrum 220 as described above. Model 314 provides one or more distributions related to calculating shadow peaks, noise, shoulders, inter-locus balance (ILB), peak height imbalance or peak height ratio (PHR), population frequency and attenuation, etc. Model 314 is usually required to generate a calculated sample spectrum. For example, model 314 can include shadow model, PHR model, ILB model, noise model, population statistics and model builder / editor. Model builder / editor allows users to generate new models or modify / update existing models. These distributions or models are usually experimental, empirical and / or statistical. They can be independent of the biological sample or evidence spectrum being analyzed and can be applied to different samples or spectra. In some embodiments, some models 314 may be relevant or applicable to specific biological samples or evidence profiles, but not to other biological samples or evidence profiles. In some embodiments, a user may select which model to apply in a specific NOC analysis of an evidence profile and / or may customize certain models for better NOC analysis.

[0067] Parameters 316 may include any variables and / or user-controllable inputs used in the NOC analysis. For example, parameters 316 may include the expected maximum number of contributors, parameters associated with the test kit, weights assigned to each selected NOC model 318, and parameters associated with the test machine used to obtain the evidence spectrum, etc. Some of these parameters will be described in more detail below.

[0068] A combination of multiple selected NOC models 318 is used to perform NOC analysis. In some embodiments, a user interface listing available NOC models is provided to the user for selecting a NOC model. For example, the user interface may display a check box in front of each available NOC model. Therefore, the user may select a desired NOC model by clicking the corresponding check box. In other embodiments, a default group of multiple NOC models is provided. The user may simply use the default group of multiple NOC models for NOC analysis, or may customize the default group by removing certain models and / or adding other models. The user interface may also provide the ability to add a new model to the list of available models and / or remove an existing model from the list of available models. In some embodiments, a NOC model may be automatically selected without user intervention based on, for example, default settings, past experimental results, predefined strategies, etc.

[0069] In some embodiments, the list of available NOC models includes, for example, models based on artificial neural networks (ANNs), models based on decision trees, models based on random forest algorithms, and peak count distribution models. ANNs are computational networks or systems that can be trained to perform NOC predictions. As described in more detail below, a large number of computational sample spectra (e.g., simulated sample spectra similar to evidence spectra) and / or features of real biological sample spectra can be used to train ANNs. ANNs include a collection of artificial neurons. Artificial neurons receive signals and process them. Artificial neurons can transmit signals to neurons connected to them. The signal at the connection can be a real number, and the output of each neuron is calculated by a certain nonlinear function of the sum of its inputs. The connection between neurons is also referred to as an edge. Neurons and edges typically have associated weights that are adjusted as the network is trained. Weights can increase or decrease the strength of the signal at the edge. Neurons can have a threshold value so that a signal is sent only when the collective signal exceeds the threshold value. Typically, neurons are clustered into layers. Different layers can perform different conversions on their inputs. The signal in the ANN can travel from the first layer (input layer) to the last layer (output layer) through the intermediate layer (hidden layer). Features extracted from a large number of computational sample spectra can be used to train, test, and validate a particular ANN model. The features extracted from the evidence spectrum can then be provided to the trained ANN model, and NOC predictions are generated based on the probabilities of various possible NOCs. Some examples of ANNs include supervised neural networks (e.g., convolutional neural networks (CNNs), long short-term memory (LSTM) networks, etc.) and / or unsupervised neural networks (e.g., generative adversarial networks (GANs), etc.). It should be understood that the ANN can be any type of desired neural network with any number of desired layers or neurons.

[0070] The decision tree-based model uses a decision tree as a prediction model to predict the number of contributors. The decision tree-based model predicts the value of the target variable (e.g., NOC) based on several input variables. Similar to the ANN model, the training of the decision tree-based model can be completed by using a large number of calculated sample spectra (e.g., simulated sample spectra similar to evidence spectra) and / or features of real biological sample spectra. For example, decision trees can be used for classification. In the decision tree-based model for classification, all input variables (e.g., input features extracted from the calculated sample spectra) can have a limited discrete domain, and there is a single target variable (e.g., target feature) called classification. Each element of the classified domain is called a class. A decision tree or classification tree is a tree in which each internal (non-leaf) node is labeled with an input feature. The arc from the node labeled with the input feature is labeled with each possible value of the target feature, or the arc leads to a subordinate decision node on different input features. Each leaf of the tree is labeled with a class or a probability distribution on the class, indicating that the data set has been classified by the tree into a specific class or a specific probability distribution (if the decision tree is well constructed, it is skewed toward certain subsets of the class).

[0071] A decision tree can be constructed by splitting the source set constituting the root node of the tree into subsets constituting subsequent child nodes. The split is based on a set of splitting rules based on classification features. The process is repeated for each derived subset in a recursive manner known as recursive segmentation. When the subset at the node has all the same values ​​of the target variable, or when the split no longer adds value to the prediction, the recursion is completed. This process of top-down induction of a decision tree is an example of a greedy algorithm. A decision tree can be a classification tree, a regression tree, a classification and regression tree (CART), a boosting tree, a guided aggregation (or bagging) decision tree, a rotation forest tree, etc. as described above. In some embodiments, a large number of selected features of the sample spectrum are calculated to be used for training, testing and verifying a specific decision tree-based model. The features extracted from the evidence spectrum can then be provided to the trained decision tree-based model, and NOC predictions (e.g., the probability of various possible NOCs) are generated as target variables.

[0072] A random forest is a bootstrapped ensemble of decision trees. Random forests, or random decision forests, are ensemble learning methods for classification, regression, and other tasks that operate by constructing a large number of decision trees at training time. For classification tasks, the output of a random forest is the class selected by the majority of the trees. For regression tasks, the mean or average prediction of the individual trees is provided. Random decision forests can correct or improve the tendency of decision trees to overfit to their training set.

[0073] The ANN model, decision tree model, and random forest model described above are all examples of machine learning based models, where the models require training and / or training updates to provide accurate predictions of NOC. Figure 3AIn the embodiment of the present invention, the selected NOC model 318 may also include a non-machine learning based model. An example of a non-machine learning based model is a peak distribution model, which uses a large number of calculated sample spectra and / or real sample spectra to calculate the peak count distribution. The peak count distribution model is described in more detail below.

[0074] Still reference Figure 3A , in step 320, a NOC analysis is performed using a combination of selected NOC models based on the evidence spectrum 312, the model 314, and the parameters 316. In step 330, analysis results of different NOC models are generated and provided (e.g., in the form of probabilities of possible NOCs) for performing a NOC assessment. NOC assessments are often required because multiple NOC models can provide different NOC estimates. In some embodiments, NOC assessments can be performed using weights assigned to different NOC models. Assigned weights (e.g., in parameters 316) are provided for selected NOC models based on user input, default settings, past experimental results, predefined strategies, etc. The weights can be used to scale the analysis results of the selected NOC model to obtain scaled analysis results. The NOC assessment process 330 provides a NOC prediction based on the scaled analysis results associated with multiple selected NOC models. The predicted NOC can be visualized in step 340 (e.g., displayed to the user as a table, graph, chart, etc.). Below in Figures 14 to 19 The NOC analysis step 320 and the NOC evaluation step 330 are described in more detail in the context of FIG. In some embodiments, the predicted NOC is provided to a downstream process for performing, for example, a deconvolution analysis 350. The results 317 of the deconvolution analysis 350 can be provided to a process 360 for training a machine learning model and / or for simulation.

[0075] Figure 3B 3 is a flowchart 300 illustrating a method of performing a deconvolution analysis. Figure 3BIn the step 351 shown, various inputs for performing deconvolution analysis are provided. Such inputs include evidence spectrum (or its data representation) 312, one or more models 314, parameters 316, and scenarios 318 including a combination of hypotheses used in deconvolution analysis. Evidence spectrum 312 is a mixed DNA spectrum corresponding to the spectrum of several contributors. Evidence spectrum 312 can be, for example, the above-mentioned evidence spectrum 230. One or more models 314 provide one or more distributions related to calculating shadow peaks, noise, shoulders, inter-locus balance (ILB), peak height imbalance or peak height ratio (PHR), population frequency and attenuation, etc. For example, model 314 can include shadow model, PHR model, ILB model, noise model, population statistics and model maker. Model maker allows users to generate new models or modify / update existing models. These distributions or models are usually experimental, empirical or statistical. They can be independent of the biological sample or evidence spectrum being analyzed, and can be applied to different samples or spectra. In some embodiments, users can choose which model to apply in a specific deconvolution analysis and / or can customize certain models for better deconvolution analysis.

[0076] Parameters 316 may include any variable or user-controllable input used in the deconvolution analysis. For example, parameters 316 may include the number of contributors, contribution rate combinations, sample degradation rates, loci in the analysis, etc. Some of these parameters will be described in more detail below.

[0077] Figure 4 Scenario 318 is illustrated in more detail. Scenario 318 refers to a combination of hypotheses used in deconvolution analysis. For example, Figure 4 As shown, in a forensic investigation, a scenario may include a prosecution hypothesis 1 410 and a defense hypothesis 1 420. For example, the prosecution hypothesis 1 410 may indicate that there are three contributors, including a suspect, a victim, and an unidentified contributor. The prosecution hypothesis may also indicate the contribution rate of each of the three contributors. The defense hypothesis 1 420 may indicate that there are three contributors, including a victim and two unidentified contributors. The defense hypothesis may also indicate the contribution rate of each of the three contributors. In some cases, the scenario 318 may include a complex scenario. For example, Figure 4 As shown, scenario 318 may include multiple prosecution hypotheses 1-n and multiple defense hypotheses 1-n.

[0078] In some embodiments, the hypothesis can also have clues about the known or assumed true of the unidentified contributors. For example, for the unidentified contributor #3 in the prosecution hypothesis 1 410, the clue can be that this person has allele 13 at locus SE33. Therefore, it is necessary to add this allele to the allele combination of the selected genotype used in the deconvolution analysis. Such clues can help to limit the number of possible allele combinations or genotypes corresponding to the unidentified contributors, and therefore further reduce simulation iterations and time consumption.

[0079] like Figure 3B As shown, scenario 318 can be verified in step 351. In view of other evidence or evidence, the scenario proposed by the prosecution and / or the defense may be valid or invalid. And therefore, the scenario may need to be verified before performing the deconvolution analysis. The scenario needs to be consistent and logical. The verification of the scenario checks the potential conflicts of the input data. For example, during the verification process, scenario 318 can be checked against parameters 316. If the number of contributors included in parameters 316 is only two, but scenario 318 has four known contributors, there is a conflict between the input data. When there is a conflict, the verification process can mark the conflict and / or provide conflict information to the user. The verification process can also be re-executed after the conflict is resolved.

[0080] After validating scenario 318, one or more of evidence spectrum 312, model 314, parameters 316, and scheme 318 are used to compute deconvolution in step 352. Deconvolution analysis generates solutions 355, and at least some of the deconvolution solutions may be provided to a user and visually displayed in step 354. Fig.12 Such visualization using a heat map is illustrated.The deconvolution analysis performed in step 352 and the deconvolution 355 and visualization steps 354 are described in more detail below.

[0081] Figure 5 is a flow chart illustrating an example deconvolution method 500. The method 500 is a computer-implemented method and may be used to implement Figure 3B Step 352 (calculating deconvolution) of method 350 shown. In some embodiments, method 500 begins with step 501, in which, for a given number of contributors, contribution rate scenarios to be considered are determined. As described above, hypotheses of the number of contributors and their contribution rates (e.g., by the prosecution and the defense) are provided as scenarios to be considered. Therefore, contribution rate scenarios can be determined from the provided data. In other cases, contribution rate scenarios can be calculated as described below.

[0082] In some embodiments, contribution rate scenarios may or may not be easily determined and provided. For example, if there are multiple unidentified contributors and there is a high degree of uncertainty in the contribution amount from one or more unidentified contributors, it may be desirable to consider additional and / or different contribution rate scenarios in the deconvolution analysis. In some embodiments, contribution rate scenarios are determined based on a user-configurable ratio resolution step. For example, if there are two contributors and if the ratio resolution step is configured to 10%, possible contribution rate scenarios include 10%-90%, 20%-80%, 30%-70%, 40%-60%, and 50%-50%. For another example, if there are three contributors and if the ratio resolution step is set to 10%, possible contribution rate scenarios include (for simplicity, "1" is used to represent "10%, "2" is used to represent 20%, etc.): 1-1-8, 1-2-7, 1-3-6, 1-4-5, 2-2-6, 2-3-5, 2-4-4, and 3-3-4. Similarly, if there are four contributors and if the ratio resolution step is set to 10%, possible contribution rate scenarios include 1-1-1-7, 1-1-2-6, 1-1-3-5, 1-1-4-4, 1-2-2-5, 1-2-3-4, 2-2-2-4, and 2-2-3-3. And if there are five contributors and if the ratio resolution step is set to 10%, possible contribution rate scenarios include 1-1-1-1-6, 1-1-1-2-5, 1-1-1-3-4, 1-1-2-2-4, 1-1-2-3-3, 1-2-2-2-3, and 2-2-2-2-2. It should be understood that the above contribution rate scenarios are only examples, and other scenarios may be possible (for example, by changing the ratio resolution step to 5% or 15%).

[0083] Step 501 of method 500 also determines a sample degradation rate or value. The sample degradation rate or value is used in the deconvolution analysis as a factor to explain the fact that the collected biological sample may be degraded. Sample degradation can be caused by, for example, exposing the sample to sunlight, storing the sample for a long time, sample evaporation, sample drying, etc. In the deconvolution analysis, the sample degradation rate or value is used to scale the expected peak height or calculate the peak height. In some embodiments, the sample degradation rate or value is configurable and optionally used.

[0084] refer to Figure 5, after determining the contribution rate scenario and the sample degradation rate, deconvolution analysis is performed for each locus in the multiple loci considered (e.g., the 20 loci mentioned above, or the loci selected by the user). In step 502, the adjusted evidence spectrum is calculated by subtracting the calculated expected contribution of all known contributors (if any) from the evidence spectrum. For a known contributor, his or her DNA spectrum can be determined by, for example, analyzing a biological sample from the known contributor. Then the peak height in the DNA spectrum of the known contributor is scaled according to the contribution rate scenario currently analyzed (e.g., scaled according to the percentage of the known contributor relative to other contributors). The scaled DNA spectrum of the known contributor can be used as a calculated spectrum of the known contributor, which represents the expected contribution from the known contributor. In some embodiments, shadow peaks are added to the calculated spectrum of the known contributor. As described above, shadow peaks can be calculated based on a shadow model constructed empirically or statistically. This process of calculating the spectrum of the known contributor can be repeated multiple times to calculate the spectrum of all known contributors, which represents the expected contribution of all known contributors.

[0085] The calculated spectrum of the known contributor is then subtracted from the evidence spectrum to obtain an adjusted evidence spectrum. Thus, the adjusted evidence spectrum includes only the spectrum of the unidentified contributor. The adjusted evidence spectrum is then used for further deconvolution analysis. In some embodiments, the full peak height of the calculated spectrum of the known contributor is subtracted from the evidence spectrum. In some embodiments, due to peak height imbalance (PHI), the calculated spectrum of the known contributor is subtracted from the evidence spectrum using the lower limit of the range of expected peak heights, rather than the full peak height. That is, the lower limit of the range of expected peak heights in the calculated spectrum is subtracted from the evidence spectrum.

[0086] Peak height imbalance is usually calculated in heterozygous loci (containing two alleles from different parents). When the height difference between the two peaks of the two alleles in the heterozygous loci is greater than a certain percentage (e.g., 30%), peak height imbalance may occur. Typically, samples from a single person should contain peaks with approximately the same peak height. Therefore, if there is a peak height imbalance, it can indicate that the sample is a mixture. However, peak height imbalance can also be caused by other factors, such as the variability of amplification, pipetting, and electric injection. When subtracting the calculated spectrum of known contributors, if the lower limit of the range of the expected peak height in the calculated spectrum is subtracted from the evidence spectrum, the remaining peaks in the evidence spectrum can have a higher peak height than if the full peak height in the calculated spectrum is subtracted. Therefore, in some embodiments, further deconvolution analysis can explain this situation by scaling the upper limit of the expected peak height range higher using a scaling factor. For example, as described in more detail below, peak height imbalance is used to determine the expected range of peak heights for selecting a subset of allele combinations for further analysis. The upper limit of the expected range of peak heights for this selection can be adjusted by subtracting the calculated spectrum from the evidence spectrum (eg, subtracting the full peak height, or subtracting the lower limit of the expected range of peak heights).

[0087] refer to Figure 5 , for the current contribution rate scenario, step 504 of method 500 selects the highest remaining unidentified contributor corresponding to the adjusted evidence spectrum. For example, there may be a total of three contributors, including two unidentified contributors and one known contributor. The contribution rates of the three contributors may be 10%-30%-60%, where 10% is the percentage of the contribution of the known contributor. In this contribution rate scenario, the unidentified contributor with a 60% contribution is first selected for analysis, and the unidentified contributor with a 30% contribution will be selected for analysis in the next iteration.

[0088] Once the highest remaining unidentified contributor is selected, a first range of expected peak heights for that unidentified contributor and a second range of expected peak heights for all other remaining unidentified contributors are calculated.

[0089] Figure 6 A method 600 is illustrated for calculating a first range 615 and a second range 625. The expected peak height first range represents a range of expected peak heights accounting for the peak height imbalance of the highest remaining unidentified contributor. The expected peak height second range represents a range of expected peak heights for all other remaining unidentified contributors combined, accounting for the peak height imbalance.

[0090] Method 600 may be used to implement Figure 5 Steps 506 and 507 of method 500 in step 602 of method 600 calculate the sum of the peak heights in the adjusted evidence spectrum. Figure 6 As an example, in the adjusted evidence spectrum 620, for alleles 7, 8, 9, 10, and 11, there are a total of 5 peaks, whose peak heights are about 50, 200, 800, 80, and 400, respectively. Therefore, the sum of these peaks is about 1530.

[0091] Step 604 calculates the expected allele peak height that will be generated from the contributor of the current analysis with one of the alleles (i.e., the first or highest remaining unidentified contributor in the current iteration). The calculation is based on the sum of the peak heights (from the adjusted evidence spectrum), the contribution rate scenario of the current analysis, and the selected degradation value. For example, in the current analysis scenario assuming two unidentified contributors and 80%-20% contribution rate, the sum of the expected peak heights of the main contributors is approximately 0.8*1530=1224. The sum of the expected peak heights is for two alleles, because the DNA spectrum of the person has two alleles, each allele from each of the parents of the person. Therefore, for an allele of the main contributor, the expected peak height is about 1224*0.5=612. In some embodiments, the degradation value is used to scale down the contribution rate. For example, a degradation function and parameters can be used to scale down 0.8 contributions to 0.6 or some other value lower than 0.8.

[0092] In a similar manner, step 614 calculates the expected allele peak height that will be produced by all other unidentified contributors having one of the alleles. In the above example, only one other unidentified contributor has a 20% contribution. For a contributor with two identical alleles, the expected height will be 0.2*1530=306. For one allele of this minor contributor, the expected peak height is about 306*0.5=153. If there are two such "minor" unidentified contributors, the result is the same, with each contributor having a 10% contribution. Similarly, taking into account sample degradation, the contribution factor 0.2 in this example will be scaled to a value lower than 0.2.

[0093] Steps 606 and 616 adjust the first expected peak height and the second expected peak height calculated above by removing the estimated shadow peak. A shadow peak is a small peak that appears immediately before or after the true peak. During PCR amplification, when replicating the DNA chain, the polymerase may lose its position, such as sliding forward or backward by multiple base pairs. This may result in shadow peaks. Therefore, in steps 606 and 616, the expected peak heights calculated above are adjusted to account for the shadow peak heights. Continuing with the above example, using the shadow model, the expected peak heights can be adjusted, for example, from about 612 to about 550 for major contributors and from about 153 to about 137 for minor contributors.

[0094] Steps 608 and 618 use the first expected peak height and the second expected peak height and the peak height imbalance (PHI) factor to calculate an expected first range of peak heights 615 (for the unidentified contributors currently being analyzed, sometimes also referred to herein as “primary” unidentified contributors) and an expected second range of peak heights 625 (for all other unidentified contributors combined, sometimes referred to herein as “secondary” contributors).

[0095] As described above, the peak height imbalance ratio distribution can be obtained from an empirical model or a statistical model. Ideally, the peaks of the same contributors will have approximately the same peak height. However, due to random effects, the peaks are not exactly the same height. Random effects are effects that occur by chance due to various factors and variables associated with analyzing biological samples to obtain evidence spectra. Peak height imbalance (PHI) or changes due to random effects can usually be modeled using a gamma distribution. The gamma distribution can be defined by its shape, proportion, and threshold parameters. In some embodiments, the threshold parameter of the gamma distribution for representing peak height imbalance or change is set to about 0.4-0.6.

[0096] exist Figure 6In the example shown, a peak height imbalance value of 0.6 is selected using a gamma distribution and used to calculate the first range 615 and the second range 625. Using a PHI value of about 0.6 and the adjusted expected peak heights of the major contributors, the change in the adjusted expected peak heights of the major contributors can be calculated to be about 0.6*550=330. Therefore, the expected peak height first range 615 can be calculated to be between about 385 (i.e., 550-330 / 2) and about 715 (i.e., 550+330 / 2).

[0097] Using the same PHI value of about 0.6 and the adjusted expected peak heights for all minor unidentified contributors (in this case, one minor contributor), the change in the adjusted expected peak height for the minor contributor can be calculated to be about 0.6*274≈164. Thus, the second range of expected peak heights 625 can be calculated to be between about 192 (i.e., 274-164 / 2) and about 356 (i.e., 274+164 / 2).

[0098] In the above example, two unidentified contributors (i.e., a major contributor with 80% contribution and a minor contributor with 20% contribution) are used for illustration purposes. It should be understood that more unidentified contributors are possible and the calculation of the first range and the second range can be performed similarly. For example, if there are 3, 4, or 5 unidentified contributors in total, the first range can still be calculated in a manner similar to that described above using the contribution rate of the unidentified contributor currently analyzed (e.g., the first or highest remaining unidentified contributor). The second range can be calculated in a manner similar to that described above using the contribution rates of all other remaining unidentified contributors rather than the contribution rate of only one minor contributor as in the above example.

[0099] exist Figure 6 In the above example shown, the first range 615 is higher than the second range 625, and the two ranges do not overlap. However, depending on various scenarios such as total peak height, contribution rate, peak height imbalance distribution, degradation rate, etc., the first range and the second range may have other relationships. For example, the first range and the second range may be farther apart or may overlap each other. In other examples, the second range may be higher than the first range.

[0100] exist Figure 5 The first range and the second range are used in step 508 of the method 500 to select which genotypes correspond or may correspond to the unidentified contributors (e.g., the first or highest remaining unidentified contributors) of the current analysis. Importantly, the selected genotypes include fewer genotypes than the total number of genotypes potentially associated with the current locus in the general population.

[0101] Fig. 7A Illustrate the implementation Figure 5Method 700 of step 508 of method 500. FIG. 7B to FIG. 7G Different examples of adjusted evidence spectra to which method 700 may be applied are shown.

[0102] refer to Fig. 7A , step 702 calculates the threshold of the second range. The threshold of the second range is the value of the upper limit of the second range. For example, if the upper limit of the second range 765A is about 360, Figure 7B As shown, the threshold of the second range 765A may therefore be configured or selected to be approximately 360. Step 704 determines whether any peak in the adjusted evidence spectrum has a peak height that is at least twice the threshold of the second range.

[0103] Figure 7B This scenario is illustrated in Figure 7B In the adjusted evidence spectrum 760A shown, allele 9 has a peak with a peak height of about 800, which is greater than twice the threshold 360 of the second range 765A. Figure 7B In the example shown in FIG. 7 , step 706 selects allele 9 twice as a pair (9 9), which is the only combination selected for further analysis. In this example, the peak height of allele 9 is so high that the peak can only be explained by the contributor of the current analysis being homozygous for the allele. Therefore, it is necessary to select two alleles (for example, 9 and 9 in this case). In this case, two alleles of the unidentified contributor of the current analysis have been found. Therefore, the count of the required allele increases by 2.

[0104] Step 722 checks whether the count of the desired allele is less than 2. In the above case, two alleles have been found, so the count is 2. The answer determined in step 722 is therefore "no". Therefore, method 700 proceeds to step 732, in which the only allele combinations are calculated using the same allele twice. Thereafter, the process moves on to further analysis for that allele combination (e.g., Figure 5 509 in step 509).

[0105] In some embodiments, using the adjusted evidence spectrum and one or more of the first range and the second range, determine whether the allele in the adjusted evidence spectrum has a peak that is at least above the threshold of the second range but less than twice the threshold. If so, the allele is selected as the desired allele of one of the two alleles of the selected genotype. For example, referring to Fig. 7A and Figure 7CIf the determination in step 704 is "no" (i.e., there is no peak with a height at least twice the threshold of the second range), the method 700 proceeds to step 708. Step 708 determines whether there are any adjusted evidence spectrum peaks above the threshold of the second range. If so, step 710 selects the allele as the desired allele once.

[0106] Figure 7C illustrates one such scenario. Figure 7C In , it is assumed that the threshold of the second range 765B is about 360 (ie, the upper limit of the second range 765B). Figure 7C In the adjusted evidence spectrum 760B shown, the two peaks of alleles 8 and 9 each have a peak height higher than the threshold of the second range 765B (e.g., higher than the upper limit of the second range 765B). Therefore, the two peaks are each selected once. In other words, the contributor of the current analysis is determined to be heterozygous, having both alleles 8 and alleles 9.

[0107] In this scenario, for each selection of two alleles, the count of the desired allele is increased by "1". Therefore, the total count is increased by 2. Therefore, this allele pair is the only combination selected for further analysis. This is because the two peaks can only be interpreted as being heterozygous for the allele of this unidentified contributor. Therefore, two alleles need to be selected (for example, 8 and 9 in this case). In this case, both alleles of the unidentified contributor currently analyzed have been found. Fig. 7A Method 700 proceeds to step 722 to check whether the count of the desired allele is less than 2. In this case, the answer determined in step 722 is also "no". Therefore, method 700 proceeds to step 732, in which alleles 8 and 9 are used to calculate the only allele combination. Thereafter, the process proceeds to further analysis for this allele combination (e.g., Figure 5 509 in step 509).

[0108] Fig.7D Another scenario is illustrated in which only one peak, but not both peaks, is determined to be above the threshold of the second range 765C. Fig.7D , again assuming that the threshold of the second range 765C is about 360 (i.e., the upper limit of the second range 765C). Fig.7DIn the adjusted evidence spectrum 760C shown, only the peak of allele 8 has a peak height above the threshold of the second range 765C. The peak of allele 9 is below the threshold of the second range 765C. Therefore, the peak of allele 8 is selected once. The peak of allele 9 is not selected. In this case, allele 8 can form a heterozygous allele pair with one of the other possible alleles. Therefore, the count of the desired allele is increased by 1. Allele 8 is the only allele that needs to be selected. Therefore, only one allele is found for the unidentified contributors currently analyzed.

[0109] Return to reference Fig. 7A , step 722 of method 700 checks whether the count of the desired allele is less than 2. In the above scenario where only one allele is found, the answer determined in step 722 is "yes". Method 700 therefore proceeds to step 712 and other subsequent steps to select other possible alleles, as described in detail below. After selecting one or more possible alleles as described below, step 734 of method 700 uses the desired allele (e.g., allele 8) and all other possible alleles to calculate the allele combination. Thereafter, the process proceeds to further analysis for this allele combination (e.g., Figure 5 509 in step 509).

[0110] As mentioned above, step 704 of method 700 determines whether the allele in the evidence spectrum adjusted has a peak sufficiently high (e.g., at least twice) than the threshold value of the second scope, so that the unidentified contributor analyzed should be considered as being homozygous for the allele. Step 708 determines whether any two alleles in the evidence spectrum adjusted have corresponding peaks higher than the threshold value of the second scope (but not twice). If any one of these two determinations is "yes", a pair of alleles of the unidentified contributor of the current analysis only includes the required allele. Then select these two required alleles, and no further selection is needed. The process can continue to perform further analysis, and then proceed to the next iteration of analyzing the next unidentified contributor.

[0111] refer to Fig. 7A , if the count of the desired allele is less than 2, the method 700 proceeds to steps 712 and 714 to determine whether the allele in the adjusted evidence spectrum has a peak that is at least twice the threshold of the first range. If so, the allele is selected twice as a possible allele for each selected genotype in one or more selected genotypes. For example, if the determination in step 708 is "no" or the determination in step 722 is "yes", the method 700 proceeds to step 712 to calculate the threshold of the first range. The threshold of the first range is the value of the lower limit of the first range. For example, if the lower limit of the first range 763D is about 400, as shown in FIG. Fig. 7EAs shown, the threshold of the first range 763D can therefore be configured or calculated to be about 400. Using the threshold of the first range, step 714 of method 700 determines whether there are any adjusted evidence spectrum peaks with a peak height at least twice the threshold of the first range. If so, step 716 selects the allele twice as two possible alleles.

[0112] Fig. 7E This example illustrates such a situation. In the adjusted evidence spectrum 760D, if Fig. 7E As shown, the peak of allele 8 has a peak height of about 850, which is more than twice the lower limit of the first range 763D (e.g., about 400). Therefore, the peak of allele 8 is selected twice as a possible allele because when it is selected twice, the allele can form a possible homozygous allele pair (e.g., allele pair (8 8)). In this case, two possible alleles of the unidentified contributors of the current analysis have been found (e.g., allele 8 8). The allele can also form one or more possible allele pair combinations with other possible alleles (e.g., combinations (8,9), (8,11), etc.).

[0113] refer to Fig. 7A , after selecting allele twice as possible allele in step 716, method 700 proceeds to step 724 to check whether there is no required allele. If the answer in step 724 is "no" (for example, currently finding a required allele), then method 700 proceeds to step 734 to use a required allele and all possible alleles to calculate all allele combinations. But if required allele is not found (that is, the answer in step 724 is "yes"), then method 700 proceeds to step 736 to calculate allele combinations using only possible alleles. In step 738, each allele combination in these allele combinations is added to selected subset for further deconvolution analysis. Then repeat the above process many times, until there is no remaining unidentified contributor.

[0114] refer to Fig. 7A If there is no peak in the adjusted evidence spectrum that is higher than twice the first range (i.e., the answer to step 714 is "no"), the method 700 proceeds to step 718 to determine whether the allele in the adjusted evidence spectrum has a peak higher than the threshold of the first range. If so, step 720 selects the allele as at least one of the two possible alleles for each selected genotype.

[0115] Figure 7F This scenario is illustrated in Figure 7FIn the adjusted evidence spectrum 760E shown, it is assumed that the threshold of the first range 763E is its lower limit of about 400. The peak of allele 11 has a peak height that is higher than the threshold but lower than twice the threshold. The peak of allele 9 will be higher than twice the threshold and should have been selected in the previous step (e.g., step 714). Therefore, the peak of allele 11 is selected once as a possible allele of the unidentified contributor of the current analysis, because the allele 11 can form a heterozygous allele pair with another allele. In this case, a possible allele (e.g., allele 11) of the unidentified contributor of the current analysis has been found.

[0116] After step 720, method 700 proceeds to step 724 to check whether there is not required allele.If the answer in step 724 is " no " (that is, there is required allele), then method 700 proceeds to step 734 to use a required allele and all possible alleles to calculate all allele combinations.If there is not required allele (that is, the answer in step 724 is " yes "), then method 700 proceeds to step 736 to use only possible alleles to calculate allele combinations.In step 738, each allele combination in these allele combinations (for example, there is required allele or only has the allele combination of possible allele) is added to selected subset for further deconvolution analysis.In other words, for each allele combination in these allele combinations, deconvolution analysis proceeds to next iteration to analyze the next unidentified contributor.In step 718, if the answer is " no ", then mean that there is not peak to be selected. The method 700 proceeds to the next iteration (eg, goes to step 511 to see if there are any more unidentified contributors to analyze).

[0117] As described above, step 714 determines whether the allele in the evidence spectrum being adjusted has a peak that is at least twice the threshold value of the first scope. Step 718 determines whether the allele in the evidence spectrum being adjusted has a peak that is higher than the threshold value of the first scope but is not twice the threshold value. If any one of the two determinations is "yes", then at least one allele is selected for the unidentified contributor of the current analysis, because it is a possible allele for explaining the peak in the spectrum of the current unidentified contributor. Using the selected possible allele and required allele (if any), method 700 calculates the allele combination in step 734 or step 736. For each allele combination in the allele combination, the analysis of the next unidentified contributor is carried out. For example, if allele 11 is a required allele and alleles 5, 6, 7 are possible alleles (but not required alleles), then allele 11 is selected for the unidentified contributor of the current analysis. For each of the possible alleles 5, 6, and 7, step 734 calculates three possible allele combinations for the unidentified contributor currently being analyzed, namely (11, 5), (11 6), and (11 7). For each of these allele combinations, the analysis of the next unidentified contributor is performed. The above process is then repeated multiple times until there are no remaining unidentified contributors.

[0118] Figure 7G Another example in which only possible alleles are found is illustrated. For example, in Figure 7G In the evidence spectrum 760F of the adjustment shown, the peaks of alleles 7, 8 and 11 have a peak height higher than the threshold value of the first range 763F (e.g., the lower limit of the first range 763F) but not twice the threshold value of the first range 763. Therefore, these peaks of alleles 7, 8 and 11 are selected as alleles to form possible allele combinations of unidentified contributors of the current analysis, because these alleles can form heterozygous allele pairs (e.g., (7 8), (7 11) and (8 11)) with each other. In this case, only possible alleles (i.e., alleles 7, 8 and 11) of the unidentified contributors of the current analysis are found, but the required alleles are not found. Because there is no required allele, the answer determined in step 724 (i.e., whether there is no required allele) is "yes". Therefore, method 700 proceeds to step 736 to calculate all allele combinations using only possible alleles. In step 738, each of these possible allele combinations is added to the selected subset for further deconvolution analysis. In other words, for each of these allele combinations, the deconvolution analysis proceeds to the next iteration to analyze the next unidentified contributor.

[0119] Fig. 7AAn example method 700 for selecting a subset of allele combinations for further analysis for the unidentified contributors of the current analysis is illustrated. It should be understood that additional steps can be added to the method 700, existing steps can be removed from the method 700, and / or the order of the steps can be changed. For example, steps 704 and 708 can be performed in parallel, and steps 714 and 718 can also be performed in parallel.

[0120] use Fig. 7A The above selection method shown selects fewer genotypes than the total number of genotypes potentially associated with loci in the general population for further deconvolution analysis. Therefore, only deconvolution analysis is required for scenarios of allele combinations that can at least potentially be used to explain the mixed DNA evidence spectrum of one or more unidentified contributors. In other words, the above selection method eliminates many allele combinations that are random or unlikely to explain the mixed DNA evidence spectrum. Therefore, full deconvolution analysis is more efficient and can provide analysis results more quickly.

[0121] Return to reference Figure 5 For each allele combination in the allele combination of the selected subset genotypes, step 509 of method 500 calculates a theoretical contribution from each allele combination in the allele combination. Step 510 stores the theoretical contribution for use in calculating a theoretical spectrum. Figure 8 An example method 800 for calculating theoretical contributions is illustrated. Method 800 may be used to implement Figure 5 Steps 509 and 510 of method 500.

[0122] Method 800 can be performed for each of the allele combinations of the corresponding unidentified contributors at the locus. For example, for each unidentified contributor, one or more allele combinations can be selected (e.g., (55), (56), (57), etc.). Then, for each of these allele combinations of the particular unidentified contributor, method 800 can be performed to calculate the theoretical contribution of the alleles. The process then proceeds to Figure 5The next iteration (step 511) shown in, to determine the next unidentified contributor and select allele combination for the next unidentified contributor. And method 800 can be repeated for all allele combinations to calculate their theoretical contribution. As an example, assuming that for the first unidentified contributor, the first allele combination is (5 5), method 800 is performed to calculate the theoretical contribution of this allele to (5 5). Then, the process goes to the next iteration of the next unidentified contributor and selects the allele combination of (6 6). Then method 800 is performed to calculate the theoretical contribution of this allele to (6 6). Further assuming that there are only these two unidentified contributors in this scene, the theoretical contribution of the calculation is stored for further calculating the theoretical spectrum, which will be described in more detail below.

[0123] refer to Figure 8 , for an allele combination (e.g., 55) of the corresponding unidentified contributor, step 802 calculates the peak corresponding to the allele in the allele combination. In particular, in step 802, the theoretical peak height is calculated using the contribution ratio scenario, sample degradation rate, and peak height imbalance (PHI) value of the corresponding unidentified contributor currently analyzed. For example, if the sum of the total peak heights in the adjusted evidence spectrum is 1000 and the contribution rate of the corresponding unidentified contributor is 80%, the theoretical peak height of the homozygous allele combination (e.g., 5 5) is calculated to be 400. The theoretical peak height can be scaled or adjusted by the sample degradation rate and / or peak height imbalance value. Using the theoretical peak height, the expected peak is calculated.

[0124] Step 804 stores the calculated peak for calculating the corresponding theoretical spectrum or directly adds the calculated peak to the corresponding theoretical spectrum. Step 806 determines whether the locus of current analysis has a shadow peak. This determination can be performed using, for example, a shadow model for the locus of current analysis. If the answer is "yes", then in step 808, calculate the shadow peak. For example, depending on the locus, one or more of a forward shadow peak, a reverse shadow peak, a double reverse shadow peak, a half forward shadow peak, and a half reverse shadow peak can be calculated. Step 810 stores the shadow peaks of these calculations for calculating the corresponding theoretical spectrum or directly adds them to the corresponding theoretical spectrum. Then, method 800 can proceed to the next iteration of the next unidentified contributor.

[0125] Return to reference Figure 5After storing the theoretical contribution calculated for the current unidentified contributor in step 510, method 500 proceeds to step 511 to determine whether there are more unidentified contributors in the current contribution rate scenario. If so, method 500 proceeds to step 513 for the next unidentified contributor. In step 513, the most recently analyzed unidentified contributor is considered a known contributor. Next, the theoretical contribution calculated for the current unidentified contributor is subtracted from the adjusted evidence spectrum (just as the known contributors are subtracted from the original evidence spectrum as described above). The further adjusted evidence spectrum (i.e., minus the contributions from the known contributors and the most recently analyzed unidentified contributors) is used for the next iteration. Method 500 continues from step 504 to repeat the process as described above.

[0126] Still reference Figure 5 , if the answer in step 511 is "no" (i.e., there are no more unidentified contributors in the current contribution rate scenario), the method 500 proceeds to step 512 to determine whether there are more contribution rate scenarios. As described above, even if the number of contributors remains the same, different combinations of contribution rate scenarios may exist. For example, for the same 3 contributors, the contribution rate scenarios may include 1-1-8, 1-2-7, 1-3-6, 1-4-5, 2-2-6, 2-3-5, 2-4-4, etc. Therefore, it may be necessary to analyze each of these contribution rate scenarios.

[0127] If step 512 determines that there are more contribution scenarios, method 500 proceeds to step 514, where the next contribution scenario is considered the current contribution scenario. The process is then repeated from step 502, where the original evidence spectrum is adjusted by subtracting the spectrum of the known contributors. The rest of the process is similar to those described above and therefore will not be repeated.

[0128] If step 512 determines that there are no more contribution scenarios, method 500 proceeds to step 515. In step 515, the process continues to calculate the theoretical spectrum using the theoretical contributions of all allele combinations and contribution scenarios previously stored. The calculated theoretical spectrum is compared with the original evidence spectrum and scored. The process of calculating and scoring the theoretical spectrum is described below.

[0129] Fig. 9It is a flowchart illustrating an example method 900 of a theoretical spectrum of an allele combination of the contribution rate of the current analysis for calculating the locus. Method 900 can be repeated to calculate the theoretical spectrum at each allele combination, each contribution rate scenario and each locus. Step 902 of method 900 calculates the theoretical contribution of the allele corresponding to all known contributors. As described above, for a known contributor, his or her DNA spectrum can be determined by, for example, analyzing a biological sample from the known contributor. Then the peak height in the DNA spectrum of the known contributor can be scaled according to the contribution rate scenario of the current analysis (for example, scaling according to the percentage of the known contributor relative to other contributors). The scaled DNA spectrum of the known contributor is used as the theoretical spectrum of the known contributor. The theoretical spectrum represents the expected contribution from the known contributor. In some embodiments, the theoretical shadow peak is added to the theoretical spectrum of the known contributor. As described above, the shadow peak can be calculated based on a shadow model constructed empirically or statistically.

[0130] Step 904 of method 900 obtains the stored theoretical contributions of the alleles corresponding to the unidentified contributors in the corresponding allele combination. As described above, these theoretical contributions can be calculated using, for example, method 800. These theoretical contributions of unidentified contributors include calculated allele peaks and calculated shadow peaks.

[0131] The step 906 of method 900 uses the theoretical contribution from unidentified contributors and the theoretical contribution from known contributors to calculate the theoretical spectrum of corresponding allele combination, corresponding contribution rate scenario and corresponding locus.For example, theoretical spectrum can be constructed by merging the peak (for example, allele peak and shadow peak) of unidentified contributors and the peak of known contributors.Use theoretical spectrum, spectrum score can be calculated.The matching degree between spectrum score indication theory spectrum and evidence spectrum.

[0132] Fig.10 is a flow chart illustrating an example method 1000 for calculating scores for bins in a theoretical spectrum. The method 1000 may be repeated to calculate a score for each bin, and the bin scores may then be used to calculate a spectrum score. Fig.10 As shown, step 1002 calculates the signal strength in a bin of the evidence spectrum and the signal strength in a corresponding bin of the theoretical spectrum. A bin is a measurement interval along the horizontal axis of the spectrum. The signals in the bins are measured (e.g., integrated or summed) together. Fig.11 As an example, the horizontal axis of a spectrum (theoretical spectrum or evidence spectrum) represents the number of alleles. The bins can be configured or customized to have any desired interval (eg, less than the peak width, equal to the peak width, or greater than the peak width). Fig.11The theoretical spectrum 1110 and the evidence spectrum 1120 are shown to be configured with approximately 43 bins (represented by the intervals between the short vertical lines below the horizontal axis). The signal in a particular bin may or may not have an allele peak. For example, the bin between the peaks of alleles 8 and 9 in the theoretical spectrum 1110 (e.g., bin #16) has only noise and / or only shadow peaks. Regardless of whether the signal in the bin has noise, shadow peaks, and / or allele peaks, the signal strength in the bin can be calculated. Fig.11 In the example shown, the vertical axis represents signal intensity in RFU (relative fluorescence units).

[0133] Return to reference Fig.10 , step 1004 determines whether the signal intensity in the bin of the theoretical spectrum and the corresponding bin of the evidence spectrum is lower than the analysis threshold AT. The analysis threshold AT is typically configured to be higher than the noise baseline of the analytical instrument. Therefore, the analysis threshold AT is a threshold used to distinguish signal from noise. Examples of analysis threshold AT are 50RFU or 75RFU. In probabilistic genotyping, lower values ​​of the analysis threshold AT (e.g., 5RFU or 10RFU) are used to reduce or eliminate the possibility of unexpected loss of shadow peaks. The analysis threshold AT can be user-configured or automatically configured based on statistics or models. If the signal intensity in a bin is lower than the analysis threshold AT, the particular bin may include only noise, and therefore there is no need to calculate a score for that bin. The analysis can simply move to the next bin. Using Fig.11 Bin #16 is shown as an example, this particular bin of the theoretical spectrum 1110 and the corresponding bin of the evidence spectrum 1120 has only noise but no allele peaks or shadow peaks. Therefore, the calculation of the signal intensity in bin #16 can be skipped. The process can move to the next bin to calculate the score.

[0134] If a particular bin has a signal strength above the analysis threshold ΔT in one or both of the evidence and theoretical spectra, then the bin may not only consist of noise. Fig.10 As shown, method 1000 proceeds to step 1006. Step 1006 sets the initial genotype probability of the bin to 1. Genotype probability (given by P freq ) represents the frequency of a particular allele. In other words, it represents the probability of seeing the same allele N times. The probability of seeing a particular allele (e.g., allele #8) once represents whether the allele is a common allele seen in a large population (e.g., 80% of the general population), a rare allele seen only in a small population (e.g., 2% of the general population), or anything in between. The probability of seeing the same allele N times (i.e., P freq ) is therefore the probability of seeing the allele once raised to the power of N. For example, if the probability of seeing the allele once is 0.2 and the allele is selected twice, then P freqis 0.04. The probability of seeing an allele once is also adjusted to account for the relationship between the contributors (e.g., if one contributor is a sibling of another contributor, this probability increases significantly). Adjustment of the probability of seeing an allele once can be performed using the probability adjustment parameter and is described in more detail below.

[0135] Method 1000 then proceeds to step 1008 to determine the number of alleles in the bin. Fig.11 In the example shown, bin #36 has a signal corresponding to an allele peak in both the evidence spectrum 1120 and the theoretical spectrum 1110. Therefore, the number of alleles in bin #36 is determined to be 1 for both the evidence spectrum 1120 and the theoretical spectrum 1110. Therefore, for bin #36, the total number of alleles is 2, which explains the evidence spectrum 1120 and the theoretical spectrum 1110. For another example, for bin #4, the number of alleles in the theoretical spectrum 1110 is "0". However, the number of alleles in this bin in the evidence spectrum 1120 is "1". Therefore, the total number of alleles for bin #4 is 1.

[0136] refer to Fig.10 , method 1000 next proceeds to step 1010 to determine whether the number of alleles in the bin is greater than 0. Continuing the above example of bin #4 and bin #36, the answer in step 1010 is "yes" for both bins. In the example of bin #4, only the evidence spectrum 1120 has an allele peak (allele 7). There is no allele peak in bin #4 of the theoretical spectrum 1110.

[0137] Still refer to Fig.10 If the number of alleles in a particular bin is greater than 0, then method 1000 proceeds to step 1012 to determine one or more probability adjustment parameters. Using the probability adjustment parameters and a model (e.g., the Balding and Nichols model), step 1014 of method 1000 calculates the genotype probability (P freq ). The Balding and Nichols model is a probability model that takes into account the shared ancestry in a population. In this model, it is assumed that the distribution of alleles in a population is known. In order to account for the common ancestor of individuals, a probability adjustment parameter (represented by θ) is used. Typically, θ is in the range of 0.01 to 0.03. By using the probability adjustment parameter θ, the genotype probabilities are perturbed away from the genotype probabilities obtained under the Hardy-Weinberg equilibrium.

[0138] As mentioned above, the genotype probability P Freq represents the probability of seeing the same allele N times. In general, for a given locus, the allele frequency in a population can be represented by K alleles (A1, A2, ..., A k ) of the vector P = (p1, p2, ..., p k) represents. Randomly select an A k The probability of a typos allele is P Freq =p k Assuming that the second allele of the same type has been selected once, the probability of randomly selecting this allele is P Freq =θ+(1-θ)p k Assuming that the third allele of the same type has been selected once, the probability of randomly selecting this allele is P Freq =[2θ+(1-θ)p k ] / (1+θ). Generally speaking, assuming that the a k Alleles have been selected k -1 times, then the probability of randomly selecting this allele is P Freq =[(a k -1)θ+(1-θ)p k ] / [(a k -1)θ+1-θ]. By using the Balding and Nichols model, the initial genotype probabilities are adjusted by θ and the adjusted probability P is calculated. Freq In some embodiments, multiple values ​​of θ may be used because different known or unknown contributors may each have different θ. Thus, a different value of θ may be used for each contributor. After step 1014 for calculating genotype probabilities, method 1000 proceeds to step 1016.

[0139] refer to Fig.10 If, in step 1010, the number of alleles in a bin is determined to be 0 (ie, no allele peak exists in the bin), method 1000 may skip steps 1012 and 1014 and proceed directly to step 1016. Step 1016 determines whether the evidence spectrum has a missing peak. Fig.11 Such a scenario is illustrated using bin #28. By comparing the signal intensity of bin #28 of theoretical spectrum 1110 and the signal intensity of bin #28 of evidence spectrum 1120, it can be determined that evidence spectrum 1120 does not have a peak at bin #28. It can also be determined that theoretical spectrum 1110 does have a peak at bin #28. In other words, there is a missing peak at bin #28 in evidence spectrum 1110.

[0140] refer to Fig.10 If there is a missing peak in the evidence spectrum, method 1000 proceeds to step 1018 to calculate the peaks represented by P 脱扣The trip probability represented by . In some embodiments, before calculating the trip probability, step 1018 also checks whether the possible trip peak is above the random threshold ST, and if so, it is determined that the trip probability does not need to be calculated or the trip probability can be assumed to be zero. The random threshold ST is generally used to explain the signal or noise that may occur accidentally due to various factors and variables associated with analyzing biological samples. If the peak is above the random threshold ST, the probability that the peak is caused by noise or is a trip peak is almost zero. If the peak is below the random threshold ST, there is a non-zero probability that the peak may be a trip peak. If step 1018 determines that the trip probability needs to be calculated, the model or distribution is continued to be used to calculate the trip probability. When calculating the trip probability, distinguish between allele tripping or shadow tripping. A specific peak may include contributions from both the allele peak and the shadow peak. Therefore, the trip probability of the allele peak and the trip probability of the shadow peak can be calculated separately. The calculation of the trip probability can be performed by applying one or more models, error functions, etc. to the theoretical spectrum of the current analysis. For example, based on a gamma distribution that models the peak height imbalance (PHI) ratio, for the expected height of the allele peak (e.g., about 400RFU), the probability that the peak height is below the analysis threshold AT (i.e., the probability of seeing a peak between zero and AT) can be calculated. This probability can be calculated by calculating the integral of the probability density between zero and AT to obtain the cumulative probability density. The probability that the peak height is below the analysis threshold AT can be regarded as the trip probability. For shadow peaks, the trip probability can be similarly calculated using the distribution of the shadow peak.

[0141] If there are no missing peaks in the evidence spectrum, method 1000 proceeds to step 1020 to determine if there are missing peaks in the theoretical spectrum. Fig.11 Such a scenario using bin #4 is illustrated. By comparing the signal intensity of bin #4 of theoretical spectrum 1110 and the signal intensity of bin #4 of evidence spectrum 1120, it can be determined that theoretical spectrum 1110 does not have a peak at bin #4. It can also be determined that evidence spectrum 1120 has a peak at bin #4. In other words, there is a missing peak at bin #4 in theoretical spectrum 1110. Or in other words, there is an extra peak at bin #4 in evidence spectrum 1110.

[0142] refer to Fig.10 If there are missing peaks in the theoretical spectrum, method 1000 proceeds to step 1022 to calculate the peaks 插入The insertion probability represented. In some embodiments, before calculating the insertion probability, step 1022 also checks whether the possible insertion peak is higher than the random threshold ST, and if so, it is determined that it is not necessary to calculate the insertion probability or it can be assumed that the insertion probability is zero. Similar to the above, if the peak is higher than the random threshold ST, the probability that the peak is caused by noise or the insertion peak is almost zero. If the peak is lower than the random threshold ST, there is a non-zero probability that the peak may be the insertion peak. If step 1022 determines that it is necessary to calculate the insertion probability, then continue to use the model or distribution to calculate the insertion probability. The insertion peak may be caused by noise. In some embodiments, the calculation of the insertion probability can be performed by applying one or more models, error functions, user-defined distributions, etc. to the theoretical spectrum of the current analysis. For example, the noise probability can be calculated based on the user-defined distribution (for example, lognormal distribution) depending on the DNA amount and the locus. The noise probability can be used as the insertion probability. For example, based on the distribution (for example, lognormal distribution) of the random variable modeling, the probability of seeing the noise peak with a height equal to or greater than the insertion peak height can be calculated by calculating the integral of the probability density from zero to a specific peak height. Therefore, the taller the peak, the less likely it is that the peak is caused by noise alone.

[0143] If there is no missing peak in both the theoretical spectrum and the evidence spectrum at a particular bin, method 1000 proceeds to step 1026 to calculate the peak height mismatch probability. Fig.11 Such a scenario using bin #20 is illustrated. By comparing the signal intensity of bin #20 of theoretical spectrum 1110 and the signal intensity of bin #20 of evidence spectrum 1120, it can be determined that bin #20 of theoretical spectrum 1110 has a peak that is smaller than the corresponding peak at bin #20 of evidence spectrum 1120. A peak height mismatch occurs when there is a peak height mismatch between the theoretical spectrum and the evidence spectrum at the corresponding bin. Although Fig.11 It is shown that at box #20, the peak height in the theoretical spectrum 1110 is less than the peak height in the evidence spectrum 1120, but it should be understood that other ways are also possible. It should also be understood that for one or both of the allele peak and the shadow peak, a peak height mismatch may occur. The peak height mismatch probability can be calculated by applying one or more models, error functions, etc. to the theoretical spectrum of the current analysis. The error function can include, for example, an error square function (e.g., squaring the peak height difference and normalizing it to 0-1), a percentage difference function, and / or a function for calculating a probability ratio based on a distribution (e.g., using a gamma distribution to calculate the ratio of the probability of a peak with a peak height in the theoretical spectrum and in the evidence spectrum). After calculating the peak height mismatch probability, method 1000 proceeds to step 1028.

[0144] Step 1028 uses one or more genotype probabilities (P Freq), trip probability, insertion probability, peak height mismatch probability, or a combination thereof to calculate the bin score. For example, if there is a trip, the bin score is equal to P Freq and P 脱扣 Similarly, if there is an insertion, the score of the bin is equal to P Freq and P 插入 If there is no tripping or insertion, the score of the box is equal to P Freq and P 峰_高_错配 The product of .

[0145] Method 1000 can be repeated for each bin of the spectrum to calculate a score. Using the scores of all bins, step 1030 calculates a spectrum score for the theoretical spectrum of the locus. For example, the spectrum score can be the product of the scores of all bins. Then, scores for multiple theoretical spectra corresponding to multiple scenarios can be obtained by repeating the above method 1000 and step 1030. These scores for multiple theoretical spectra can be sorted from high to low, with higher scores indicating a better match between the theoretical spectrum and the evidence spectrum.

[0146] In other embodiments, Fig.12 As shown, a heat map can be used to visualize the scores of multiple theoretical spectra. In heat map 1210, the score highlighted in green (e.g., represented by a line pattern from the upper right to the lower left) is a high score (e.g., 100%) indicating a good match between the theoretical spectrum and the evidence spectrum. The score highlighted in yellow (e.g., represented by a small square pattern) is a medium score (e.g., 70% to 79%) indicating a medium match between the theoretical spectrum and the evidence spectrum. The score highlighted in red (e.g., represented by a horizontal line pattern or a diamond pattern) is a low score (e.g., 50% to 60%) indicating a lower match between the theoretical spectrum and the evidence spectrum. Therefore, a heat map can be used to display the scores. It should be understood that the heat map can use any number of colors to represent different scores, thereby providing a visual display of scores for different contributor ratio scenarios. The highest-scoring theoretical spectrum can be used to provide the possibility of a match between an unidentified contributor and a person of interest (POI).

[0147] The above methods improve the efficiency of genotyping for forensic analysis, such as identifying POIs. By intelligently eliminating random or allele combinations that are unlikely to result in a good match, the amount of deconvolution analysis is significantly reduced. In turn, the analysis time is also reduced. Therefore, these methods result in improved efficiency without sacrificing accuracy.

[0148] Fig.13 An example user interface 1300 for receiving user input and displaying deconvolution analysis results is illustrated. Fig.13As shown, user interface 1300 may include input fields, such as unknown number, ratio increment, etc. The unknown number field allows the user to input the number of unidentified contributors. The ratio increment field allows the user to input the contribution rate resolution step (e.g., 10%, 5%, etc.). User interface 1300 also provides menu items or selectable fields for viewing detailed information. For example, on the left side of user interface 1300, several menu items and submenu items are provided. These menu items include trace / model, matching, contributor, possibility, mixture, etc. When one of these menu items or submenu items is selected, user interface 1300 can display relative information. For example, if the "model" menu is selected, user interface 1300 can display one or more models 314, including distributions related to calculating shadow peaks, noise, shoulders, balance between loci (ILB), peak height imbalance or peak height ratio (PHR), population frequency and attenuation, etc. For example, user interface 1300 can provide detailed information about shadow models, PHR models, ILB models, noise models, population statistics, etc. used in deconvolution analysis. For another example, if the "Contributor" menu item is selected, the user interface 1300 displays detailed information about the contributor, including, for example, his role (victim, suspect, etc.), allele information for each locus, clues, whether it is known or unknown, etc.

[0149] The user interface 1300 may also display the deconvoluted spectrum (eg, theoretical spectrum) along with the evidence spectrum, such as Fig.13 As shown. In this way, user interface 1300 provides a visual direct comparison between evidence spectrum 1310 and theoretical spectrum 1320. In some embodiments, user interface 1300 also provides contribution rate information and locus information for a particular theoretical spectrum. It should be understood that Fig.13 This is merely an illustration of a possible user interface 1300. The user interface 1300 may include more or fewer icons, menu items, tables, charts, graphs, etc., and may display them in any desired manner.

[0150] Fig.14 A flow chart illustrating a method 1400 for predicting the number of contributors to a biological sample using an evidence profile obtained from the biological sample. The method 1400 may be used to implement, for example, Figure 3A Steps 320 and 330 are shown. Fig.14 In step 1402, from a system that generates an evidence spectrum (e.g., Figure 1 The system shown in FIG. 10 ) obtains the evidence spectrum or data representing the evidence spectrum. Figure 2AIn one example shown, the evidence spectrum includes data representing peaks at multiple loci. Peaks in the evidence spectrum can include allele peaks and / or shadow peaks. Shadow peaks are small peaks that usually appear one repeat unit before or after the allele peak. During PCR amplification, when replicating the DNA chain, the polymerase can lose its position, such as sliding forward or backward by multiple base pairs, thereby causing shadow peaks.

[0151] Step 1404 of method 1400 receives an indication of a plurality of selected NOC models. As described above, for example, a user interface may be provided for a user to select a plurality of NOC models. An indication of which NOC models are selected is received, and subsequent processes are performed using the selected NOC models. Using a combination of multiple NOC models can generally provide a more accurate NOC prediction than using a single model.

[0152] Step 1406 of method 1400 obtains the weight assigned to each selected NOC model. The weight is used to improve the NOC prediction accuracy. For example, if the first NOC model generally or in some cases tends to provide more accurate results than the second NOC model, the first NOC model can be assigned a higher weight than the second NOC model. In some embodiments, a NOC model that provides more accurate predictions in some cases may not provide accurate predictions in other cases. Many variables may affect the accuracy of a particular model, including, for example, equipment changes, the training status of the NOC model, the past prediction performance of the NOC model in similar or different situations, the features selected for the model, etc. Therefore, relying on a single model to predict NOC in all cases may not produce accurate or acceptable results. Using a combination of multiple NOC models with assigned weights can reduce the possibility of erroneous predictions or inaccurate predictions.

[0153] In some embodiments, the weights assigned to each NOC model also reflect the user's preferences or experience. For example, if the user decides that a particular NOC model is the most credible based on past experience, a higher weight may be assigned to that particular NOC model. The weights may be any desired numbers that collectively reflect the estimated ability or evaluation ability of the various NOC models to produce accurate NOC predictions. In one example, the weights assigned to the peak count distribution model, the ANN model, and the decision tree model are 0.9, 0.2, and 0.5. In this example, the peak count distribution model is evaluated to produce more accurate NOC predictions than the other two models, and is therefore assigned a higher weight. If the NOC estimates are different between different models, the estimate produced by the peak count distribution model may be more credible than the other models.

[0154] Still reference Fig.14 In some embodiments, method 1400 continues by obtaining contextual information from the evidence spectrum in step 1408. Fig.15 , such contextual information includes, for example, parameters 1502, total peak height 1504, and population statistics 1506. Parameters 1502 are associated with the test kit and / or test machine used to obtain the evidence spectrum. Different test kits and / or test machines may result in different models and statistics. These models and statistics may include one or more of a shadow model, a PHR model, an ILB model, a noise model, etc. Therefore, knowing the specific test kit and / or test machine used for the specific biological sample from which the evidence spectrum was obtained, an appropriate statistical model can be generated, or an appropriate statistical model can be selected from a model / statistics database for subsequent generation of a calculated sample spectrum.

[0155] Total peak height 1504 is the sum of the peak heights of the peaks in the evidence spectrum. The total peak height of the peaks in the evidence spectrum is a type of context information used to generate the calculated sample spectrum. The total peak height represents the total signal intensity and is related to the amount of DNA contained in the biological sample from which the evidence spectrum was obtained.

[0156] In some embodiments, population statistics 1506 are associated with the evidence spectrum and obtained as part of the contextual information. Different contributors to a biological sample may come from different populations. Therefore, for a particular evidence spectrum, the genotypes corresponding to the contributors may include fewer genotypes than the total number of genotypes potentially associated with the current locus in the general population. For example, certain alleles may be common in a large population (e.g., 80% of the general population), while other alleles may only be seen in a smaller population (e.g., 2% of the general population). In addition, population statistics may also include models that take into account common ancestry in a population. Therefore, for different evidence spectra obtained from different biological samples with different contributors, appropriate population statistics are obtained for subsequent generation of calculated sample spectra.

[0157] Return to reference Fig.14 Based on the context information, parameters and model, step 1410 of method 1400 generates a calculated sample spectrum, sometimes also referred to as a simulated sample spectrum. When using a NOC model to estimate NOC probabilities in a more accurate manner, a calculated sample spectrum is required to provide sufficient training for the NOC model and / or to provide enough data points to obtain a reasonably accurate statistical distribution for NOC estimation. Typically, the data available from real biological samples is limited and not always available in large quantities. Therefore, the real sample spectrum may not be sufficient to train a machine learning-based NOC model and / or to generate a statistically significant distribution. In order to alleviate the lack of data available from the real sample spectrum, simulations can be performed to generate a calculated sample spectrum.

[0158] Fig.15 A method 1500 for generating a sample spectrum based on context information, parameters and / or models is illustrated. The method 1500 may be used to implement Fig.14At least step 1410 of method 1400 in. In some embodiments, in order to generate the calculation sample spectrum, the simulation engine establishes, updates and / or uses the simulation model. Step 1508 of method 1500 provides various context information, parameters and models to the simulation engine to establish or update the simulation model for generating the calculation sample spectrum. For example, step 1508 can provide context information to the simulation engine, including, for example, distribution related to the calculation shadow peak (forward shadow peak, reverse shadow peak, half shadow peak), noise mode defined for each locus, shoulder, balance between loci (ILB), total peak height (representing the total amount of DNA in the evidence spectrum), peak height ratio (PHR), population frequency and attenuation (for each locus and each allele) etc. The simulation engine uses context information to establish and / or update the simulation model. For example, the simulation engine can establish and / or update the simulation model by obtaining some available spectra (for example, real sample spectrum or previously calculated spectrum) and applying parameter changes to these available spectra. These parameter changes can be predefined as being in certain boundaries. For example, if the simulation engine changes the allele peak position, the change is limited to those possible or very likely allele quantities of any particular locus. For another example, if the simulation engine changes the total peak height and assigns the peak height to each peak in the calculation spectrum, the total peak height change can be limited to the specific percentage of the peak height change of the evidence spectrum. By defining the range of variation of the parameters, the simulation engine can establish and / or update the simulation model so that it generates a calculation spectrum similar to the evidence spectrum. In some embodiments, the simulation model uses the same or similar distribution (for example, normal distribution, gamma distribution, lognormal distribution), which is also used to define the statistical model of some context information. For example, if the statistical model related to the shadow peak is defined using the normal distribution, the simulation model can use the same distribution to generate the simulated shadow peak with the same normal distribution. Similarly, the simulation model can use the statistical model related to noise to generate the simulated noise with the same distribution. It should be understood that the simulation model can be configured with any desired rules, strategies, restrictions, etc.

[0159] Using the simulation model, the simulation engine generates a calculated sample spectrum with contextual information similar to the contextual information of the evidence spectrum. For example, the calculated sample spectrum can have a total peak height (which reflects the total amount of DNA) that is the same or similar to the total peak height of the evidence spectrum (for example, within a predetermined percentage change range). The calculated sample spectrum can also have a population statistics and shadow distribution similar to the population statistics and shadow distribution of the evidence spectrum. For example, if the biological sample from which the evidence spectrum is obtained has one or more known contributors, the calculated sample spectrum can be generated while considering the known peaks of the one or more known contributors. When generating the calculated sample spectrum, the simulation engine generates simulated peaks for all unknown contributors, and includes the known peaks of the one or more known contributors in the calculated sample spectrum. Depending on the contribution rate scenario, the known peaks of the one or more known contributors can have different peak heights in different calculated sample spectra. In some embodiments, the known peaks of the one or more known contributors can be allele peaks, and the simulation engine can generate simulated shadow peaks, noise, etc. Therefore, the calculated sample spectrum and the evidence spectrum have at least some commonalities with respect to contextual information. The commonality between the evidence spectrum and the calculated spectrum is beneficial to improve the accuracy of the NOC prediction process, because the calculated sample spectrum is used for subsequent training of the machine learning-based NOC model and / or for obtaining the statistical distribution for predicting NOC.

[0160] In some embodiments, the simulation engine builds and / or updates the simulation model based on a set of samples from which one or more evidence profiles are obtained. In one embodiment, the simulation model is built or updated by using the default settings from a specific laboratory. If some loci or some alleles in the evidence profile differ from those of the default settings, the default settings can be adjusted to take into account the contextual information of the specific evidence profile.

[0161] In some embodiments, step 1510 of method 1500 receives user input providing an expected maximum number of contributors. The expected maximum number of contributors can be used to establish an upper limit on the likelihood of a NOC. For example, if a user has information to believe that a biological sample has contributions from a victim, a suspect, and no more than two or three witnesses, the user can specify that the number of possible contributors is no greater than five. The expected maximum number of contributors can limit the range of possible NOCs. In turn, this reduces the requirements for the number of calculated sample profiles to be generated, reduces the computational workload, and increases the likelihood of providing an accurate NOC prediction. The user input providing the expected maximum number of contributors can be parameter 316 ( Figure 3A ).

[0162] In step 1512 of method 1500, for each possible number of contributors that is less than or equal to the expected maximum number of contributors, one or more user inputs are received that provide a requested number of calculated sample spectra to be generated. For example, if the expected maximum number of contributors is 3, the possible number of contributors can be 1, 2, or 3. For each of the possible numbers of contributors, the user input specifies the requested number of calculated sample spectra to be generated. For example, the user input can specify that for each of the possible numbers of contributors 1, 2, or 3, 1000 calculated sample spectra should be generated. Therefore, the total number of calculated sample spectra is 3000. As another example, the user input can specify different numbers of calculated sample spectra (e.g., 500, 1000, 1500) to be generated for different possible numbers of contributors (e.g., 1, 2, and 3). The user input that provides the requested number of calculated sample spectra to be generated can be parameter 316 ( Figure 3A ).

[0163] In some embodiments, the requested number of calculated sample spectra to be generated may need to meet a threshold number. For example, the threshold number may be 500. Thus, if the user input specifies a number less than 500, the threshold number or a default number may be used instead. Using the threshold number reduces the likelihood that the user input may specify a number that is too low, such that the number of calculated sample spectra is insufficient to perform subsequent training of the model or to obtain a statistically significant distribution for NOC prediction.

[0164] Based on the calculated sample spectrum of the requested number, step 1514 of method 1500 generates the calculated sample spectrum of the requested number. The calculated sample spectrum is a simulated spectrum generated by using a simulation model. As described above, the simulation engine can generate a calculated sample spectrum with contextual information similar to the contextual information of the evidence spectrum. Therefore, these simulated spectra and the evidence spectrum have at least some commonalities about the contextual information. For example, these simulated spectra can have a total peak height that is the same or similar to the total peak height of the evidence spectrum, which represents the total amount of DNA. The simulated spectrum can have the same or similar shadow peak statistics as the evidence spectrum. The simulated spectrum can have an allele peak with the same or similar population statistics as the evidence spectrum. In one embodiment, the simulated spectrum is generated so that their contextual information is as close to the evidence spectrum as possible. Therefore, the calculated sample spectrum can be used as if they were a spectrum obtained from a real biological sample. These calculated sample spectra can therefore be used for subsequent processing (e.g., training of a model based on machine learning) to improve the accuracy of NOC prediction.

[0165] Return to reference Fig.14As described above, step 1404 of method 1400 receives an indication of a selected NOC model (e.g., two or more of an ANN model, a decision tree model, a random forest model, and a peak distribution model). For each selected NOC model in the selected NOC models, step 1412 of method 1400 determines whether the model is a machine learning-based model. If the answer is "yes", method 1400 proceeds to step 1416. For example, if the user selects an ANN model, a decision tree model, and / or a random forest model, method 1400 determines that the models are machine learning-based NOC models. Based on such determinations, step 1416 of method 1400 uses the generated calculation sample spectra to form a training data set, a test data set, and a validation data set. The calculation sample spectra can be divided in any desired manner to form a data set. For example, 1000 calculation sample spectra can be divided into 400 training data sets, 300 test data sets, and 300 validation data sets, respectively. These data sets can therefore be used to train the selected machine learning-based NOC model or update its training. In some embodiments, the number of computed sample spectra is large enough to provide training, testing, and validation data sets and large enough that a machine learning-based NOC model can be trained to a number sufficient to provide accurate NOC predictions (e.g., 1000 or more).

[0166] refer to Fig.14 In some embodiments, in order to train the selected machine learning-based NOC model, step 1418 of method 1400 determines one or more features of the computed sample profile. Fig.16A , Fig. 16B , Fig.17A , Fig. 17B and Fig.18Describe the determination of features in more detail. Step 1420 uses the features determined from the calculated sample spectrum to train the selected NOC model based on machine learning or update its training. The training of the selected NOC model may be important because it affects the accuracy of NOC prediction. Sufficient training data is often required to avoid overfitting. In the present disclosure, a desired number of calculated sample spectra are generated, and features are extracted from these calculated sample spectra. As described above, the calculated sample spectrum is a simulated spectrum similar to the real sample spectrum. Therefore, if there is not a sufficient number of real sample spectra, the calculated sample spectrum can be used to provide enough training data for training the selected NOC model based on machine learning. In some embodiments, the real sample spectrum and the calculated sample spectrum are combined to provide enough training data for training the selected NOC model based on machine learning. Using the training data, the training of the model can be performed, and then tested and verified. In some embodiments, the calculated sample spectrum is provided for updating the previously trained model. For example, if the context information associated with the new evidence spectrum is different from the context information used to train the NOC model, the calculated sample spectrum generated based on the new evidence spectrum can be used to update the training. A calculated sample spectrum for a new evidence spectrum may be generated in a similar manner as described above based on the changed context information.

[0167] like Fig.14 As shown, after using the features extracted from the calculated sample spectrum to train each of the selected NOC models or update its training, step 1422 of method 1400 uses the trained (or training-updated) machine-learning-based NOC model to estimate the probability of multiple NOC possibilities. For example, the ANN model can have an input layer, one or more hidden layers, and an output layer. The ANN model can have any desired number of neurons in the hidden layer. The output layer of the trained ANN model provides the probability of several NOC possibilities up to the expected maximum number of contributors. If the expected maximum number of contributors is, for example, three, the trained ANN model provides the probability that the biological sample from which the evidence spectrum is obtained has 1, 2, or 3 contributors (e.g., 24%, 76%, 0%, respectively). Other trained machine-learning-based models can similarly provide the probability of multiple NOC possibilities.

[0168] As described above, for each of the selected NOC models, step 1412 of method 1400 determines whether the model is a machine learning-based NOC model. If the answer is "no", method 1400 proceeds to step 1414. Step 1414 also determines whether the selected NOC model is a peak count distribution model. The peak count distribution model is a statistical model that can be used to predict NOC based on comparing the total number of peaks to the peak count distribution of the number of possible contributors derived from a large number of calculated sample spectra. In method 1400, if the answer to the determination in step 1414 is yes, method 1400 proceeds to step 1424. Steps 1424, 1426, 1428, and 1430 are used to perform NOC prediction based on the peak count distribution model.

[0169] As described above, when generating the calculated sample spectra, method 1400 receives an expected maximum number of contributors (e.g., 3), which may represent a user's expectation or estimate that the NOC cannot be greater than the maximum number. Therefore, for each possible number of contributors that is less than or equal to the expected maximum number of contributors, step 1424 counts the number of peaks in each of the calculated sample spectra. As described above, the calculated sample spectra may be generated so that they include a first spectrum set corresponding to one contributor (e.g., 1000 spectra), a second spectrum set corresponding to two contributors (e.g., 1000 spectra), and so on. Therefore, in each set of these spectra, step 1424 may count the total number of peaks for each of the calculated sample spectra. For example, using the first spectrum set corresponding to one contributor, step 1424 may count the total number of peaks as 45, 50, 55, and 60 for various different spectra. In the second spectrum set corresponding to two contributors, step 1424 may count the total number of peaks as 65, 70, 75, and 79 for various different spectra. In this way, peak counts for different spectrum sets may be determined and all possible numbers of contributors may be considered.

[0170] When counting the number of peaks, step 1424 of method 1400 may need to distinguish signal peaks from noise. As described in more detail below, in some embodiments, method 1400 distinguishes signal peaks from noise by using a signal strength threshold (also referred to as a peak height threshold). If the signal strength (or peak height) of the data at a particular location of the spectrum is less than the signal strength threshold, it may be noise rather than a signal. Therefore, in some embodiments, step 1424 identifies signal peaks before counting the number of signal peaks.

[0171] Based on the number of peaks for each calculated sample spectrum with a corresponding possible number of contributors, step 1426 of method 1400 determines an expected peak count distribution. As described above, step 1424 determines the peak counts in each spectrum in different sets of spectra corresponding to all possible numbers of contributors. Based on these peak counts, the number of calculated sample spectra with a specific number of peak counts can be obtained. For example, for a first set of spectra corresponding to one contributor, step 426 can determine that there are about 80 spectra with a peak count of about 45; about 350 spectra with a peak count of about 50; about 300 spectra with a count of about 55; and so on. Similarly, for a second set of spectra corresponding to two contributors, step 1426 can determine that there are about 100 spectra with a peak count of about 65; about 380 spectra with a peak count of about 70; about 100 spectra with a count of about 75; and so on. Therefore, the expected peak count distribution for all possible numbers of contributors can be determined in a similar manner.

[0172] Fig.19 Example expected peak count distributions for multiple possible numbers of contributors are illustrated. Fig.19 , the horizontal axis represents the peak counts in the calculated sample spectra, and the vertical axis represents the number of calculated sample spectra. Fig.19 Curve 1902 in is an example expected peak count distribution associated with a spectrum having one contributor. Curve 1904 is an example expected peak count distribution associated with a spectrum having two contributors. Curve 1906 is an example expected peak count distribution associated with a spectrum having three contributors, and so on. These curves of expected peak count distributions provide a relationship between the number of spectra corresponding to the peak counts in each calculated sample spectrum set corresponding to the corresponding number of contributors. For example, curve 1902 indicates that for spectra corresponding to one contributor, the peak counts are between about 10 and about 65; and the majority of spectra have peak counts between about 45 and 60. Similarly, curve 1904 indicates that for spectra corresponding to two contributors, the peak counts are distributed between about 60 and 80; and the majority of spectra have peak counts between about 65 and 75. For Fig.19 Similar indications can be obtained for the other curves in .

[0173] Return to reference Fig.14 , step 1428 counts the total number of peaks in the evidence spectrum. Based on the expected peak count distribution and the total number of peaks in the evidence spectrum, step 1430 estimates the probability of multiple NOC possibilities. For example, Fig.19 As shown, step 1428 of method 1400 can determine that the evidence spectrum has a total peak count of about 62, as shown by line 922. This peak count is compared to the peak distribution (e.g., curves 902, 904, 906, 908, and 910). Based on the comparison results, the probability of the evidence spectrum having 1, 2, 3, 4, and 5 contributors is estimated. Fig.19 In the example shown, since the total peak counts of the evidence spectrum are between the peak count distributions of one contributor and two contributors, it is unlikely that the evidence spectrum in this example has more than two contributors. Therefore, the probability that the evidence spectrum has 3 to 5 contributors is estimated to be almost zero. The probability that the evidence spectrum has one or two contributors can be estimated to be, for example, 40% and 60%, respectively.

[0174] Return to reference Fig.14 , if step 1414 determines that the selected NOC model is not a peak distribution model, it proceeds to step 434 to use any other desired model to estimate the probability of NOC possibility. These models can include, for example, a model based on maximum allele count (MAC), an expert-based heuristic rule model and / or a model based on performing deconvolution analysis. The MAC-based model uses the MAC distribution extracted from the calculated sample spectrum. The heuristic rule model is a customized decision tree model defined by an expert. For example, the rule defined by the expert can specify that if there is a specific number of peaks (e.g., 3) in a specific locus (e.g., D8S1179), a specific action should be performed (e.g., a specific path in the decision tree should be followed or a specific classification should be generated). Based on the model performing deconvolution analysis, it can be used to generate a score using the deconvolution analysis results of all possible contributors. The score can be used to predict NOC.

[0175] As described above, the machine learning based NOC model (step 1422), the peak count distribution model (step 1430), and / or other non-machine learning based models (step 1434) may be used to estimate the probabilities of multiple NOC possibilities (e.g., 1, 2, 3 contributors, etc.). Because multiple selected NOC models are used to estimate the probabilities, their estimated results may or may not have the same or similar probabilities. In some embodiments, the method 400 may perform an evaluation of the NOC probabilities provided by multiple NOC models to generate a NOC prediction.

[0176] exist Fig.14In the step 1422, 1430 and / or 1434, the probabilities of multiple NOC possibilities are estimated. These estimated probabilities are provided to step 1432 for evaluation. In order to evaluate the estimated probabilities provided by different selected NOC models, step 1432 calculates the product of the weight assigned to the selected model and the estimated probability provided by the specific selected model for each selected model. The product result is called the scaled probability. As described above, in step 1406, the weight of the selected NOC model can be obtained. The weight can be assigned based on user input, default settings, past prediction results, etc. As an example, the weights assigned to the peak count distribution model, the ANN model, and the decision tree model are 0.9, 0.2, and 0.5, respectively. For each model, each estimated probability of multiple NOC possibilities is multiplied by the corresponding weight assigned to the specific selected NOC model. For example, one of the selected NOC models can be a peak count distribution model and can be assigned a weight of 0.9. For this peak count distribution model, the estimated probabilities of multiple NOC possibilities are, for example, 40% for one contributor, 50% for two contributors, 10% for three contributors, and 0% for four or more contributors. To obtain scaled probabilities (also called weighted probabilities), a weight of 0.9 is multiplied by these probabilities. Therefore, the scaled probabilities are 36% for one contributor, 45% for two contributors, 9% for three contributors, and 0% for four or more contributors. The scaled probabilities for other selected NOC models can be calculated in a similar manner. For example, the ANN model may estimate the probabilities of multiple NOC possibilities as 70% for one contributor, 20% for two contributors, 10% for three contributors, and 0% for four or more contributors. If the assigned weight of the ANN model is 0.2, the scaled probabilities are therefore 14% for one contributor, 4% for two contributors, 2% for three contributors, and 0% for four or more contributors. As another example, the decision tree model may estimate the probability of multiple NOC possibilities to be 20% for one contributor, 60% for two contributors, 20% for three contributors, and 0% for four or more contributors. If the decision tree model has an assigned weight of 0.5, the scaled probability is therefore 10% for one contributor, 30% for two contributors, 10% for three contributors, and 0% for four or more contributors.

[0177] Next, for each NOC possibility in the plurality of NOC possibilities, step 1436 of method 400 calculates a weighted sum using the scaled probabilities of the plurality of selected models. Using the above example, if the peak count distribution model, the ANN model, and the decision tree model are the selected models, the weighted sum of one contributor can be calculated by summing the scaled probabilities of one contributor in all three models. Continuing with the above example, the weighted sum of one contributor in all three models is calculated to be 60% (i.e., 36%+14%+10%). Similarly, the weighted sum of two contributors is calculated to be 79% (i.e., 45%+4%+30%). The weighted sum of three contributors is calculated to be 21% (i.e., 9%+2%+10%).

[0178] In some embodiments, step 436 may also calculate a normalized weighted sum, which is obtained by dividing the weighted sum of the scaled probabilities by the sum of the weights assigned to all selected models. Continuing with the above example, the sum of the weights assigned to the peak count distribution model, the ANN model, and the decision tree model is 1.6 (i.e., 0.9+0.2+0.5). Therefore, the normalized sums of the scaled probabilities for one contributor, two contributors, and three contributors in all three models are 37.5% (i.e., 60% divided by 1.6), 49.4% (i.e., 79% divided by 1.6), and 13.1% (i.e., 21% divided by 1.6), respectively.

[0179] Based on the weighted sum or the normalized weighted sum, step 1440 of method 1400 predicts the number of contributors associated with the biological sample from which the evidence spectrum is obtained. Continuing with the above example, step 1440 can sort the weighted sum or the normalized weighted sum of all possible numbers of contributors in descending order and / or determine the maximum value. In the above example, the weighted sum of two contributors is the highest value (i.e., 79% or 49.4, if normalized) among the weighted sums of all possible numbers of contributors. Therefore, method 1400 can predict that the biological sample from which the evidence spectrum is obtained is most likely to have two contributors.

[0180] In the above example, the peak distribution model has moderately higher probability estimates for two contributors (50%) than for one contributor (40%). And the ANN model has a much higher probability estimate for one contributor (70%) than for two contributors (20%). Therefore, the two models do not provide the same or similar estimates. By using the above-described NOC evaluation process that takes into account the weights assigned to each NOC model, the estimation differences between different NOC models can be resolved and the accuracy of NOC predictions can be improved. In the above example, because the weights assigned to the peak count distribution model are much higher than the weights assigned to the ANN model, the NOC prediction gives more weight to the probability estimate provided by the peak count distribution model. Therefore, method 400 can predict a NOC that is the same or similar to the NOC estimated by the peak count distribution model (e.g., the two contributors in the above example). Feature determination based on machine learning models

[0181] As described above, for each NOC model in the NOC model selected by the user, step 1412 of method 1400 determines whether the model is a NOC model based on machine learning. If the answer is "yes", method 1400 proceeds to step 1416. Step 1416 uses the calculated sample spectrum to form a data set, and then proceeds to step 1418 to determine one or more features from the calculated sample spectrum. Features are attributes of the spectrum (e.g., calculated sample spectrum, real sample spectrum, or evidence spectrum). The features used to train the NOC model based on machine learning can be a default feature set for training, a feature set used in past training, and / or a feature set selected by the user. The features of the spectrum include features that can be directly extracted from the spectrum and features that can be calculated or derived using the spectrum. Features that can be directly extracted from the spectrum include, for example, the total number of peaks, total peak height, etc. Features that can be calculated or derived from the spectrum include, for example, peak height distribution of the spectrum, relative peak height, the number of loci and / or markers for "N" peaks (N≥2), etc. If the features are complementary and / or orthogonal to each other, they are usually useful when training a NOC model based on machine learning. For example, DNA amount and double DNA amount are not orthogonal features because they are essentially the same attribute of the spectrum. Features are also generally more useful or more important if they are attributes that have an impact on predicting the target variable. For example, to estimate NOC probability, the day of the week or the first letter of the sample name may not have any impact on the estimate. However, the peak height distribution may have some impact on the estimate. Therefore, the peak height distribution feature is more useful.

[0182] In some embodiments, features for training a machine-learning-based NOC model are selected based on past experiments (e.g., validation data from a previously trained model), contextual information, and a particular machine-learning-based NOC model. To estimate NOC probabilities, for example, a set of features may be identified such that they are likely to provide good estimates using a reasonable number of computed sample spectra. The identified feature set also does not result in a substantial increase in required training data; require a more complex neural network; result in unacceptably long training times (e.g., days or weeks); or result in a significant risk of overfitting.

[0183] In some embodiments, the features used to train the NOC model based on machine learning are selected based on the type of sample. Different features can be used for different types of biological samples (e.g., DNA samples, RNA samples, protein samples, etc.). The same or different features can be used for different NOC models based on machine learning. For example, ANN models and random forest models can use different features so that they can each provide improved NOC estimates. In some embodiments, the features used to train the NOC model based on machine learning are derived from a large number of samples representing many or substantially all possible scenarios. For example, samples for training can be generated corresponding to many or substantially all possible scenarios, including different amounts of DNA, different test kits and / or machines, different population frequencies, etc. Therefore, the machine learning model is trained by using these samples that potentially cover substantially all possible scenarios. Therefore, the trained machine learning model can be used for many different types of biological samples and / or samples with various other different contextual information (e.g., DNA amount, test kit / machine, population, etc.). For example, if the update to the model cannot be easily performed, such a trained model is sometimes desired; limited device storage can be used to store many different models and data; and / or it is impractical to distribute many different models.

[0184] In some embodiments, features used to train machine learning-based models include the peak height distribution of the profile, the total number of peaks identified, the total peak height, the number of loci and / or markers with "N" peaks (N≥2), the percentage of the number of peaks with peak heights below a predetermined peak height threshold relative to the total number of peaks, etc. Each of these features is described in more detail below.

[0185] Fig.16A A method 1600 is illustrated for determining a characteristic of a peak height distribution of each calculated sample spectrum. The method 1600 may be used to implement Fig.14At least a portion of step 1418 of method 1400 shown. As described above, the characteristics of the peak height distribution can be useful and important features that may have an impact on the NOC probability estimate. Therefore, the peak height distribution characteristics are attributes of the spectrum that have some predictive properties. Fig.16A , method 1600 can be applied to each calculated sample spectrum (or actual sample spectrum) to determine the peak height distribution characteristics. For a particular spectrum, step 1602 of method 1600 identifies the peak in the spectrum. The identification of the peak in the spectrum can be based on a signal intensity threshold. For example, the signal intensity at each allele position (or shadow position) in the calculated sample spectrum can be compared with a predetermined signal intensity threshold. If the specific signal intensity is greater than or equal to the signal intensity threshold, the signal peak is identified. Fig. 16B For example, the signal strength of the signal at allele 13 of locus 1 is about 400RFU. If the predetermined signal strength threshold is 50RFU, the peak at allele 13 is identified. This process can be performed at all positions of all loci to identify peaks (alleles or shadows) by distinguishing signals from noise. The signal strength of a peak is also called peak height.

[0186] Return to reference Fig.16A In some embodiments, for each locus in which a sample profile is calculated, method 1600 performs steps 1606, 1608, 1610, 1612, and 1614. Step 1606 determines the maximum peak height at the locus. For example, for Fig. 16B For the locus 2 in the calculated sample spectrum 1640 shown, for the peak at allele 30, the maximum peak height can be determined to be about 800 RFU. Next, step 1608 calculates the percentage of the peak height relative to the maximum peak height for the first or next peak at the locus. Continue Fig. 16B In the example of , at locus 2, step 1608 determines that the peak height of the peak at allele 25 is about 400RFU, and then calculates the percentage of the peak height relative to the maximum peak height (i.e., 50%). Next, step 610 determines the corresponding histogram bin based on the calculated peak height relative to the maximum peak height. The histogram bins can be automatically determined or selected by the user. In some embodiments, the increment of the histogram bins is predetermined to be 5%. If the horizontal axis of the histogram represents the number of histogram bins, the bin counts are arranged in increments of 5%. Fig.17A and Fig. 17B Two example histograms 710 are shown ( Fig.17A ) and 720( Fig. 17B. In the above example, after step 1608 calculates that the percentage of the peak height of the peak at allele 25 is 50%, step 1610 determines that the corresponding histogram bin is bin #50. Next, step 1612 increases the number of peak counts in the corresponding bin by one. Thus, for example, the peak count in the 50% bin of Figure 1710 is increased by one. If the percentage of the peak height is a number between two bins (e.g., between bin #45 and bin #50), the percentage of the peak height is rounded up or down and the corresponding bin is identified. For example, if the percentage is 46%, the corresponding bin is identified as bin #45. The peak count in bin #45 is then increased by one. Similarly, if the percentage is 58%, the corresponding bin is identified as bin #60. The peak count in bin #60 is then increased by one.

[0187] Return to reference Fig.16A , step 1614 determines whether there is another peak at the locus. If the answer is yes, the process repeats steps 1608, 1610, and 1612. If the answer is no, the process proceeds to step 1616 to generate histogram data using the identified histogram bins and the corresponding peak counts for all loci of the calculated sample profile. Based on the histogram data, a histogram can be generated. Fig. 7A Figure 1710 in Figure 1710 represents an example histogram of all loci corresponding to a calculated sample spectrum of a contributor. In Figure 1710, the horizontal axis represents the number of histogram bins, and the vertical axis represents the peak count or normalized peak count. The spectrum represented by Figure 1710 has a single contributor, and explains the shadow peak and allele peak. The shadow peak tends to have a lower peak height (e.g., less than 15%) than the corresponding allele peak. Therefore, Figure 1710 indicates that there are many peaks with small peak heights (e.g., in box #5 and box #10). In Figure 1710, the height of the bar represents the peak count with a specific relative peak height percentage (e.g., 5% of the maximum peak height). For example, in Figure 1710, about 60% of all peaks in the calculated sample spectrum are between 0% and 5% of the peak height. In Figure 1710, the peak height of most allele peaks is about 50%. Therefore, the spectrum corresponding to Figure 1710 has allele peaks (e.g., heterozygous peaks) with approximately the same peak height. Typically, a sample from a single person has a spectrum containing peaks with approximately the same peak height. In graph 1710, some peaks have a peak height percentage of about 100%, and they are homozygous peaks.

[0188] Fig. 17BFigure 1720 in represents another example histogram of all loci corresponding to another calculated sample spectrum of two contributors. In Figure 1720, the horizontal axis represents the number of histogram bins, and the vertical axis represents peak counts. The spectrum represented by Figure 1720 has two contributors, and shadow peaks and allele peaks are explained. Similar to Figure 1710, Figure 1720 indicates that there are many peaks with small peak heights (for example, in box #5, box #10 and box #15). In Figure 1720, the height of the bar represents the peak count with a specific relative peak height percentage (for example, 5% of the maximum peak height). Depending on the contribution rate between the two contributors, the allele peak may no longer have a peak height percentage of about 50%. In Figure 1720, compared with Figure 1710 (for example, from box #5 to box #10 and from box #40 to box #65), the value of the peak height percentage is distributed in more boxes (for example, from box #5 to box #20 and from box #30 to box #55). In some embodiments, for spectra with two or more contributors, fewer peaks are homozygous peaks, also shown in graph 1720 (e.g., there is no bar at the high peak height percentage range corresponding to bin #90 to bin #100 compared to graph 710).

[0189] As also shown in Figure 17, the peak height distribution of the peaks calculated for all loci in the sample profile varies with the number of contributors, and is therefore an influential feature on the NOC probability estimate. Fig.18 Other example features that have an impact on NOC probability estimates are illustrated in .

[0190] Fig.18 A method 1800 is illustrated for determining one or more other characteristics of each calculated sample spectrum. The method 1800 may be used to implement Fig.14 At least a portion of step 1418 of method 1400 shown. Step 1802 of method 1800 identifies peaks in the calculated sample spectrum. Step 1802 is similar to step 1602 of method 1600 and is therefore not described again. In one embodiment, the feature that can be extracted from the calculated sample spectrum is the total number of peaks. Therefore, step 1804 counts the total number of peaks identified in the calculated sample spectrum. As described above, the peak identification process distinguishes signal peaks from noise. Therefore, the identified peaks may include allele peaks and shadow peaks, but not noise. Using Fig. 16B The calculated sample spectrum 1640 is shown as an example, and if the spectrum 1640 only includes the three loci shown, the total peak count is nine. It should be understood that the spectrum 1640 is used to illustrate one example. The calculated sample spectrum can include more or fewer peaks, and can also have both allele peaks and shadow peaks.

[0191] In one embodiment, the feature that can be extracted from the computational sample spectrum is the total peak height. Based on the identified peaks, step 806 of method 800 obtains the peak height of each of the identified peaks in the computational sample spectrum. Using Fig. 16B As an example, if the calculated sample spectrum 1640 only includes the three loci shown, step 1806 can obtain the peak heights of all nine peaks. Next, step 1808 of method 1800 calculates the total peak height by summing the peak heights of all peaks in the calculated sample spectrum across all loci. For example, for spectrum 1640, by summing Fig. 16B The peak heights of all nine peaks shown are summed to calculate the total peak height. The total peak height of the spectrum is related to the total amount of DNA in the sample and is therefore an influential feature in the NOC probability estimate.

[0192] In one embodiment, a feature that can be extracted from the computational sample profile is the number of loci and / or markers having a predetermined number of peaks (e.g., "N" peaks, N≥2). Typically, if a sample has more contributors, there may be more peaks at some loci. Therefore, the number of loci and / or markers having a predetermined number has some impact on the NOC probability estimate. A locus is a specific fixed position on a chromosome where a specific gene or genetic marker is located. Based on the identified peaks, step 810 of method 800 counts the number of loci and / or markers having a predetermined number of peaks (e.g., "N" peaks, N≥2). For example, if Figure 6 The profile 640 shown in B has only the three loci shown, then the number of loci with three or more peaks is two (ie, locus 1 and locus 2).

[0193] In one embodiment, a feature that can be extracted from the calculated sample spectrum is the percentage of peaks with peak heights below a predetermined peak height threshold relative to the total number of peaks. As described above using FIG. 17 as an example, generally, if there are two or more contributors, there are a higher number of peaks with smaller peak heights. Therefore, the percentage of peaks with smaller peak heights has some impact on the NOC probability estimate. Fig.18 As shown, based on the identified peaks, step 1804 counts the total number of peaks in the calculated sample spectrum. Step 1812 identifies peaks in the calculated sample spectrum that have a peak height below a peak height threshold. The peak height threshold may be, for example, 150 RFU. Next, step 1814 calculates the percentage of peaks having a peak height below a predetermined peak height threshold relative to the total number of peaks. Continue Fig. 16BIn the example spectrum 1640 shown, there are a total of nine peaks in the spectrum, and two peaks (e.g., the peak of allele 34 at locus 2 and the peak of allele 8 at locus 3) have a peak height of about 100 RFU. These two peaks therefore have a peak height below a predetermined peak height threshold (e.g., 150 RFU). Therefore, the percentage of peaks with peak heights below the threshold is calculated to be about 22%.

[0194] It should be understood that Fig.16A , Fig. 16B , Fig.17A and Fig. 17B Some example features that can be extracted or calculated from the spectrum are shown. Other features can be determined in step 1820 of method 1800. For example, some features can be specific to a locus or a marker. These features can include, for example, the number of peaks at a specific locus or marker (e.g., marker SE33) and / or the size of the maximum peak of a specific locus or marker (e.g., marker SE33). In some embodiments, the features used in the NOC prediction analysis can be customized based on past analysis results, user input, default settings, and / or any other relevant factors.

[0195] FIG. 20A to FIG. 20B A computer interface 2000 is illustrated for visualizing the effects of different selections of the minimum relative fluorescence unit (RFU) for counting peaks according to some embodiments of the present disclosure. As shown, the interface 2000 allows a user to select a minimum RFU value for a peak in the user input area 2020.

[0196] exist Fig. 20A In the visual presentation of , 50RFU is selected as the minimum value for peaks to be counted in the calculated sample spectrum and evidence spectrum. Expected peak count distribution curves are shown for 1, 2, and 3 contributor scenarios. Specifically, curve 2001 shows an example of the expected peak count distribution for a calculated spectrum with one contributor. Curve 2002 shows an example of the expected peak count distribution for a calculated spectrum with two contributors. Curve 2003 shows the expected peak count distribution for a calculated spectrum with three contributors. Line 2010 shows the total peak count for the evidence spectrum.

[0197] Fig. 20B The visual representation of shows the expected peak count distribution and evidence spectrum if the minimum RFU value used to define the peaks to be counted is changed to 500RFU. Curves 2021, 2022 and 2023 respectively show the expected peak count distribution for one, two and three contributor spectra using 500RFU instead of 50RFU as the minimum value for counting peaks. Line 2015 shows the peak counts for the evidence spectrum.

[0198] like Fig. 20B As shown, the curve has Fig. 20A The corresponding curves 2001, 2002 and 2003 in FIG. 1 have slightly different shapes. Moreover, the distribution curves and lines of the evidence spectrum are relatively Fig. 20A The distribution curves and lines shown are slightly shifted to the left. Fig. 20B The visualization shows that Fig. 20A The same overall results, in the sense that it still suggests that the spectrum of evidence probably reflects 2 contributors but not 3. However, at the same time, there are some potentially meaningful differences. Fig. 20B Although the evidence spectrum Fig. 20B The line in 2015) is still closer to the double contributor curve ( Fig. 20B 2022) instead of the single contributor curve ( Fig. 20B curve in 2021), but it is already higher than in Fig. 20A is closer to the single contributor curve ( Fig. 20A 2010 and curve 2001 in FIG. 4 ), which indicates that when 500 RFU is set as the minimum value for counting peaks (relative to setting 50 as the minimum RFU for peaks), there is some increase in the probability that the evidence spectrum comes from one contributor.

[0199] An interactive computer interface such as interface 2000 may help a user who is a forensic expert visualize and demonstrate whether the selection of a minimum RFU value affects the overall results.

[0200] FIG. 21A to FIG. 21B A graphical user interface 2100 is shown for visualizing the relationship between the minimum RFU value and the maximum allele count ("MAC") in different analyzed samples. The MAC value is the maximum number of different alleles in any one of the STR loci being analyzed. In this example, the user can set a range of minimum RFU values ​​for the peak (peak height or "PH" in the illustration) to visualize the MAC and peak height. The user can also select the granularity (interval, i.e., "increment" between peak height values) for visualization. In these examples, data from minimum peak heights of 400 RFU to 4800 RFU are displayed at intervals of 200 RFU. In the illustrated examples, the user can change these settings by interacting with the user input area 2102.

[0201] Fig.21AA computerized visualization of data for a sample with two contributors and a contribution ratio of about 1:3 is shown. As shown, the data below 3600RFU is consistent with a mixture of two contributors because the MAC value is 4 or 3. For a single contributor, the maximum allele count at any one locus is expected to be 2. At the peak height cutoff of 3600RFU and above, the MAC value is 2. However, this may be due to the fact that if the minimum peak height counted as an allele is 3600RFU or higher, the minor contributor does not produce a high enough peak to count any allele. Therefore, the data as a whole is consistent with two contributors and unequal contribution ratios, and visualization helps experts confirm this.

[0202] Fig.21B The same visualization of a two contributor sample is shown, but the sample has a different mixture ratio of about 1 : 7. As shown, when the minimum RFU used to count peaks is 2400 or higher, the minor contributor's allele disappears from the data.

[0203] although Fig.21A and Fig.21B The visualizations shown may not necessarily be used alone to unambiguously determine the number of contributors to a mixture or to represent the ratios of these contributors in a sample, but these visualizations in a computerized graphical user interface can be useful for an expert to narrow down the range of plausible possibilities. Fig.21B The visualization of the sample data analyzed in itself does not explicitly show a mixture ratio of 1:7, but Fig.21A Compared to the samples visualized in , it is consistent with minor contributors representing a smaller portion of the sample relative to the major contributors. Assuming the data is above the range of the potential shadow peak, an earlier (lower RFU value) drop in MAC from one level to a lower level indicates that the ratio between minor and major contributors is smaller than a later (higher RFU value) drop. Therefore, this data can help experts identify which NOCs and contribution ratios are more likely than others, which in turn can be used to speed up deconvolution by prioritizing more likely scenarios over less likely scenarios for full deconvolution analysis. Additionally, in some examples, such visualizations can help experts identify anomalies in the data and / or confirm the results of other analyses, or help others better understand the data.

[0204] Fig. 22 is an example block diagram of a computing device 1400 that may incorporate embodiments of the present disclosure. Fig.14The machine systems that perform various aspects of the technical processes described herein are merely exemplified and do not limit the scope of the claims. Those skilled in the art will recognize other variations, modifications, and alternatives. In one embodiment, computing device 1400 typically includes a monitor or graphical user interface 1402, a data processing system 1420, a communication network interface 1412, an input device 1408, an output device 1406, and the like.

[0205] like Fig. 22 As shown, the data processing system 2220 may include one or more processors 2204 that communicate with a number of peripheral devices via a bus subsystem 2218. These peripheral devices may include input devices 2208, output devices 2206, a communication network interface 2212, and storage subsystems such as volatile memory 2210 and non-volatile memory 2214. The volatile memory 2210 and / or the non-volatile memory 2214 may store computer executable instructions and thus form logic 2222 that, when applied to and executed by the processor 2204, implements various embodiments of the processes disclosed herein.

[0206] Input devices 2208 include devices and mechanisms for inputting information into data processing system 2220. These may include keyboards, keypads, touch screens incorporated into monitor or graphical user interface 2202, audio input devices such as voice recognition systems, microphones, and other types of input devices. In various embodiments, input device 2208 may be embodied as a computer mouse, trackball, trackpad, joystick, wireless remote control, drawing tablet, voice command system, eye tracking system, etc. Input device 2208 typically allows a user to select objects, icons, control areas, text, etc. that appear on monitor or graphical user interface 2202 via commands such as clicking a button, etc.

[0207] Output devices 2206 include devices and mechanisms for outputting information from data processing system 2220. These may include a monitor or graphical user interface 2202, speakers, printers, infrared LEDs, etc. as are well known in the art.

[0208] The communication network interface 2212 provides an interface to a communication network (e.g., communication network 2216) and devices outside the data processing system 2220. The communication network interface 2212 may serve as an interface for receiving data from other systems and transmitting data to other systems. Embodiments of the communication network interface 2212 may include an Ethernet interface, a modem (telephone, satellite, cable, ISDN), (asynchronous) digital subscriber line (DSL), FireWire, USB, wireless communication interfaces such as Bluetooth or WiFi, near field communication wireless interfaces, cellular interfaces, etc. The communication network interface 2212 may be coupled to the communication network 2216 via an antenna, a cable, etc. In some embodiments, the communication network interface 2212 may be physically integrated on a circuit board of the data processing system 2220, or in some cases may be implemented in software or firmware such as a "soft modem", etc. The computing device 2200 may include logic capable of communicating on a network using protocols such as HTTP, TCP / IP, RTP / RTSP, IPX, UDP, etc.

[0209] Volatile memory 2210 and non-volatile memory 2214 are examples of tangible media configured to store computer-readable data and instructions, which form logic for implementing various aspects of the processes described herein. Other types of tangible media include removable memory (e.g., pluggable USB memory devices, mobile device SIM cards), optical storage media such as CD-ROMS, DVDs, semiconductor memories such as flash memory, non-transitory read-only memory (ROMS), battery-supported volatile memory, networked storage devices, etc. Volatile memory 2210 and non-volatile memory 2214 can be configured to store basic programming and data structures, which provide the functions of the disclosed processes and other embodiments thereof within the scope of the present disclosure. The logic 2222 implementing the embodiments of the present disclosure can be formed by volatile memory 2210 and / or non-volatile memory 2214 storing computer-readable instructions. The instructions can be read from volatile memory 2210 and / or non-volatile memory 2214 and executed by processor 2204. The volatile memory 2210 and the non-volatile memory 2214 may also provide a repository for storing data used by the logic 2222. The volatile memory 2210 and the non-volatile memory 2214 may include multiple memories, including a main random access memory (RAM) for storing instructions and data during program execution and a read-only memory (ROM) in which read-only non-transitory instructions are stored. The volatile memory 2210 and the non-volatile memory 2214 may include a file storage subsystem that provides permanent (non-volatile) storage for program and data files. The volatile memory 2210 and the non-volatile memory 2214 may include a removable storage system, such as a removable flash memory.

[0210] The bus subsystem 2218 provides a mechanism for allowing the various components and subsystems of the data processing system 2220 to communicate with each other as intended. Although the communication network interface 2212 is schematically depicted as a single bus, some embodiments of the bus subsystem 2218 may utilize multiple different busses.

[0211] It will be apparent to one of ordinary skill in the art that computing device 2200 may be a device such as a smart phone, a desktop computer, a laptop computer, a rack-mounted computer system, a computer server, or a tablet computer device. As is generally known in the art, computing device 2200 may be implemented as a series of multiple networked computing devices. In addition, computing device 2200 will typically include operating system logic (not shown), the type and nature of which are well known in the art.

[0212] One embodiment of the present disclosure includes systems, methods, and non-transitory computer-readable storage media or media tangibly storing computer program logic executable by a computer processor.

[0213] Those skilled in the art will appreciate that computer system 2200 illustrates only one example of a system in which a computer program product according to an embodiment of the present disclosure may be implemented. To cite only one example of an alternative embodiment, the execution of instructions contained in a computer program product according to an embodiment of the present disclosure may be distributed across multiple computers, such as, for example, computers of a distributed computer network.

[0214] Although the present disclosure has been specifically described with respect to the illustrated embodiments, it should be understood that various changes, modifications and adjustments may be made based on the present disclosure, and these changes, modifications and adjustments are intended to fall within the scope of the present disclosure. Although the present disclosure has been described in conjunction with what are currently considered to be practical embodiments, it should be understood that the present disclosure is not limited to the disclosed embodiments, but on the contrary, the present disclosure is intended to cover various modifications and equivalent arrangements included within the scope of the basic principles of the present disclosure as described in the various embodiments referenced above and below.

Claims

1. A computer-implemented method for selecting a subset of possible allele combinations for further deconvolution analysis using an evidence profile obtained from a biological sample to improve computer efficiency in genotyping one or more unidentified contributors from a plurality of contributors to the biological sample, the evidence profile comprising genetic signal data corresponding to a short tandem repeat (STR) allele at each locus of a plurality of loci, the method comprising: At each locus, for a currently analyzed contribution scenario among a plurality of contribution scenarios, performing processing using one or more computer processors, the processing comprising: (a) calculating an adjusted evidence spectrum by subtracting the calculated expected contributions of all known contributors, if any, from the evidence spectrum; (b) for a first or next unidentified contributor having a predetermined highest remaining contribution rate in the currently analyzed contribution rate scenario for the plurality of contributors, calculating a first range of expected peak heights using at least the predetermined highest remaining contribution rate, the selected degradation value, and the peak height ratio distribution; (c) calculating a second range of expected peak heights for all other remaining unidentified contributors, if any, using at least the predetermined contribution ratios in the currently analyzed contribution ratio scenario, the selected degradation values, and the peak height ratio distribution for all other remaining unidentified contributors; and (d) using the adjusted evidence spectrum and one or more of the first range and the second range to select one or more selected genotypes that at least potentially correspond to the first or next unidentified contributor for further deconvolution analysis, wherein the one or more selected genotypes include fewer genotypes than the total number of genotypes potentially associated with the current locus in the general population.

2. The computer-implemented method of claim 1 , wherein (d) comprises: If an allele in the adjusted evidence profile has a peak that is at least twice the threshold of the second range, the allele is selected twice as the desired allele in the allele pair, such that the one or more selected genotypes are a single homozygous genotype.

3. The computer-implemented method of any one of claims 1 and 2, wherein (d) comprises: If an allele in the adjusted evidence profile has a peak above a threshold of the second range and below twice the threshold of the second range, the allele is selected as the desired allele in the allele pair for each of the one or more selected genotypes.

4. The computer-implemented method of any one of claims 1 to 3, wherein (d) comprises: If an allele in the adjusted evidence profile has a peak that is at least twice the threshold of the first range, the allele is selected twice as two possible alleles for each of the one or more selected genotypes.

5. The computer-implemented method of any one of claims 1 to 4, wherein (d) comprises: If an allele in the adjusted evidence profile has a peak above a threshold of the first range but below twice the threshold of the first range, the allele is selected as a possible allele in an allele pair for each of the one or more selected genotypes.

6. The computer-implemented method of any one of claims 1 to 5, further comprising: Based on a predetermined number of contributors in the plurality of contributors and the genetic signal data, one or more contribution rate scenarios are determined that represent possible proportions of biological material contributed to the biological sample by each of the plurality of contributors.

7. The computer-implemented method of any one of claims 1 to 6, further comprising using one or more computer processors to perform a process comprising: For each corresponding genotype of the one or more selected genotypes that at least potentially corresponds to the first unidentified contributor, if a next unidentified contributor exists, repeat (a)-(d) for the next unidentified contributor, while treating the first unidentified contributor as a known contributor having a corresponding genotype of the one or more selected genotypes.

8. The computer-implemented method of claim 7, further comprising using the one or more computer processors to perform a process comprising: Before repeating (a)-(d) for the next unidentified contributor: A determination is made as to whether a corresponding genotype of the one or more selected genotypes includes a pair of desired alleles corresponding to the first unidentified contributor.

9. The computer-implemented method of any one of claims 7 and 8, further comprising using the one or more computer processors to perform a process comprising: Before repeating (a)-(d) for the next unidentified contributor: determining whether a corresponding genotype of the one or more selected genotypes includes only one desired allele corresponding to the first unidentified contributor and one or more possible alleles potentially corresponding to the first unidentified contributor; as well as If so, one or more allele combinations are calculated using the one desired allele and the one or more possible alleles, wherein for each of the one or more allele combinations, (a)-(d) are repeated for the next unidentified contributor.

10. The computer-implemented method of any one of claims 7 and 8, further comprising using one or more computer processors to perform a process comprising: Before repeating (a)-(d) for the next unidentified contributor: determining whether a corresponding genotype of the one or more selected genotypes includes only possible alleles that potentially correspond to the first unidentified contributor; as well as If so, then allele combinations are calculated using only the possible alleles, wherein for each of the allele combinations, the process according to claim 1 is repeated for the next unidentified contributor.

11. The computer-implemented method of any one of claims 7 to 10, further comprising using one or more computer processors to perform a process comprising: calculating multiple allele combinations of unidentified contributors by repeating (a)-(d) one or more times until no unidentified contributors remain; as well as For each allele combination in the plurality of allele combinations for each unidentified contributor in the unidentified contributors, a theoretical contribution to the corresponding theoretical spectrum is calculated.

12. The computer-implemented method of claim 11, wherein for each allele combination in the plurality of allele combinations for each of the unidentified contributors, calculating a theoretical contribution to a corresponding theoretical spectrum comprises: For the corresponding allele combinations of the corresponding unidentified contributors: calculating an allele peak corresponding to an allele in the corresponding allele combination, the allele peak being calculated using a contribution rate scenario of the current analysis for the corresponding unidentified contributor; storing the allele peaks for use in calculating a corresponding theoretical spectrum; calculating ghost peaks in at least some of the allele peaks, if any; and The shadow peaks are stored for use in calculating the corresponding theoretical spectrum.

13. The computer-implemented method of any one of claims 11 and 12, further comprising using one or more computer processors to perform a process comprising: calculating theoretical contributions of alleles corresponding to said all known contributors, said theoretical contributions having peak heights calculated using said contribution rate scenario of the current analysis; calculating the corresponding theoretical spectrum using the theoretical contributions from the unidentified contributors and the theoretical contributions from all known contributors; as well as A degree of match between the evidential spectrum and a corresponding theoretical spectrum is determined.

14. The computer-implemented method of claim 13, wherein determining a degree of match between the evidential spectrum and a corresponding theoretical spectrum comprises: For each corresponding bin associated with the evidence spectrum and the corresponding theoretical spectrum: Determine the number of alleles in the corresponding bin; determining whether the number of said alleles in said corresponding bin is greater than zero, and if so, determining one or more probability adjustment parameters; as well as One or more genotype probabilities are calculated based on the one or more probability adjustment parameters and the genotype probability model.

15. The computer-implemented method of claim 14, further comprising: determining whether there are missing peaks in the evidence spectrum, and if so, calculating a trip probability; as well as It is determined whether there are missing peaks in the theoretical spectrum, and if so, the probability of insertion is calculated.

16. The computer-implemented method of claim 15, further comprising: If there are no missing peaks in the theoretical spectrum and there are no missing peaks in the evidence spectrum, determining whether the peak heights in the corresponding bins of the theoretical spectrum and the evidence spectrum are greater than a predefined threshold; If yes, calculating the peak height mismatch probability between the peaks at the corresponding bins of the theoretical spectrum and the evidence spectrum; and A score for the corresponding bin is calculated using one or more of the one or more genotype probabilities, the dropout probability, the insertion probability, and the peak height mismatch probability.

17. The computer-implemented method of claim 16, further comprising: calculating a spectrum score using the scores of all bins associated with the theoretical spectrum and the evidence spectrum; using the spectrum score to determine whether the degree of match between the evidence spectrum and the corresponding theoretical spectrum satisfies a matching threshold; as well as If so, a possible match between the one or more unidentified contributors and one or more persons of interest (POIs) is provided.

18. The computer-implemented method of any one of claims 1 to 17, wherein (b) comprises: calculating the sum of the peak heights of the peaks in the adjusted evidence spectrum; calculating a first expected peak height associated with the first or next unidentified contributor using the sum of peak heights, the predetermined highest remaining contribution, and the selected degradation value; Using one or more expected shadow peak heights to adjust the first expected peak height; and The expected peak height first range is calculated using the adjusted first expected peak height and the peak height ratio distribution.

19. The computer-implemented method of any one of claims 1 to 18, wherein (c) comprises: Calculating the sum of the peak heights of the peaks in the evidence spectrum; calculating a second expected peak height associated with all other remaining unidentified contributors using the sum of the peak heights, the at least predetermined contribution rate in the currently analyzed contribution rate scenario of all other remaining unidentified contributors, and the selected degradation value; Using the one or more expected shadow peak heights to adjust one or more second expected peak heights; and The expected peak height second range is calculated using the adjusted one or more second expected peak heights and the peak height ratio distribution.

20. A non-transitory computer-readable medium storing one or more instructions that, when executed by one or more processors of at least one computing device, perform processing to select a subset of possible allele combinations for further deconvolution analysis using an evidence profile obtained from a biological sample to improve computer efficiency in genotyping one or more unidentified contributors from a plurality of contributors to the biological sample, the evidence profile comprising genetic signal data corresponding to a short tandem repeat (STR) allele at each locus of a plurality of loci, the processing comprising: At each locus, for the contribution scenario currently analyzed among multiple contribution scenarios: (a) calculating an adjusted evidence spectrum by subtracting the calculated expected contributions of all known contributors, if any, from the evidence spectrum; (b) for a first or next unidentified contributor having a predetermined highest remaining contribution rate in the currently analyzed contribution rate scenario for the plurality of contributors, calculating a first range of expected peak heights using at least the predetermined highest remaining contribution rate, the selected degradation value, and the peak height ratio distribution; (c) calculating a second range of expected peak heights for all other remaining unidentified contributors, if any, using at least the predetermined contribution ratios in the currently analyzed contribution ratio scenario, the selected degradation values, and the peak height ratio distribution for all other remaining unidentified contributors; and (d) using the adjusted evidence spectrum and one or more of the first range and the second range to select one or more selected genotypes that at least potentially correspond to the first or next unidentified contributor for further deconvolution analysis, wherein the one or more selected genotypes include fewer genotypes than the total number of genotypes potentially associated with the current locus in the general population.

21. A system for selecting a subset of possible allele combinations for further deconvolution analysis using an evidence profile obtained from a biological sample to improve computer efficiency in genotyping one or more unidentified contributors from a plurality of contributors to the biological sample, the evidence profile comprising genetic signal data corresponding to a short tandem repeat (STR) allele at each locus of a plurality of loci, the system comprising: one or more processors of at least one computing device; as well as a memory storing one or more instructions that, when executed by the one or more processors, cause the one or more processors to perform processing at each locus for a currently analyzed contribution scenario among a plurality of contribution scenarios, the processing comprising: (a) calculating an adjusted evidence spectrum by subtracting the calculated expected contributions of all known contributors, if any, from the evidence spectrum; (b) for a first or next unidentified contributor having a predetermined highest remaining contribution rate in the currently analyzed contribution rate scenario for the plurality of contributors, calculating a first range of expected peak heights using at least the predetermined highest remaining contribution rate, the selected degradation value, and the peak height ratio distribution; (c) calculating a second range of expected peak heights for all other remaining unidentified contributors, if any, using at least the predetermined contribution ratios in the currently analyzed contribution ratio scenario, the selected degradation values, and the peak height ratio distribution for all other remaining unidentified contributors; and (d) using the adjusted evidence spectrum and one or more of the first range and the second range to select one or more selected genotypes that at least potentially correspond to the first or next unidentified contributor for further deconvolution analysis, wherein the one or more selected genotypes include fewer genotypes than the total number of genotypes potentially associated with the current locus in the general population.

22. A computer-implemented method for predicting a number of contributors associated with a biological sample using an evidence profile obtained from the biological sample to improve computer efficiency in genotyping at least one contributor, the evidence profile comprising genetic signal data, the method comprising processing using one or more computer processors, the processing comprising: receiving an indication of a combination of selected models for estimating the number of contributors associated with the biological sample, the selected models being different models; generating a predetermined number of calculated sample spectra based on the evidence spectrum; for each selected model in the combination of selected models, estimating probabilities of a plurality of number of contributors (NOC) likelihoods using the selected model and the calculated sample spectrum; as well as Based on the weights assigned to the selected models, the estimated probabilities of the plurality of NOC likelihoods obtained based on a combination of the selected models are used to predict the number of contributors associated with the biological sample.

23. The computer-implemented method of claim 22, wherein the combination of selected models includes two or more of an artificial neural network-based model, a decision tree-based model, a random forest algorithm-based model, and a peak count distribution model.

24. The computer-implemented method of any one of claims 22 and 23, further comprising obtaining contextual information based on the evidence spectrum, wherein generating the predetermined number of calculated sample spectra based on the evidence spectrum comprises: Receive the maximum number of expected contributors; for each NOC possibility less than or equal to the expected maximum number of contributors, receiving a requested number of calculated sample spectra; as well as The predetermined number of calculated sample spectra are generated based on the requested number of calculated sample spectra for each of the plurality of NOC possibilities, the calculated sample spectra and the evidence spectra having at least some commonality with respect to the contextual information.

25. The computer-implemented method of claim 24, wherein the context information comprises one or more of: parameters associated with a test kit and a testing machine used to obtain said evidence profile; the total peak height associated with the peaks in the evidence spectrum; and Population statistics associated with the evidence spectrum.

26. The computer-implemented method of any one of claims 22 to 25, wherein for each selected model in the combination of selected models, estimating the probability of a plurality of NOC possibilities using the selected model and the calculated sample spectrum comprises: determining whether the selected model is a machine learning based model; Based on determining that the selected model is a machine learning based model, performing the following process: Determining one or more features of the calculated sample spectrum for use in training the machine learning-based model or updating the training of the machine learning-based model; Using the calculated sample profile to train the machine learning-based model or update training of the machine learning-based model based on the one or more features to provide a trained model; as well as The probabilities of the plurality of NOC possibilities are estimated using the trained model.

27. The computer-implemented method of claim 26, wherein determining one or more features of the computational sample spectrum for use in training the machine learning-based model or updating training of the machine learning-based model comprises: For each locus in the calculated sample profile: determining a maximum peak height at the locus; For each peak of the calculated sample spectrum, Calculate the percentage of peak height relative to the maximum peak height, determining corresponding histogram bins based on the percentage of peak height relative to the maximum peak height, and updating the number of peak counts for the corresponding histogram bin; and For all loci of the calculated sample profile, histogram data is generated using the histogram bins and corresponding peak counts.

28. The computer-implemented method of claim 26, wherein determining one or more features of the computational sample spectrum for use in training the machine learning-based model or updating training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: identifying peaks in the calculated sample spectrum; as well as The total number of identified peaks in the calculated sample spectrum is determined.

29. The computer-implemented method of claim 26, wherein determining one or more features of the computational sample spectrum for use in training the machine learning-based model or updating training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: identifying peaks in the calculated sample spectrum; obtaining peak heights of peaks identified in the calculated sample spectrum; and The peak heights of the identified peaks are used to calculate the total peak height.

30. The computer-implemented method of claim 26, wherein determining one or more features of the computational sample spectrum for use in training the machine learning-based model or updating training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: identifying peaks in the calculated sample spectrum; as well as A number of loci having a predetermined number of peaks in the calculated sample profile is determined based on the identified peaks, wherein the predetermined number of peaks is greater than or equal to two.

31. The computer-implemented method of claim 26, wherein determining one or more features of the computational sample spectrum for use in training the machine learning-based model or updating training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: determining the total number of peaks in the calculated sample spectrum; identifying peaks in the calculated sample spectrum having a peak height below a peak height threshold; and The percentage of the peaks having a peak height below a predetermined peak height threshold relative to the total number of the peaks is calculated.

32. The computer-implemented method of any one of claims 22 to 31, further comprising: An expected maximum number of contributors is received, wherein generating the predetermined number of calculated sample spectra is further based on the expected maximum number of contributors.

33. The computer-implemented method of any one of claims 22 to 32, wherein for each user-selected model in the combination of selected models, estimating the probability of a plurality of NOC possibilities using the selected model and the calculated sample spectrum comprises: determining whether the selected model is a peak count distribution model; Based on determining that the selected model is a peak count distribution model, performing the following process: for each possible number of contributors that is less than or equal to the expected maximum number of contributors, determining the number of peaks for each calculated sample spectrum having the corresponding possible number of contributors; determining an expected peak count distribution based on said number of peaks of said calculated sample spectrum; determining the total number of peaks in the evidence spectrum; and The probabilities of the plurality of NOC possibilities are estimated based on the expected peak count distribution and the total number of peaks in the evidence spectrum.

34. The computer-implemented method of any one of claims 22 to 33, wherein predicting the number of contributors associated with the biological sample using the estimated probabilities of the plurality of NOC likelihoods obtained based on the combination of the selected models comprises: For each selected model in the combination of selected models, calculating a product of the weight assigned to the selected model and the estimated probability obtained based on the selected model to obtain a scaled probability; for each NOC possibility of the plurality of NOC possibilities, calculating a weighted sum using the scaled probabilities of the combination of the selected models; as well as The weighted sum of the plurality of selected models is used to predict the number of contributors associated with the biological sample.

35. A computer-implemented method for improving computer efficiency of genotyping at least one contributor using an evidence profile obtained from a biological sample to predict a number of contributors associated with the biological sample, the evidence profile comprising genetic signal data, the method comprising processing using one or more computer processors, the processing comprising: Receive the maximum number of expected contributors; generating a predetermined number of calculated sample spectra based on the evidence spectrum and the expected maximum number of contributors; for each possible number of contributors that is less than or equal to the expected maximum number of contributors, determining the number of peaks for each calculated sample spectrum having the corresponding possible number of contributors; obtaining an expected peak count distribution based on the number of peaks for each calculated sample spectrum with a corresponding number of possible contributors; determining the total number of peaks in the evidence spectrum; and The probabilities of the plurality of NOC possibilities are estimated based on the expected peak count distribution and the total number of peaks in the evidence spectrum.

36. The computer-implemented method of claim 35, further comprising obtaining contextual information based on the evidence spectrum, wherein generating the predetermined number of computational sample spectra based on the evidence spectrum and the expected maximum number of contributors comprises: for each NOC possibility less than or equal to the expected maximum number of contributors, receiving a requested number of calculated sample spectra; as well as The predetermined number of calculated sample spectra are generated based on the requested number of calculated sample spectra for each of the plurality of NOC possibilities, the calculated sample spectra and the evidence spectra having at least some commonality with respect to the contextual information.

37. The computer-implemented method of claim 36, wherein the context information comprises one or more of: parameters associated with a test kit and a testing machine used to obtain said evidence profile; the total peak height associated with the peaks in the evidence spectrum; and Population statistics associated with the evidence spectrum.

38. A non-transitory computer readable medium storing one or more instructions that, when executed by one or more processors of at least one computing device, perform processing to predict a number of contributors associated with a biological sample using an evidence profile obtained from the biological sample to improve computer efficiency in genotyping at least one of the contributors, the evidence profile comprising genetic signal data, the processing comprising: receiving an indication of a combination of selected models for estimating the number of contributors associated with the biological sample, the selected models being different models; generating a predetermined number of calculated sample spectra based on the evidence spectrum; for each selected model in the combination of selected models, estimating probabilities of a plurality of number of contributors (NOC) likelihoods using the selected model and the calculated sample spectrum; as well as Based on the weights assigned to the selected models, the estimated probabilities of the plurality of NOC likelihoods obtained based on a combination of the selected models are used to predict the number of contributors associated with the biological sample.

39. The computer-readable medium of claim 38, wherein the combination of selected models comprises two or more of an artificial neural network-based model, a decision tree-based model, a random forest algorithm-based model, and a peak count distribution model.

40. The computer-readable medium of any one of claims 38 and 39, wherein the processing further comprises obtaining contextual information based on the evidence spectrum, wherein generating the predetermined number of calculated sample spectra based on the evidence spectrum comprises: Receive the maximum number of expected contributors; for each NOC possibility less than or equal to the expected maximum number of contributors, receiving a requested number of calculated sample spectra; as well as The predetermined number of calculated sample spectra are generated based on the requested number of calculated sample spectra for each of the plurality of NOC possibilities, the calculated sample spectra and the evidence spectra having at least some commonality with respect to the contextual information.

41. The computer-readable medium of claim 40, wherein the context information comprises one or more of: parameters associated with a test kit and a testing machine used to obtain said evidence profile; the total peak height associated with the peaks in the evidence spectrum; and Population statistics associated with the evidence spectrum.

42. The computer readable medium of any one of claims 38 to 41, wherein for each selected model in the combination of selected models, estimating the probability of a plurality of NOC possibilities using the selected model and the calculated sample spectrum comprises: determining whether the selected model is a machine learning based model; Based on determining that the selected model is a machine learning based model, performing the following process: Determining one or more features of the calculated sample spectrum for use in training the machine learning-based model or updating the training of the machine learning-based model; Using the calculated sample profile to train the machine learning-based model or update training of the machine learning-based model based on the one or more features to provide a trained model; as well as The probabilities of the plurality of NOC possibilities are estimated using the trained model.

43. The computer-readable medium of claim 42, wherein determining one or more features of the computed sample spectrum for use in training the machine learning-based model or updating training of the machine learning-based model comprises: For each locus in the calculated sample profile: determining a maximum peak height at the locus; For each peak of the calculated sample spectrum, Calculate the percentage of peak height relative to the maximum peak height, determining corresponding histogram bins based on the percentage of peak height relative to the maximum peak height, and updating the number of peak counts for the corresponding histogram bin; and For all loci of the calculated sample profile, histogram data is generated using the histogram bins and corresponding peak counts.

44. The computer-readable medium of claim 42, wherein determining one or more features of the computed sample spectrum for use in training the machine learning-based model or updating training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: identifying peaks in the calculated sample spectrum; as well as The total number of identified peaks in the calculated sample spectrum is determined.

45. The computer-readable medium of claim 42, wherein determining one or more features of the computed sample profile for use in training the machine learning-based model or updating training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: identifying peaks in the calculated sample spectrum; obtaining peak heights of peaks identified in the calculated sample spectrum; and The peak heights of the identified peaks are used to calculate the total peak height.

46. ​​The computer-readable medium of claim 42, wherein determining one or more features of the computed sample profile for use in training the machine learning-based model or updating training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: identifying peaks in the calculated sample spectrum; as well as A number of loci having a predetermined number of peaks in the calculated sample profile is determined based on the identified peaks, wherein the predetermined number of peaks is greater than or equal to two.

47. The computer-readable medium of claim 42, wherein determining one or more features of the computed sample spectrum for use in training the machine learning-based model or updating training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: determining the total number of peaks in the calculated sample spectrum; identifying peaks in the calculated sample spectrum having a peak height below a peak height threshold; and The percentage of the peaks having a peak height below a predetermined peak height threshold relative to the total number of the peaks is calculated.

48. The computer readable medium of any one of claims 38 to 47, wherein the processing further comprises: An expected maximum number of contributors is received, wherein generating the predetermined number of calculated sample spectra is further based on the expected maximum number of contributors.

49. The computer-readable medium of any one of claims 38 to 48, wherein for each user-selected model in the combination of selected models, estimating probabilities of a plurality of NOC possibilities using the selected model and the calculated sample spectrum comprises: determining whether the selected model is a peak count distribution model; Based on determining that the selected model is a peak count distribution model, performing the following process: for each possible number of contributors that is less than or equal to the expected maximum number of contributors, determining the number of peaks for each calculated sample spectrum having the corresponding possible number of contributors; determining an expected peak count distribution based on said number of peaks of said calculated sample spectrum; determining the total number of peaks in the evidence spectrum; and The probabilities of the plurality of NOC possibilities are estimated based on the expected peak count distribution and the total number of peaks in the evidence spectrum.

50. The computer-readable medium of any one of claims 38 to 49, wherein predicting the number of contributors associated with the biological sample using the estimated probabilities of the plurality of NOC likelihoods obtained based on the combination of the selected models comprises: For each selected model in the combination of selected models, calculating a product of the weight assigned to the selected model and the estimated probability obtained based on the selected model to obtain a scaled probability; for each NOC possibility of the plurality of NOC possibilities, calculating a weighted sum using the scaled probabilities of the combination of the selected models; as well as The weighted sum of the plurality of selected models is used to predict the number of contributors associated with the biological sample.

51. A system for predicting a number of contributors associated with a biological sample using an evidence profile obtained from the biological sample to improve computer efficiency in genotyping at least one of the contributors, the evidence profile comprising genetic signal data, the system comprising: one or more processors of at least one computing device; as well as A memory storing one or more instructions that, when executed by the one or more processors, cause the one or more processors to perform the following process: receiving an indication of a combination of selected models for estimating the number of contributors associated with the biological sample, the selected models being different models; generating a predetermined number of calculated sample spectra based on the evidence spectrum; for each selected model in the combination of selected models, estimating probabilities of a plurality of number of contributors (NOC) likelihoods using the selected model and the calculated sample spectrum; as well as Based on the weights assigned to the selected models, the estimated probabilities of the plurality of NOC likelihoods obtained based on a combination of the selected models are used to predict the number of contributors associated with the biological sample.

52. The system of claim 51, wherein the combination of selected models comprises two or more of an artificial neural network-based model, a decision tree-based model, a random forest algorithm-based model, and a peak count distribution model.

53. The system of any one of claims 51 and 52, wherein the process further comprises obtaining contextual information based on the evidence spectrum, wherein generating the predetermined number of calculated sample spectra based on the evidence spectrum comprises: Receive the maximum number of expected contributors; for each NOC possibility less than or equal to the expected maximum number of contributors, receiving a requested number of calculated sample spectra; as well as The predetermined number of calculated sample spectra are generated based on the requested number of calculated sample spectra for each of the plurality of NOC possibilities, the calculated sample spectra and the evidence spectra having at least some commonality with respect to the contextual information.

54. The system of claim 53, wherein the context information comprises one or more of: parameters associated with a test kit and a testing machine used to obtain said evidence profile; the total peak height associated with the peaks in the evidence spectrum; and Population statistics associated with the evidence spectrum.

55. The system of any one of claims 51 to 54, wherein for each selected model in the combination of selected models, estimating the probability of a plurality of NOC possibilities using the selected model and the calculated sample spectrum comprises: determining whether the selected model is a machine learning based model; Based on determining that the selected model is a machine learning based model, performing the following process: Determining one or more features of the calculated sample spectrum for use in training the machine learning-based model or updating the training of the machine learning-based model; Using the calculated sample profile to train the machine learning-based model or update training of the machine learning-based model based on the one or more features to provide a trained model; as well as The probabilities of the plurality of NOC possibilities are estimated using the trained model.

56. The system of claim 55, wherein determining one or more features of the computed sample profile for use in training the machine learning-based model or updating the training of the machine learning-based model comprises: For each locus in the calculated sample profile: determining a maximum peak height at the locus; For each peak of the calculated sample spectrum, Calculate the percentage of peak height relative to the maximum peak height, determining corresponding histogram bins based on the percentage of peak height relative to the maximum peak height, and updating the number of peak counts for the corresponding histogram bin; and For all loci of the calculated sample profile, histogram data is generated using the histogram bins and corresponding peak counts.

57. The system of claim 56, wherein determining one or more features of the computed sample profile for use in training the machine learning-based model or updating the training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: identifying peaks in the calculated sample spectrum; as well as The total number of identified peaks in the calculated sample spectrum is determined.

58. The system of claim 56, wherein determining one or more features of the computed sample profile for use in training the machine learning-based model or updating the training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: identifying peaks in the calculated sample spectrum; obtaining peak heights of peaks identified in the calculated sample spectrum; and The peak heights of the identified peaks are used to calculate the total peak height.

59. The system of claim 56, wherein determining one or more features of the computed sample profile for use in training the machine learning-based model or updating the training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: identifying peaks in the calculated sample spectrum; as well as A number of loci having a predetermined number of peaks in the calculated sample profile is determined based on the identified peaks, wherein the predetermined number of peaks is greater than or equal to two.

60. The system of claim 56, wherein determining one or more features of the computed sample profile for use in training the machine learning-based model or updating the training of the machine learning-based model comprises: For each calculated sample spectrum of the calculated sample spectrum: determining the total number of peaks in the calculated sample spectrum; identifying peaks in the calculated sample spectrum having a peak height below a peak height threshold; and The percentage of the peaks having a peak height below a predetermined peak height threshold relative to the total number of the peaks is calculated.

61. The system of any one of claims 51 to 60, wherein the process further comprises: An expected maximum number of contributors is received, wherein generating the predetermined number of calculated sample spectra is further based on the expected maximum number of contributors.

62. The system of any one of claims 51 to 61, wherein for each user-selected model in the combination of selected models, estimating the probability of a plurality of NOC possibilities using the selected model and the calculated sample spectrum comprises: determining whether the selected model is a peak count distribution model; Based on determining that the selected model is a peak count distribution model, performing the following process: for each possible number of contributors that is less than or equal to the expected maximum number of contributors, determining the number of peaks for each calculated sample spectrum having the corresponding possible number of contributors; determining an expected peak count distribution based on said number of peaks of said calculated sample spectrum; determining the total number of peaks in the evidence spectrum; and The probabilities of the plurality of NOC possibilities are estimated based on the expected peak count distribution and the total number of peaks in the evidence spectrum.

63. The system of any one of claims 51 to 62, wherein predicting the number of contributors associated with the biological sample using the estimated probabilities of the plurality of NOC likelihoods obtained based on the combination of the selected models comprises: For each selected model in the combination of selected models, calculating a product of the weight assigned to the selected model and the estimated probability obtained based on the selected model to obtain a scaled probability; for each NOC possibility of the plurality of NOC possibilities, calculating a weighted sum using the scaled probabilities of the combination of the selected models; as well as The weighted sum of the plurality of selected models is used to predict the number of contributors associated with the biological sample.

64. A non-transitory computer readable medium storing one or more instructions that, when executed by one or more processors of at least one computing device, perform processing to predict a number of contributors associated with a biological sample using an evidence profile obtained from the biological sample to improve computer efficiency in genotyping at least one of the contributors, the evidence profile comprising genetic signal data, the processing comprising: Receive the maximum number of expected contributors; generating a predetermined number of calculated sample spectra based on the evidence spectrum and the expected maximum number of contributors; for each possible number of contributors that is less than or equal to the expected maximum number of contributors, determining the number of peaks for each calculated sample spectrum having the corresponding possible number of contributors; obtaining an expected peak count distribution based on the number of peaks for each calculated sample spectrum with a corresponding number of possible contributors; determining the total number of peaks in the evidence spectrum; and The probabilities of the plurality of NOC possibilities are estimated based on the expected peak count distribution and the total number of peaks in the evidence spectrum.

65. The computer-readable medium of claim 64, wherein the processing further comprises obtaining contextual information based on the evidence spectrum, wherein generating the predetermined number of calculated sample spectra based on the evidence spectrum and the expected maximum number of contributors comprises: for each NOC possibility less than or equal to the expected maximum number of contributors, receiving a requested number of calculated sample spectra; as well as The predetermined number of calculated sample spectra are generated based on the requested number of calculated sample spectra for each of the plurality of NOC possibilities, the calculated sample spectra and the evidence spectra having at least some commonality with respect to the contextual information.

66. The computer-readable medium of claim 65, wherein the context information comprises one or more of: parameters associated with a test kit and a testing machine used to obtain said evidence profile; the total peak height associated with the peaks in the evidence spectrum; and Population statistics associated with the evidence spectrum.

67. A system for predicting a number of contributors associated with a biological sample using an evidence profile obtained from the biological sample to improve computer efficiency in genotyping at least one of the contributors, the evidence profile comprising genetic signal data, the system comprising: one or more processors of at least one computing device; as well as A memory storing one or more instructions that, when executed by the one or more processors, cause the one or more processors to perform the following process: Receive the maximum number of expected contributors; generating a predetermined number of calculated sample spectra based on the evidence spectrum and the expected maximum number of contributors; for each possible number of contributors that is less than or equal to the expected maximum number of contributors, determining the number of peaks for each calculated sample spectrum having the corresponding possible number of contributors; obtaining an expected peak count distribution based on the number of peaks for each calculated sample spectrum with a corresponding number of possible contributors; determining the total number of peaks in the evidence spectrum; and The probabilities of the plurality of NOC possibilities are estimated based on the expected peak count distribution and the total number of peaks in the evidence spectrum.

68. The system of claim 67, wherein the processing further comprises obtaining contextual information based on the evidence spectrum, wherein generating the predetermined number of calculated sample spectra based on the evidence spectrum and the expected maximum number of contributors comprises: for each NOC possibility less than or equal to the expected maximum number of contributors, receiving a requested number of calculated sample spectra; as well as The predetermined number of calculated sample spectra are generated based on the requested number of calculated sample spectra for each of the plurality of NOC possibilities, the calculated sample spectra and the evidence spectra having at least some commonality with respect to the contextual information.

69. The system of claim 68, wherein the context information comprises one or more of: parameters associated with a test kit and a testing machine used to obtain said evidence profile; the total peak height associated with the peaks in the evidence spectrum; and Population statistics associated with the evidence spectrum.

70. A computer-implemented method for providing a graphical user interface (GUI) on an electronic display, the computer-implemented method comprising: receiving, via the GUI, a minimum peak height value of fluorescence data input by a user, the fluorescence data corresponding to alleles at a plurality of loci in a biological sample analyzed via the GUI; calculating a plurality of distribution visual representations using the minimum peak height value and displaying the plurality of distribution visual representations in the GUI on the electronic display, each of the plurality of distribution visual representations showing a distribution of a calculated sample spectrum of a hypothesized number of contributors for a range of peak counts; as well as An evidence spectrum visual representation is calculated using the minimum peak height values ​​and displayed in the GUI on the electronic display, the evidence spectrum visual representation showing peak counts of the evidence spectrum, wherein the evidence spectrum visual representation is displayed relative to the plurality of distribution visual representations.

71. The computer-implemented method of claim 70, further comprising: Two or more visual representations are displayed in the GUI on the electronic display, each visual representation comprising a plurality of distribution visual representations and an evidence spectrum visual representation, wherein each of the two or more visual representations is calculated and displayed using a different minimum peak height value.

72. A computer-implemented method according to claim 71, wherein the two or more visual representations are displayed relative to each other in a manner that facilitates comparison of a first visual presentation corresponding to a first minimum peak value and a second visual presentation corresponding to a second minimum peak value.

73. The computer-implemented method of claim 72, wherein the two or more visual representations are horizontally aligned and displayed one above the other.

74. A computer-implemented method for providing a graphical user interface (GUI) on an electronic display, the computer-implemented method comprising: receiving display parameters including at least a minimum peak height value range; as well as A visual representation of the maximum allele count value of the analyzed sample is calculated for each minimum peak height value in a plurality of minimum peak height values ​​within the range of minimum peak height values ​​using the display parameters and displayed in the GUI on the electronic display.

75. The computer implemented method of claim 74, wherein the display parameters are inputtable by a user via the GUI.

Images