Using relatives' information to determine genetic risk for non-mendelian phenotypes

Abstract

To provide methods, systems and non-transitory machine-readable media for using relatives' genetic information to determine a genetic risk for non-mendelian phenotypes.SOLUTION: Methods for outputting a non-Mendelian risk score comprises: receiving, from a first dataset, genotype data for a subject and genotype data and phenotype data for one or more blood relatives of a subject having a gene of interest; receiving, from a second dataset, genotype population data and phenotype population data, where the population comprises two or more blood relatives; training a model on the first and second datasets to determine a genetic risk in the subject associated with one or more non-Mendelian gene of interest; and outputting a phenotypic risk score for the subject.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. Provisional Patent Application No. 62 / 820,286, filed March 19, 2019, which is incorporated by reference in its entirety herein.

[0002] Methods are described for determining genetic risk for non-Mendelian phenotypes using genetic information from relatives. [Background technology]

[0003] For Mendelian genes, the probability of developing a phenotype is roughly 0 or 1, depending on whether the subject inherits 0, 1 or 2 versions of the mutant gene and whether the gene exhibits dominant or recessive inheritance. For Mendelian phenotypes, the risk to a subject is established by analyzing the family tree and the medical history of the subject's relatives in a well-defined manner. For non-Mendelian genes, the probability that a subject with a specific gene mutation will develop a phenotype is not absolutely 0 or 1. Furthermore, non-Mendelian phenotypes are generally influenced by multiple genes. The influence of multiple genes is generally captured in polygenic risk models, which tend to be inaccurate and use population-level data to calibrate the influence of each gene. There is a need in the art for more accurate methods for determining whether a subject is at risk for a non-Mendelian phenotype, especially methods that can incorporate family medical history. Summary of the Invention

[0004] Methods are provided for outputting a non-Mendelian phenotypic risk score for each subject that is more accurate by using the disease or phenotypic status of the subject's relatives. Some embodiments include receiving from a first dataset (i) genotypic data for the subject with one or more non-Mendelian genes of interest and (ii) genotypic and phenotypic data for one or more relatives of the subject with one or more non-Mendelian genes of interest. Some embodiments include receiving genotypic population data and phenotypic population data from a second dataset, the populations including one or more sets of two or more relatives. Some embodiments include training a model on the first and second datasets to determine risk in the subject associated with one or more of the non-Mendelian genes of interest. Some embodiments include outputting a phenotypic risk score for the subject.

[0005] In some embodiments, the second dataset includes genotypic population data and phenotypic population data for multiple sets of two or more relatives.

[0006] In some embodiments, the blood relatives in the first dataset include one or more of the subject's mother, father, brother, sister, son, daughter, grandfather, grandmother, aunt, uncle, niece, nephew, and first cousin. In some embodiments, the second dataset includes two or more subjects who have the same blood relationship as the subject in the first dataset.

[0007] In some embodiments, one or more of the relatives is a male relative. In some embodiments, one or more of the relatives is a female relative.

[0008] In some embodiments, the first dataset includes data for a plurality of relatives of the subject, hi some embodiments, one or more of the relatives is a male relative and one or more of the relatives is a female relative.

[0009] In some embodiments, the gene of interest is a gene variant of interest.

[0010] In some embodiments, the first dataset and the second dataset comprise data associated with phenotype age of onset.

[0011] Also provided is a system that includes a processor; and a memory coupled to the processor for storing instructions that, when executed by the processor, cause the processor to perform operations including receiving, from a first dataset, (i) genotype data for a subject having one or more non-Mendelian genes of interest and (ii) genotype data and phenotype data for one or more relatives of the subject having one or more of the genes of interest; receiving, from a second dataset, genotype population data and phenotype population data, the population including one or more sets of two or more relatives; training a model on the first and second datasets to determine a risk in the subject associated with one or more of the non-Mendelian genes of interest, and outputting a phenotypic risk score for the subject.

[0012] Also provided is a non-transitory machine-readable medium having instructions stored thereon that, when executed by a processor, causes the processor to perform operations including receiving, from a first dataset, (i) genotype data for a subject having one or more non-Mendelian genes of interest and (ii) genotype data and phenotype data for one or more relatives of the subject having one or more of the genes of interest; receiving, from a second dataset, genotype population data and phenotype population data, the population including one or more sets of two or more relatives; training a model on the first and second datasets to determine a risk in the subject associated with one or more of the non-Mendelian genes of interest, and outputting a phenotypic risk score for the subject.

[0013] In some aspects related to the system or non-transitory machine-readable medium, the second dataset includes genotypic population data and phenotypic population data for two or more blood relatives. In some aspects, the blood relatives in the first dataset include one or more of the subject's mother, father, brother, sister, son, daughter, grandfather, grandmother, aunt, uncle, niece, nephew, and first cousin. In some aspects, the second dataset includes two or more subjects with the same blood relationship as the subject in the first dataset. In some aspects, one or more of the blood relatives is a male relative. In some aspects, one or more of the blood relatives is a female relative.

[0014] In some aspects related to the system or non-transitory machine-readable medium, the first dataset includes data for a plurality of relatives of the subject, hi some aspects, one or more of the relatives are male relatives and one or more of the relatives are female relatives.

[0015] In some aspects related to the system or non-transitory machine-readable medium, the gene of interest is a gene variant of interest.

[0016] In some aspects related to the system or non-transitory machine-readable medium, the first dataset and the second dataset include data associated with a phenotype age-of-onset.

[0017] Also provided are methods for outputting a polygenic risk score, comprising receiving from a first dataset (i) genotypic data for a subject having one or more non-Mendelian genes of interest and (ii) genotypic and phenotypic data for one or more relatives of the subject having one or more of the non-Mendelian genes of interest; receiving from a second dataset genotypic population data and phenotypic population data, the population including one or more sets of two or more relatives; training a model on the first and second datasets to determine a risk in the subject based on the one or more non-Mendelian genes of interest; and outputting a phenotypic risk score for the subject. Some embodiments include training a model on the first and second datasets to predict how risk in the subject is modified by the one or more non-Mendelian genes of interest compared to the risk in the subject given the phenotypic data of the relatives.

[0018] Methods of treating a subject based on the phenotypic risk score are also provided. [Brief description of the drawings]

[0019] [Figure 1] FIG. 1 shows a simulated histogram of expressed phenotypes with a mean age of onset of 60 years. [Diagram 2] FIG. 2 is an example block diagram of a computing device. [Diagram 3] FIG. 3 shows simulation results illustrating an embodiment of the method applied to three genes, where the third gene has a population frequency of 1.0%; FIGS. 3A and 3B show histograms of predictions for subjects where only a subset of the relevant genes are available in the model; FIG. 3C shows a histogram of predictions for subjects where all genetic variables are included. [Figure 4]FIG. 4 shows simulation results illustrating an embodiment of the method applied to three genes, where the third gene has a population frequency of 0.2%; FIGS. 4A and 4B show histograms of predictions for subjects where only a subset of the relevant genes are available in the model; FIG. 4C shows a histogram of predictions for subjects where all genetic variables are included. [Diagram 5] FIG. 5 shows simulation results illustrating an embodiment of the method applied to three genes, where the third gene has a population frequency of 0.05%; FIGS. 5A and 5B show histograms of predictions for subjects where only a subset of the relevant genes are available in the model; FIG. 5C shows a histogram of predictions for subjects where all genetic variables are included. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] Unless otherwise defined, technical and scientific terms used herein have the meaning commonly understood by one of ordinary skill in the art to which this invention belongs. Materials referred to in the following description and examples are available from commercial sources unless otherwise indicated.

[0021] As used herein, the singular forms "a," "an," and "the" refer to both the singular and the plural, unless expressly stated to refer to the singular only.

[0022] The term "about" means that the numbers covered are not limited to the exact numbers set forth herein, but rather refers to numbers substantially before or after the cited number without departing from the scope of the present invention. As used herein, "about" is understood by those of ordinary skill in the art and varies to some extent in the context in which it is used. If there are uses of the term that are not clear to those of skill in the art given the context in which it is used, "about" means up to plus or minus 10% of the particular term.

[0023] The term "blood relative" refers to two or more subjects who share one or more common ancestors. Non-limiting examples of blood relatives of a subject include the subject's mother, father, brother, sister, son, daughter, grandfather, grandmother, aunt, uncle, niece, nephew, and / or first cousin. In some embodiments, the blood relative is male. In some embodiments, the blood relative is female.

[0024] The term "gene" relates to a stretch of DNA or RNA that encodes a polypeptide or plays a functional role in an organism. A gene can be a wild-type gene or a variant or mutation of a wild-type gene. A "gene of interest" refers to a gene or variant of a gene that may or may not be known to be associated with a particular phenotype or risk of a particular phenotype.

[0025] "Expression" refers to the process by which a polynucleotide is transcribed from a DNA template (such as into mRNA or other RNA transcript) and / or the process by which the transcribed mRNA is then translated into a peptide, polypeptide, or protein. Where a nucleic acid sequence encodes a peptide, polypeptide, or protein, gene expression relates to the production of the nucleic acid (e.g., DNA or RNA, such as mRNA) and / or the peptide, polypeptide, or protein. Thus, "expression level" can refer to the amount of nucleic acid (e.g., mRNA) or protein in a sample.

[0026] Described is a new and unexpected method that uses genetic information to determine the risk that a subject has a phenotype.For non-Mendelian genes, the probability that a subject will express a phenotype can be calculated from population data.However, if a subject has a gene mutation that is the same mutation as their relatives, and if the relatives have the phenotype, the probability that a subject will express a phenotype can be calculated more accurately than if using population risk calculated without the data of relatives.

[0027] Genetic selection The gene of interest may be identified by any means known in the art. For example, the gene of interest may be selected based on the subject's personal genome. In some embodiments, the gene of interest is a known non-Mendelian gene. In some embodiments, the gene of interest is a genetic variant of interest. In some embodiments, the gene of interest is not independently statistically significantly associated with the observed phenotype. In some embodiments, the gene of interest is known to be associated with the observed phenotype.

[0028] Dataset Selection The data set for determining risk can be obtained by any means known in the art. For example, the first data set can include genotypic data and phenotypic data for the subject and also for one or more relatives of the subject. The genotypic data can include expression data for one or more genes of interest. The phenotypic data can include observable characteristics or disease traits, including specific symptoms of disease or observable characteristics of the subject that are not associated with any disease.

[0029] The first dataset may be prepared by detecting expression of one or more genes of interest in the subject and one or more relatives of the subject. In some embodiments, genotypic and / or phenotypic data from the subject and from one or more relatives of the subject are obtained from multiple sources.

[0030] In some embodiments, the first dataset further comprises information regarding the age of the subject and / or relatives, hi some embodiments, the first dataset comprises information regarding the age of onset of a phenotype (e.g., a disease or condition or a particular symptom associated with a disease or condition) in the subject and / or relatives of the subject.

[0031] In some embodiments, the subject has a particular phenotype. In some embodiments, the subject does not have the phenotype. In some embodiments, the subject carries one or more genes of interest. In some embodiments, the subject does not carry the genes of interest. In some embodiments, one or more relatives of the subject carry one or more genes of interest and exhibit the phenotype also observed in the subject. In some embodiments, one or more of the subject's relatives carry one or more genes of interest and exhibit the phenotype also not observed in the subject. In some embodiments, one or more of the subject's relatives carry one or more genes of interest and exhibit the phenotype also observed in the subject. In some embodiments, one or more of the subject's relatives do not carry one or more genes of interest and exhibit the phenotype not observed in the subject.

[0032] A second data set can be used that has genotype population data and phenotype population data. Such population data for non-Mendelian genes can be used to determine the probability of a subject expressing a phenotype. In some embodiments, the population data includes data from two or more relatives. In some embodiments, the population data includes data from one or more sets of two or more relatives, such as two, three, four, five, ten or more sets of relatives. The relationship between relatives can be the same as, different from, or overlapping with the relationship between the subject and relative in the first data set. In some embodiments, the two or more relatives from the population data are not relatives to the subject used for the first data set. In some embodiments, the data for the second data set is collected from one or more public databases. Non-limiting examples of such databases may include the United Kingdom (UK) Biobank; various genotype-phenotype data sets that are part of the Database of Genotype and Phenotype (dbGaP) maintained by the National Center for Biotechnology Information (NCBI); The European Genome-phenome Archive; OMIM; GWASdb; PheGenl; Genetic Association Database (GAD); and PhenomicDB.

[0033] The datasets may be collected using data from one or more of a variety of tissues or bodily fluids. For example, the first and / or second datasets may independently include data associated with brain tissue, heart tissue, lung tissue, kidney tissue, liver tissue, muscle tissue, bone tissue, stomach tissue, intestinal tissue, esophageal tissue, and / or skin tissue, or any combination of such tissues. Additionally or alternatively, the datasets may include data associated with bodily fluids, such as urine, blood, plasma, serum, saliva, semen, sputum, cerebrospinal fluid, mucus, sweat, vitreous humor, and / or milk, or any combination of such bodily fluids.

[0034] In some embodiments, the dataset is collected using data from subjects with a particular condition and / or a particular symptom, hi some embodiments, the dataset is collected using samples from multiple tissues and / or multiple bodily fluids.

[0035] Phenotypic Risk Score Some embodiments include determining a phenotypic risk score for the subject. The phenotypic risk score may indicate the likelihood that the subject will develop a particular phenotype (e.g., a disease or condition or a symptom of a disease or condition). The polygenic risk score may be determined using machine learning (including supervised and / or unsupervised machine learning algorithms). In some embodiments, the polygenic risk score may be calculated by training a model on a first dataset (e.g., having genotypic and phenotypic data for the subject and one or more relatives of the subject) and a second dataset (e.g., having genotypic population data and phenotypic population data). In some embodiments, the training includes a normalization (e.g., normalizing the transcript expression levels of the genes of interest to the expression levels of housekeeping genes) and / or a standardization step (e.g., via SVM to scale transcript abundance to a zero mean).

[0036] In some embodiments, the phenotypic risk score is determined using resampling techniques, such as oversampling or undersampling. Some embodiments include using binning and / or bagging techniques. In some embodiments, parametric and / or non-parametric statistical tests are used to assess expression differences between subjects.

[0037] In some embodiments, phenotypic risk score can be used to classify a subject as being at risk for a phenotype.Classification can be performed, for example, using SVM, logistic regression analysis, random forest, naive Bayes and / or Adaboost.In some embodiments, phenotypic risk score is the probability that a subject will develop a phenotype.In some embodiments, phenotypic risk score is the probability that a subject will develop a phenotype by a certain age.

[0038] In some embodiments, the phenotypic risk score is determined using an area under the curve (AUC) measurement. For example, the AUC can be greater than about 0.5, greater than about 0.55, greater than about 0.6, greater than about 0.65, greater than about 0.7, greater than about 0.75, greater than about 0.8, greater than about 0.85, greater than about 0.9, greater than about 0.95, greater than about 0.97, greater than about 0.98, or greater than about 0.99.

[0039] Execution System The methods described herein can be implemented on a variety of systems. For example, in some embodiments, a system for determining a phenotypic risk score includes one or more processors coupled to a memory. The methods can be implemented using code and data stored and executed on one or more electronic devices. Such electronic devices can store and communicate code and data (internal and / or with other electronic devices over a network) using computer-readable media such as non-transitory computer-readable storage media (e.g., magnetic disks; optical disks; random access memory; read-only memory; flash memory devices; phase-change memory) and transitory computer-readable transmission media (e.g., electrical, optical, acoustic or other forms of propagated signal-carrier waves, infrared signals, digital signals, etc.).

[0040] The memory can be loaded with computer instructions for training a model to determine a phenotypic risk score. In some aspects, the system runs on a computer, such as a personal computer, a portable computer, a workstation, a computer terminal, a network computer, a supercomputer, a massively parallel computing platform, a television, a mainframe, a server farm, a widely distributed series of loosely networked computers, or any other data processing system or user device.

[0041] The method may be performed by processing logic that may include hardware (e.g., circuitry, dedicated logic, etc.), firmware, software (e.g., embodied on a non-transitory computer-readable medium), or a combination of both. The operations described may be performed in any order or in parallel.

[0042] Typically, a processor may receive instructions and data from a read-only memory or a random access memory or both. A computer typically contains a processor capable of performing actions according to instructions and one or more storage devices for storing instructions and data. Typically, a computer also includes or is operably coupled to one or more mass storage devices for storing data, such as magnetic disks, magneto-optical disks, optical disks, or solid-state drives, for receiving data therefrom or transmitting data thereto, or both. However, a computer need not have such devices. Additionally, a computer may be embedded in another device, such as a smartphone, a portable audio or media player, a game console, a global positioning system (GPS) receiver, or a portable storage device (such as a Universal Serial Bus (USB) flash drive), to name a few. Suitable devices for storing computer program instructions and data include all forms of non-volatile memory, media, and storage devices, including, by way of example, semiconductor storage devices, such as EPROM, EEPROM, and flash memory devices; magnetic disks, such as internal hard disks or removable disks; magneto-optical disks; and CD ROM and DVD-ROM disks. The processor and the memory may be provided by, or incorporated in, special purpose logic circuitry.

[0043] One or more computer systems may be configured to perform particular operations or actions by having software, firmware, hardware, or a combination thereof installed on the system that causes the operation to be performed or that causes the system to perform an action. One or more computer programs may be configured to perform particular operations or actions by containing instructions that, when executed by a data processing device, cause the device to perform the action.

[0044] An exemplary execution system is shown in Figure 2. Such a system may be used to perform one or more of the operations described herein. The computing device may be coupled to other computing devices in a LAN, an intranet, an extranet, and / or the Internet. The computing device may operate in the capacity of a server machine in a client-server network environment or in the capacity of a client in a peer-to-peer network environment.

[0045] Diagnosis and Treatment In some embodiments, a subject (e.g., a human subject) is diagnosed as having a condition or disease, or as at risk of having a condition or disease, based on the phenotypic risk score. For example, in some embodiments, a subject with a particular phenotypic risk score is diagnosed as having a condition or disease. In some embodiments, a subject with a particular phenotypic risk score is determined to be at increased risk of developing a condition or disease or one or more symptoms thereof.

[0046] Some embodiments include treating a subject determined to have or be at increased risk for a condition or disease or one or more symptoms of a disease or condition. The term "treat" is used herein to characterize a method or process that aims to (1) delay or prevent the onset or progression of a disease or condition; (2) delay or halt the progression, aggravation, or worsening of a symptom of a disease or condition; (3) ameliorate a symptom of a disease or condition; or (4) cure a disease or condition. Treatment may be administered after the onset of a disease or condition. Alternatively, treatment may be administered before the onset of a disease or condition for prophylactic or preventative action. In this case, the term "prevention" is used. In some embodiments, the treatment includes administering a drug product listed in the latest version of FDA's Orange Book, which is incorporated herein by reference in its entirety. Representative conditions and treatments are also described in PHYSICIANS' DESK REFERENCE (PRD Network 71st ed. 2016); and THE MERCK MANUAL OF DIAGNOSIS AND THERAPY (Merck 20th ed. 2018), each of which is incorporated by reference in its entirety.

[0047] The following examples are offered to illustrate the invention, but it should be understood that the invention is not limited to the specific conditions or details of these examples.

[0048] Working Example Example 1: Refining risk using relative information As a simplified example, consider a possible mutation m on a gene g, X gm When a mutation is present, X gm = 1, and when there is no mutation, X gm = 0. For efficiency, we use a binary indicator variable, X gm are used interchangeably to refer to a mutation, the locus of a mutation, and as an indication of whether a mutation is present at that locus. gmIn a subpopulation with a phenotype, P(X gm )=p gm (This symbol will be used throughout the following examples.) gm can be determined from the test:

number

[0049] For this embodiment, X gm In addition to the mutations in the gene, only one other mutation is known to affect the phenotype (e.g., mutation n and gene h, X hn ), X hn X gm Assume that the X hn X gm and X hn If X is mutated, the subject will develop the phenotype, but if X is mutated, the subject will develop the phenotype. gm Or X hn We assume that the mutation acts like a switch, that is, if only the mutation is present, the subject will not express the phenotype. gm If the mutation is present and the mother has the phenotype, the risk to the child is p gm For this example, mutation X can be predicted more accurately than if risk was determined based on subpopulation testing. hn We assume that the probability of inheriting this mutation from a father or mother with multiple copies is rare enough that it can be ignored. hn Since there is a 50% chance of inheriting the mutation, the child has approximately a 50% chance of expressing the phenotype. For this example, if the general population risk for the phenotype is around 1%, and the mutation X gm Mutation X is a rare mutation that increases risk by 50% and does not include data from relatives. gmFor individuals with X, the risk increases to approximately 1.5%. gm Mutation, mother is X gm If you know you have the mutation and have the phenotype, your child's risk is now 50% instead of 1.5%. So even if you think of it as a moderate increase in risk of 50%, X hn X gm Given the simplified scenario in which the chromosome acts as a switch for the chromosome 1 gene, the impact of maternal information bearing the mutation and phenotype is substantial.

[0050] X gm In a scenario where we do not know that all mutations that interact with X affect the phenotype or the mechanism of their interaction, we can apply the concepts outlined above to empirically estimate the probability that a subject will develop a phenotype if blood relatives have the same mutation and associated phenotype. This involves extracting information from genotype-phenotype databases to calculate the inherent risk for a particular relationship and a particular mutation or gene. The subject is asked to identify a blood relative r (where r can be mother, father, brother, sister, son, daughter, grandfather, grandmother, aunt, uncle, niece, nephew, cousin, etc.) and a mutation X. gm Now, assuming that the subject is aged prior to the age at which the phenotype would be expressed, the subject's lifetime risk can be considered without adjustment for the effect of the subject's current age (which can be incorporated separately as discussed below). gm and has the mutation X gm and a relative r who has the phenotype and who, if the phenotype were to be expressed, is either deceased or of an age at which the phenotype would be expressed. gm,r Then, we find the number of people affected by the phenotype (so that we can calculate the complete lifetime risk). gm,r Number of people from N gm , r , affected The estimated probability that a subject will express the phenotype is:

number

[0051] Example 2 - Limited Data Management For normal approximation of the binomial distribution - exact binomial can be used for small numbers -

number

number

number

number

number

number

[0052] Or, for the sake of conservatism, gm at a few standard deviations in the direction of

number

number

number

number

number

[0053]

number

number

number

[0054] Example 3: Combining Similar Kinship Relationships Another approach is to combine data on male and female relatives and assume that genes that are present on the X chromosome but not on the Y chromosome have minimal effect on phenotypic expression.

[0055] Additionally, information from relatives who share a similar amount of genetic material with the subject of interest can be combined. In this case, let r represent each group of relatives who share the same amount of genetic information with the subject. The counts for each group r are pooled. That is, using a similar approach as above, N gm,r Here, the mutation X gm and has the mutation X gm and N represents the number of people in the database who have a relative in group r with the phenotype; gm,r,affected Here, r represents the number of affected individuals. For example, r=1 / 2 represents the group with half the genetic information of the subject - mother, father, brother, sister, son, daughter; r=1 / 4 represents the group with one-quarter of the genetic information - grandfather, grandmother, half-brother or half-sister, aunt, uncle, niece, nephew, grandson, grandson, etc.; r=1 / 8 represents the group with one-eighth of the genetic information. In this approach, X gm Any two subjects who have the phenotype and have relatives in the same relative group r have the same

number

[0056] Example 4: Gene-level mutations An alternative approach is to address the presence of mutations at the gene level, rather than treating each variant in isolation. gLet x correspond to a mutated gene g, which is a mutation X known to have the same effect, e.g., loss of function, in functional gene g. gm , m = 1...M. In this case, N is the number of people with a loss-of-function mutation in functional gene g and their relatives in group r who also have a mutation of that type, such as a loss-of-function mutation in gene g. g,r Then we can calculate the probability at the gene level:

number

[0057] Example 5: Incorporation of Age Another approach deals with the ages of the people in the database and gm,r Eliminate the need to only consider deceased people in the calculation of N gm,r Instead of N g,r can be calculated.

[0058]

number

number

[0059] Without changing the basic concept, we can apply p to subjects that have not yet expressed the phenotype. g,r Note that there are many other ways to approximate (A). For example, for limited data, p r (A) or p g (A) is calculated by p g,r (A) can be approximated, i.e., they have the mutation X g or mutation X g It does not filter subjects in the database based on requiring that they have some relative r.

[0060] Another approach with limited data is to consider whether they are mutations X g or relative r, and calculate a histogram of when the phenotype is expressed. Such a simulated example histogram is shown in the bars of Figure 1 for a phenotype with a mean age of onset of 60 years. The cumulative probability of an individual expressing the phenotype as a function of age can be calculated, shown in red, which approaches p, the population frequency of expressing the phenotype, in this case p=0.2. For individual subjects with a risk different from p, the relative probability for the age at which the phenotype will be expressed can be estimated to be invariant. The estimated lifetime risk

number

number

number

number

number

[0061] Example 6: Combining the effects of multiple relatives Another approach involves the situation where a subject has multiple relatives with the mutation and phenotype.The simplest approach is to use the same method as above, but instead of counting the cases with only one relative in the database, count all cases with the same set of multiple relatives, where the relatives are classified according to the above grouping r, such as sharing the same amount of common genetic data with the subject, and being of a certain gender.For example, when grouping by gender and by the amount of common genetic information, a subject with one father, one uncle and one grandfather who all have the mutation and disease can be counted together with a subject with two sons and one uncle who all have the mutation and disease, for example.As another example, when grouping only by the amount of common genetic information, a subject with one father, one aunt and one grandmother who all have the mutation and disease can be counted together with a subject with two sons and one uncle who all have the mutation and disease, for example.

[0062] In the case of limited data, the risk can be approximated, so that a lower bound is obtained, typically by ignoring some of the relatives of the subject who have the variant and the disease, allowing more data to be pooled. In this case, typically prioritizing relatives who share more genetic information with the subject. For example, a subject who has one father, one uncle, and one grandfather, all of whom have the variant and the disease, can be treated as a subject who has only one relative, the father, who has the variant and the disease.

[0063] Another approach is to combine data across several categories of relatives. There are many empirical or heuristic approaches to this concept. For example, one representative approach is g This is appropriate when the number of genes influencing penetrance of a gene is very large and the individual effect size of each of these genes is very small.

number

number

[0064] Then, using a series of equations for each kinship group, [ka] which gives the individual variances of each group:

number

number

number

[0065] Example 7: Application of the method to polygenic risk scores The above techniques can be used in terms of polygenic risk scores or regression models that describe the probability of developing a phenotype, or in other machine learning models to determine the probability of a phenotype. For example, phenotypes can be modeled based on the following polygenic or multivariate, regression models at the mutation or gene level:

number

[0066] As already mentioned, the indicator variable X g Let X be all mutations of a similar type, such as loss of function or a particular type of gain of function. gm If the gene has a mutation, X g = 1, otherwise X g = 0. This same concept can be extended to different classifications of mutations, such as different classes of loss-of-function or gain-of-function mutations.

[0067] The following example is valid at the mutation level without loss of generality. Regression models such as the above can be adjusted based on the probabilities derived for a particular individual using the methods outlined herein. Consider the case where P is a polygenic risk score (PRS) that is not a probability per se, but has meaning in the context of other scores, such as to determine which percentile a subject's genetic risk score falls in. In this case, one can set a bias parameter b0=0, etc. for the effect size of each gene or variant. This effect size b gm is the mutation X gm It can be estimated by taking the logarithm of the ratio of the odds of expressing disease phenotype D with and without.

number

number

number

number

number

number

number

number

number

number

[0068] When an individual of interest is known to have r affected relatives, p is the probability of expressing the phenotype given r affected relatives. r One can use effect sizes for and vary parameters to take this into account.

number

number

number

number

[0069] Why is the parameter p and not p r To understand how a polygenic model can be defined for x, consider that we are trying to model the probability of a phenotype resulting from multiple genetic variables. Suppose we have three genetic variables X1, X2, and X3:

number

number

number

number

number

number

number

[0070] logP(D|X1...X G The above explanation for calculating logP(D|X1...X G This paper outlines the derivation and concepts behind polygenic prediction models that sum the log odds ratios, or approximations thereto, for each SNP to estimate the log odds of a given gene.

number

number

number

number

number

[0071] X g If =1, then as above

number

[0072] PRS or P(D|X1...X g The goal of estimating ) is to reproduce as closely as possible the probability of a disease or phenotype for a subject and to distinguish as thoroughly as possible between subjects with different probabilities of disease. To demonstrate the value of using relative information, a more theoretical probability formula may be used in the following description and in the MATLAB simulation code discussed below. That is, the following description is based, as is commonly done, on the variables X r The effectiveness of estimating disease probability incorporating relative information available in P(D|X1...X g ) to compare their effectiveness in estimating

[0073] The above P(D|X1...X g ) in the derivation to estimate the variables X1...X gWe make some approximations based on the strong assumption of independence of the r A variable is assumed to represent whether a relative or set of relatives has a disease or phenotype of interest. This variable is generally denoted as X1...X G For example, if these are genetic variables, the presence of an affected relative determines the probability that the subject has the gene or X1 = 1,...,X G = 1. However, instead of calculating the risk relative to the population average, P(D), if we instead calculate the risk relative to the probability of having the disease or phenotype of interest, P(D|X r ), variables X1...X G Without extending the independence hypothesis in situations beyond X, the information contained in the family history can be used to generate more powerful polygenic prediction models. r Given X1, X2, and X3, we use the same independence hypothesis to find X r We can calculate the risk using the same derivation arguments as above for P(D|X1X2X3) without having to ignore the dependencies between and X1X2...X3.

number

[0074] Similarly, genetic, lifestyle, environmental or phenotypic variables X1...X G We can extend this method to any number of . If we can assume independence between these variables:

number

[0075] Similar to what was described above, one approach is to create a PRS to estimate the effect size b g,r The goal is to calculate:

number

number

[0076] The following discussion focuses on the case of three genetic variables that are approximated as independent. Instead of P(D|X1X2X3), we model P(D|X r X1X2X3) relative to X r To illustrate the value of using available data from the PRS, which is less accurate in its ability to model the probability of disease for each individual, typically resulting in more false positives, increased medical costs, poorer outcomes, etc. The following explanation uses the PRS instead of the PRS. Xr The above formula can be used similarly to calculate P(D|X1X2X3X r ) is used based on the estimate.

[0077] Consider an example where we have two genes, X1 and X2, with individual prevalences in the population of 1 / 20 and 1 / 50, where X2 acts as a switch for X1, such that a subject will have the phenotype if both X1=1 and X2=1. To make this example more illustrative, we further assume that these are not the only factors that can cause the disease, and that there is another gene, X3, that causes the disease with 100% penetrance when present. Furthermore, without loss of generality of the concept, we consider that the set of relatives considered for each subject is only its parents, i.e., if either parent has the disease, then X2 will have the phenotype. r = 1, if neither parent has the disease, X r= 0. The MATLAB code in Appendix A implements the inventive concept applied to this scenario. Note that the simulations use the same data to create the model and to test the model. This is because there are very few parameters being estimated compared to the number of subjects being simulated, so generating new test data yields roughly the same results. That is, this reduction to implement in MATLAB does not focus on the effect of limited data, but rather on the versatility of each of the modeling approaches or the ability of the models to accurately estimate disease probabilities as described above and captured in the data.

[0078] Figures 3A and 3B show histograms of predictions on the logarithmic scale of the y-axis for each subject when gene X3 has a frequency of 1 / 100 in the general population, and only a subset of related genes are available in the model. That is, Figure 3A describes a model using only genetic variables X1 and X2, and Figure 3B describes a model using only genetic variables X1 and X3. Such scenarios are common, for example, when a polygenic model covers only certain related SNPs in a subset of genes, while other related genes are not included in the model. This occurs, for example, because in a model that assumes linearity of the effect and independence of genetic variables, the genetic variables that are excluded do not reach statistical significance, or because the genes that are excluded are affected by many rare variants that have a significant effect together but are not associated with any one common variant that has a frequency high enough to be recognized as a SNP or "single nucleotide polymorphism". In both figures, the truth for each subject is included, i.e., whether each subject actually developed the disease or not, taken as 1 or 0, respectively. r Figure 3B illustrates modeling the data by estimating P(D|X1X3) and P(D|X rFigure 3C illustrates modeling of the data by estimating the relatives of the siblings, P(D|X1X2X3) and P(D|X r X1X2X3) is obtained. Figure 3C also assumes that P(X3) = 1 / 100.

[0079] Table 1 shows the relatives X, who are the parents in this example. r We present the root mean square error (RMSE) of several models from simulations with different combinations of genetic variables when different combinations of genes are used in polygenic risk models with and without information on . [Table 1]

[0080] In the latter case, represented by FIG. 3C, the disease history of the parents, i.e., X r The incorporation of changes the RMSE from 0.0846 to 0.0312, or a 63% decrease.

[0081] Figures 4A-C represent a similar situation to Figures 3A-3C, except that P(X3)=1 / 500. Figures 5A-C represent a similar situation to Figures 3A-3C, except that P(X3)=1 / 2000. The RMSE for all of these scenarios described in Figures 3, 4 and 5, along with other scenarios, are captured in Table 1. Note that in general, the incorporation of kinship information Xr generally improves performance in matching the truthful data.

[0082] Example 8: Alternative approaches to modeling phenotype probabilities When modeling the probability of a phenotype (rather than the risk score itself), the approaches described herein can also be used to calibrate parameters for individuals, for example, using an approach based on logistic regression. At the gene level, a logistic regression model is:

number

[0083] If the parameters a0 and b0 can be fitted to the data, then b g We use the concepts outlined above to select

[0084] Using nonlinear combinations of genes or variants, P(D|X r X1...X G ) Here, again without loss of generality, we work at the gene rather than the variant level. Suppose we want to capture interactions between genes, and we are only looking at two gene interactions (the same concept can be applied to three or more gene interactions, although there may be challenges with the data). Some logical combination of two genes X1 and X2:

number

number

number

[0085] Appendix A: MATLAB Expressions % rel_sim % simulates training polygenic prediction relative using relationships %simulation parameters n = 1000000; % 1000000; % number of families p_x1 = 1 / 20; %1 / 20; % P(X1) the probability of X1 variant in the general population p_x2 = 1 / 50; %1 / 50; % P(X2) the probability of X2 variant in the general population p_x3 = 1 / 2000; %1 / 100; %1 / 500; %1 / 2000; % P(X3) the probability of X3 variant in the general population % setting up variables % Assume no denovo variants % assume no homozygotes of variant in parents % ph_x1 = min(roots([1 -2 p_x1])); % probability per homolog; comment out if assume no homozygotes of variant in parents % ph_x2 = min(roots([1 -2 p_x2])); % probability per homolog; comment out if assume no homozygotes of variant in parents % create parents par1_vec_x1 = (rand(n,1)<p_x1); % 1 if have variant 0 if don't par1_vec_x2 = (rand(n,1)<p_x2); % 1 if have variant 0 if don't par1_vec_x3 = (rand(n,1)<p_x3); % 1 if have variant 0 if don't par2_vec_x1 = (rand(n,1)<p_x1); % 1 if have variant 0 if don't par2_vec_x2 = (rand(n,1)<p_x2); % 1 if have variant 0 if don't par2_vec_x3 = (rand(n,1)<p_x3); % 1 if have variant 0 if don't par1_vec_dis = (par1_vec_x1 & par1_vec_x2) | par1_vec_x3; par2_vec_dis = (par2_vec_x1 & par2_vec_x2) | par2_vec_x3; par_vec_dis = par1_vec_dis | par2_vec_dis; % create children p_inh_x1 = 0.5 *par1_vec_x1 + 0.5 * par2_vec_x1 - 0.25 * par1_vec_x1. * par2_vec_x1; chi_vec_x1 = (rand(n,1)<p_inh_x1); p_inh_x2 = 0.5 * par1_vec_x2 + 0.5 * par2_vec_x2 - 0.25 * par1_vec_x2. * par2_vec_x2; chi_vec_x2 = (rand(n,1)<p_inh_x2); p_inh_x3 = 0.5 * par1_vec_x3 + 0.5 * par2_vec_x3 - 0.25 * par1_vec_x3. * par2_vec_x3; chi_vec_x3 = (rand(n,1)<p_inh_x3); chi_vec_dis = (chi_vec_x1 & chi_vec_x2) | chi_vec_x3; % child gets sick if either (x1 and x2) or x3 %%%% train model for phenotype using standard method: P(D / X1X2) = P(D) * P(D / X1) / P(D) * P(D / X2) / P(D) * P(D / X3) / P(D) % just using child data for now; can do this also for parents p_dis_h = length(find(chi_vec_dis==1)) / n chi_vec_x1e1_ind = find(chi_vec_x1==1); p_dis_x1e1_h=length( find(chi_vec_dis(chi_vec_x1e1_ind)==1) ) / length(chi_vec_x1e1_ind); chi_vec_x1e0_ind = find(chi_vec_x1==0); p_dis_x1e0_h=length( find(chi_vec_dis(chi_vec_x1e0_ind)==1)) / length(chi_vec_x1e0_ind); chi_vec_x2e1_ind = find(chi_vec_x2==1); p_dis_x2e1_h=length( find(chi_vec_dis(chi_vec_x2e1_ind)==1) ) / length(chi_vec_x2e1_ind); chi_vec_x2e0_ind = find(chi_vec_x2==0); p_dis_x2e0_h=length( find(chi_vec_dis(chi_vec_x2e0_ind)==1) ) / length(chi_vec_x2e0_ind); chi_vec_x3e1_ind = find(chi_vec_x3==1); p_dis_x3e1_h=length( find(chi_vec_dis(chi_vec_x3e1_ind)==1) ) / length(chi_vec_x3e1_ind); chi_vec_x3e0_ind = find(chi_vec_x3==0); p_dis_x3e0_h=length( find(chi_vec_dis(chi_vec_x3e0_ind)==1) ) / length(chi_vec_x3e0_ind); % prediction on the training data % can also implement this on test data p_dis_x1_h = zeros(n,1); p_dis_x1_h(chi_vec_x1e1_ind)=p_dis_x1e1_h; p_dis_x1_h(chi_vec_x1e0_ind)=p_dis_x1e0_h; p_dis_x2_h = zeros(n,1); p_dis_x2_h(chi_vec_x2e1_ind)=p_dis_x2e1_h; p_dis_x2_h(chi_vec_x2e0_ind)=p_dis_x2e0_h; p_dis_x3_h = zeros(n,1); p_dis_x3_h(chi_vec_x3e1_ind)=p_dis_x3e1_h; p_dis_x3_h(chi_vec_x3e0_ind)=p_dis_x3e0_h; % prediction using x1 and x2 p_dis_x1x2_h = p_dis_h * (p_dis_x1_h / p_dis_h). * (p_dis_x2_h / p_dis_h); % prediction using x1 and x3 p_dis_x1x3_h = p_dis_h * (p_dis_x1_h / p_dis_h). * (p_dis_x3_h / p_dis_h); % prediction using x1,x2 and x3 p_dis_x1x2x3_h=p_dis_h * (p_dis_x1_h / p_dis_h). * (p_dis_x2_h / p_dis_h). * (p_dis_x3_h / p_dis_h); %%%% train model for phenotype using relative method: P(D / Xr / X1X2) = P(D / Xr) * P(D / XrX1) / P(D / Xr) *P(D / XrX2) / P(D / Xr) % just using child data for now to train; can train and test also for parents par_vec_dis_ind = find(par_vec_dis==1); p_dis_xr_h=length( find(chi_vec_dis(par_vec_dis_ind)==1) ) / length(par_vec_dis_ind); % computing P(D / XrX1) for all states chi_vec_xre1_x1e1_ind = find(par_vec_dis==1 & chi_vec_x1==1); p_dis_xre1_x1e1_h=length( find(chi_vec_dis(chi_vec_xre1_x1e1_ind)==1) ) / length(chi_vec_xre1_x1e1_ind); chi_vec_xre0_x1e1_ind = find(par_vec_dis==0 & chi_vec_x1==1); p_dis_xre0_x1e1_h=length( find(chi_vec_dis(chi_vec_xre0_x1e1_ind)==1) ) / length(chi_vec_xre0_x1e1_ind); chi_vec_xre0_x1e0_ind = find(par_vec_dis==0 & chi_vec_x1==0); p_dis_xre0_x1e0_h=length( find(chi_vec_dis(chi_vec_xre0_x1e0_ind)==1) ) / length(chi_vec_xre0_x1e0_ind); chi_vec_xre1_x1e0_ind = find(par_vec_dis==1 & chi_vec_x1==0); p_dis_xre1_x1e0_h=length( find(chi_vec_dis(chi_vec_xre1_x1e0_ind)==1) ) / length(chi_vec_xre1_x1e0_ind); % computing P(D / XrX2) for all states chi_vec_xre1_x2e1_ind = find(par_vec_dis==1 & chi_vec_x2==1); p_dis_xre1_x2e1_h=length( find(chi_vec_dis(chi_vec_xre1_x2e1_ind)==1) ) / length(chi_vec_xre1_x2e1_ind); chi_vec_xre0_x2e1_ind = find(par_vec_dis==0 & chi_vec_x2==1); p_dis_xre0_x2e1_h=length( find(chi_vec_dis(chi_vec_xre0_x2e1_ind)==1) ) / length(chi_vec_xre0_x2e1_ind); chi_vec_xre0_x2e0_ind = find(par_vec_dis==0 & chi_vec_x2==0); p_dis_xre0_x2e0_h=length( find(chi_vec_dis(chi_vec_xre0_x2e0_ind)==1) ) / length(chi_vec_xre0_x2e0_ind); chi_vec_xre1_x2e0_ind = find(par_vec_dis==1 & chi_vec_x2==0); p_dis_xre1_x2e0_h=length( find(chi_vec_dis(chi_vec_xre1_x2e0_ind)==1) ) / length(chi_vec_xre1_x2e0_ind); % computing P(D / XrX3) for all states chi_vec_xre1_x3e1_ind = find(par_vec_dis==1 & chi_vec_x3==1); p_dis_xre1_x3e1_h=length( find(chi_vec_dis(chi_vec_xre1_x3e1_ind)==1) ) / length(chi_vec_xre1_x3e1_ind); chi_vec_xre0_x3e1_ind = find(par_vec_dis==0 & chi_vec_x3==1); p_dis_xre0_x3e1_h=length( find(chi_vec_dis(chi_vec_xre0_x3e1_ind)==1) ) / length(chi_vec_xre0_x3e1_ind); chi_vec_xre0_x3e0_ind = find(par_vec_dis==0 & chi_vec_x3==0); p_dis_xre0_x3e0_h=length( find(chi_vec_dis(chi_vec_xre0_x3e0_ind)==1) ) / length(chi_vec_xre0_x3e0_ind); chi_vec_xre1_x3e0_ind = find(par_vec_dis==1 & chi_vec_x3==0); p_dis_xre1_x3e0_h=length( find(chi_vec_dis(chi_vec_xre1_x3e0_ind)==1) ) / length(chi_vec_xre1_x3e0_ind); % prediction on the training data % could also implement this on separate test data % computing P(D / XrX1) p_dis_xr_x1_h = zeros(n,1); p_dis_xr_x1_h(chi_vec_xre1_x1e1_ind)=p_dis_xre1_x1e1_h; p_dis_xr_x1_h(chi_vec_xre0_x1e1_ind)=p_dis_xre0_x1e1_h; p_dis_xr_x1_h(chi_vec_xre0_x1e0_ind)=p_dis_xre0_x1e0_h; p_dis_xr_x1_h(chi_vec_xre1_x1e0_ind)=p_dis_xre1_x1e0_h; % computing P(D / XrX2) p_dis_xr_x2_h = zeros(n,1); p_dis_xr_x2_h(chi_vec_xre1_x2e1_ind)=p_dis_xre1_x2e1_h; p_dis_xr_x2_h(chi_vec_xre0_x2e1_ind)=p_dis_xre0_x2e1_h; p_dis_xr_x2_h(chi_vec_xre0_x2e0_ind)=p_dis_xre0_x2e0_h; p_dis_xr_x2_h(chi_vec_xre1_x2e0_ind)=p_dis_xre1_x2e0_h; % computing P(D / XrX3) p_dis_xr_x3_h = zeros(n,1); p_dis_xr_x3_h(chi_vec_xre1_x3e1_ind)=p_dis_xre1_x3e1_h; p_dis_xr_x3_h(chi_vec_xre0_x3e1_ind)=p_dis_xre0_x3e1_h; p_dis_xr_x3_h(chi_vec_xre0_x3e0_ind)=p_dis_xre0_x3e0_h; p_dis_xr_x3_h(chi_vec_xre1_x3e0_ind)=p_dis_xre1_x3e0_h; %%% computing key results % prediction using xr, x1 and x2 p_dis_xrx1x2_h=p_dis_xr_h * (p_dis_xr_x1_h / p_dis_xr_h). * (p_dis_xr_x2_h / p_dis_xr_h); % prediction using xr, x1 and x3 p_dis_xrx1x3_h=p_dis_xr_h * (p_dis_xr_x1_h / p_dis_xr_h). * (p_dis_xr_x3_h / p_dis_xr_h); % prediction using xr, x1, x2 and x3 p_dis_xrx1x2x3_h= p_dis_xr_h * (p_dis_xr_x1_h / p_dis_xr_h). * (p_dis_xr_x2_h / p_dis_xr_h). * (p_dis_xr_x3_h / p_dis_xr_h); %%% plotting key results %%raw data disp_vec = [1:10000]; % figure; plot(chi_vec_dis(disp_vec),'b.'); hold on; plot(chi_vec_dis(disp_vec),'b'); %%prediction using xr, x1 % plot(p_dis_xr_x1_h(disp_vec),'gx'); % prediction using x1 % plot(p_dis_x1_h(disp_vec),'ro'); %%prediction using x1 and x2 % plot(p_dis_x1x2_h(disp_vec),'ro'); % prediction using xr, x1 and x2 % plot(p_dis_xrx1x2_h(disp_vec),'gx'); %%histograms using x1, x2 (and xr) figure; hold on; [t1,c1] = hist(chi_vec_dis); bar(c1, log10(t1),'b'); [t2,c2] = hist(p_dis_xrx1x2_h); bar(c2, log10(t2),'g'); [t3,c3] = hist(p_dis_x1x2_h); bar(c3, log10(t3),'r'); legend('Truth', 'Estimate of P(D|XrX1X2)', 'Estimate of P(D|X1X2)'); ylabel('log10(count)'); xlabel('probability estimate'); title('histogram of estimates P(D|X1X2), P(D|XrX1X2)'); grid; %%prediction using x1 and x3 % plot(p_dis_x1x3_h,'ro'); % prediction using xr, x1 and x3 % plot(p_dis_xrx1x3_h,'gx'); % histograms using x1, x3 (and xr) figure; hold on; [tmp3,c3] = hist(p_dis_x1x3_h); bar(c3, log10(tmp3),'r'); [tmp1,c1] = hist(chi_vec_dis); bar(c1, log10(tmp1),'b'); [tmp2,c2] = hist(p_dis_xrx1x3_h); bar(c2, log10(tmp2),'g'); legend('Estimate of P(D|X1X3)', 'Truth', 'Estimate of P(D|XrX1X3)'); ylabel('log10(count)'); xlabel('probability estimate'); title('histogram of estimates P(D|X1X3), P(D|XrX1X3)'); grid; %%prediction using x1, x2 and x3 % plot(p_dis_x1x2x3_h,'ro'); % prediction using xr, x1, x2 and x3 % plot(p_dis_xrx1x2x3_h,'gx'); % histograms using x1, x2, x3 (and xr) figure; hold on; [tm3,c3] = hist(p_dis_x1x2x3_h); bar(c3, log10(tm3),'r'); [tm2,c2] = hist(p_dis_xrx1x2x3_h); bar(c2, log10(tm2),'g'); [tm1,c1] = hist(chi_vec_dis); bar(c1, log10(tm1),'b'); legend('Estimate of P(D|X1X2X3)','Estimate of P(D|XrX1X2X3)','Truth'); ylabel('log10(count)'); xlabel('probability estimate'); title('histogram of estimates P(D|X1X2X3), P(D|XrX1X2X3)'); grid; %%% comparing RMSE accuracy of results % prediction using x1 (and xr) p_dis_xr_x1_h_e = p_dis_xr_x1_h-chi_vec_dis; p_dis_x1_h_e = p_dis_x1_h-chi_vec_dis; p_dis_xr_x1_h_RMSE = sqrt(p_dis_xr_x1_h_e'*p_dis_xr_x1_h_e / n) p_dis_x1_h_RMSE = sqrt(p_dis_x1_h_e'*p_dis_x1_h_e / n) % prediction using x1 and x2 (and xr) p_dis_xrx1x2_h_e = p_dis_xrx1x2_h-chi_vec_dis; p_dis_x1x2_h_e = p_dis_x1x2_h-chi_vec_dis; p_dis_xrx1x2_h_RMSE = sqrt(p_dis_xrx1x2_h_e'*p_dis_xrx1x2_h_e / n) p_dis_x1x2_h_RMSE = sqrt(p_dis_x1x2_h_e'*p_dis_x1x2_h_e / n) % prediction using x1, x3 (and xr) p_dis_xrx1x3_h_e = p_dis_xrx1x3_h-chi_vec_dis; p_dis_x1x3_h_e = p_dis_x1x3_h-chi_vec_dis; p_dis_xrx1x3_h_RMSE = sqrt(p_dis_xrx1x3_h_e'*p_dis_xrx1x3_h_e / n) p_dis_x1x3_h_RMSE = sqrt(p_dis_x1x3_h_e'*p_dis_x1x3_h_e / n) % prediction using x1, x2, x3 (and xr) p_dis_xrx1x2x3_h_e = p_dis_xrx1x2x3_h-chi_vec_dis; p_dis_x1x2x3_h_e = p_dis_x1x2x3_h-chi_vec_dis; p_dis_xrx1x2x3_h_RMSE = sqrt(p_dis_xrx1x2x3_h_e'*p_dis_xrx1x2x3_h_e / n) p_dis_x1x2x3_h_RMSE = sqrt(p_dis_x1x2x3_h_e'*p_dis_x1x2x3_h_e / n)

Claims

1. 1. A method for outputting a non-Mendelian phenotype risk score, comprising: receiving from a first dataset (i) genotype data for a subject having one or more non-Mendelian genes of interest, and (ii) genotype and phenotype data for one or more relatives of the subject having one or more of the genes of interest; receiving genotypic population data and phenotypic population data from a second dataset, the populations including one or more sets of two or more relatives; training a model on the first and second datasets to determine a risk in the subject associated with one or more of the non-Mendelian genes of interest; outputting a phenotypic risk score for said subject. A method comprising:

2. 2. The method of claim 1, wherein the second data set comprises genotypic population data and phenotypic population data for multiple sets of two or more relatives.

3. the blood relatives in the first dataset include one or more of the subject's mother, father, brother, sister, son, daughter, grandfather, grandmother, aunt, uncle, niece, nephew, and first cousin; 3. The method of claim 1 or 2, wherein the second dataset comprises two or more subjects that have the same ancestry as the subject in the first dataset.

4. The method of any one of claims 1 to 3, wherein one or more of said blood relatives is a male relative.

5. The method of any one of claims 1 to 3, wherein one or more of said blood relatives is a female relative.

6. The method of any one of claims 1 to 5, wherein the first dataset comprises data for a plurality of relatives of the subject.

7. The method according to any one of claims 1 to 6, wherein one or more of said blood relatives is a male relative and one or more of said blood relatives is a female relative.

8. The method of any one of claims 1 to 7, wherein the gene of interest is a gene variant of interest.

9. The method of any one of claims 1 to 8, wherein the first and second datasets comprise data relating to age of onset of a phenotype.

10. A processor; a memory coupled to the processor for storing instructions that, when executed by the processor, cause the processor to perform operations; a system comprising: receiving from a first dataset (i) genotype data for a subject having one or more non-Mendelian genes of interest, and (ii) genotype and phenotype data for one or more relatives of the subject having one or more of the genes of interest; receiving genotypic population data and phenotypic population data from a second dataset, the populations including one or more sets of two or more relatives; training a model on the first and second datasets to determine a risk in the subject associated with one or more of the non-Mendelian genes of interest; outputting a phenotypic risk score for said subject. Including, the system.

11. A non-transitory machine-readable medium having instructions stored thereon that, when executed by a processor, cause the processor to perform operations, the operations including: receiving from a first dataset (i) genotype data for a subject having one or more non-Mendelian genes of interest, and (ii) genotype and phenotype data for one or more relatives of the subject having one or more of the genes of interest; receiving genotypic population data and phenotypic population data from a second dataset, the populations including one or more sets of two or more relatives; training, by the processor, a model on the first and second datasets to determine a risk in the subject associated with one or more of the non-Mendelian genes of interest; outputting a phenotypic risk score for said subject. A non-transitory machine-readable medium, including

12. 12. The non-transitory machine-readable medium of claim 11, wherein the second dataset comprises genotypic population data and phenotypic population data for multiple sets of two or more relatives.

13. the blood relatives in the first dataset include one or more of the subject's mother, father, brother, sister, son, daughter, grandfather, grandmother, aunt, uncle, niece, nephew, and first cousin; 13. The non-transitory machine-readable medium of claim 11 or 12, wherein the second dataset includes two or more subjects that have a same kinship relationship to the subject in the first dataset.

14. The non-transitory machine-readable medium of any one of claims 11 to 13, wherein one or more of the blood relatives is a male relative.

15. The non-transitory machine-readable medium of any one of claims 11 to 13, wherein one or more of the blood relatives is a female relative.

16. 16. The non-transitory machine-readable medium of claim 11, wherein the first data set includes data for a plurality of relatives of the subject.

17. The non-transitory machine-readable medium of any one of claims 11 to 16, wherein one or more of the blood relatives is a male relative and one or more of the relatives is a female relative.

18. 18. The non-transitory machine-readable medium of any one of claims 11 to 17, wherein the gene of interest is a gene variant of interest.

19. 19. The non-transitory machine-readable medium of any one of claims 11 to 18, wherein the first and second datasets comprise data related to age of onset of a phenotype.

20. 1. A method for outputting a polygenic risk score, comprising: receiving from a first dataset (i) genotype data for a subject having one or more non-Mendelian genes of interest, and (ii) genotype and phenotype data for one or more relatives of the subject having one or more of the non-Mendelian genes of interest; receiving genotypic population data and phenotypic population data from a second dataset, the populations including one or more sets of two or more relatives; training a model on the first and second datasets to predict risk in the subject based on the one or more non-Mendelian genes of interest; outputting a polygenic risk score for said subject. A method comprising:

21. training a model on the first and second datasets to predict how risk in the subject is modified by one or more non-Mendelian genes of interest compared to risk in the subject given phenotypic data of the relatives.

21. The method of claim 20, comprising:

22. The method of any one of claims 1 to 21, further comprising treating the subject based on the risk score.

Images