SimRVPedigree (version 0.4.4)

sim_RVped: Simulate a pedigree ascertained to contain multiple disease-affected relatives

Description

sim_RVped simulates a pedigree ascertained to contain multiple affected members, selects a proband, and trims the pedigree to contain only those individuals that are recalled by the proband.

Usage

sim_RVped(hazard_rates, GRR, num_affected, ascertain_span, FamID,
  founder_byears, stop_year = NULL, recall_probs = NULL,
  carrier_prob = 0.002, RVfounder = FALSE, NB_params = c(2, 4/7),
  fert = 1, first_diagnosis = NULL, sub_criteria = NULL)

Arguments

hazard_rates

An object of class hazard, created by hazard.

GRR

Numeric. The genetic relative-risk of disease, i.e. the relative-risk of disease for individuals who carry at least one copy of the causal variant. Note: When simulating diseases with multiple subtypes GRR must contain one entry for each simulated subtype. See details.

num_affected

Numeric vector. The minimum number of disease-affected relatives required for ascertainment.

ascertain_span

Numeric vector of length 2. The year span of the ascertainment period. This period represents the range of years during which the proband developed disease and the family would have been ascertained for multiple affected relatives.

FamID

Numeric. The family ID to assign to the simulated pedigree.

founder_byears

Numeric vector of length 2. The span of years from which to simulate, uniformly, the birth year for the founder who introduced the rare variant to the pedigree.

stop_year

Numeric. The last year of study. If not supplied, defaults to the current year.

recall_probs

Numeric. The proband's recall probabilities for relatives, see details. If not supplied, the default value of four times kinship coefficient between the proband and the relative is used.

carrier_prob

Numeric. The carrier probability for all causal variants with relative-risk of disease GRR. By default, carrier_prob = 0.002

RVfounder

Logical. Indicates if all pedigrees segregate the rare, causal variant. By default, RVfounder = FALSE See details.

NB_params

Numeric vector of length 2. The size and probability parameters of the negative binomial distribution used to model the number of children per household. By default, NB_params = c(2, 4/7), due to the investigation of Kojima and Kelleher (1962).

fert

Numeric. A constant used to rescale the fertility rate after disease-onset. By default, fert = 1.

first_diagnosis

Numeric. The first year that reliable diagnoses can be obtained regarding disease-affection status. By default, first_diagnosis = NULL so that all diagnoses are considered reliable. See details.

sub_criteria

List. Additional subtype criteria required for ascertainment. The first item in sub_criteria is expected to be a character string indicating a subtype label and the second is a numeric entry indicating the minimum number of relatives affected by the identified subtype for ascertianment. By default, sub_criteria = NULL so that no additional criteria is applied. See details.

Value

A list containing the following data frames:

full_ped

The full pedigree, prior to proband selection and trimming.

ascertained_ped

The ascertained pedigree, with proband selected and trimmed according to proband recall probability. See details.

See Also

sim_ped, trim_ped, sim_life

Details

When RV_founder = TRUE, all simulated pedigrees will segregate a genetic susceptibility variant. In this scenario, we assume that the variant is rare enough that it has been introduced by one founder, and we begin the simulation of the pedigree with this founder. Alternatively, when RV_founder = FALSE we simulate the starting founder's causal variant status with probability carrier_prob. When RV_founder = FALSE pedigrees may not segregate the genetic susceptibility variant. The default selection is RV_founder = FALSE. Additionally, we note that sim_RVpedigree is intended for rare causal variants; users will recieve a warning if carrier_prob > 0.002.

We note that when GRR = 1, pedigrees do not segregate the causal variant regardless of the setting selected for RVfounder. When the causal variant is introduced to the pedigree we transmit it from parent to offspring according to Mendel's laws.

When simulating diseases with multiple subtypes GRR is a numeric list indicating the genetic-relative risk for each subtype specified in the hazard object supplied to hazard_rates. For example, for a disease with two disease subtypes, if we set GRR = c(20, 1) individuals who inherit the causal variant are 20 times more likely than non-carriers to develop the first subtype and as likely as non-carriers to develop the second subtype.

We begin simulating the pedigree by generating the year of birth, uniformly, between the years specified in founder_byears for the starting founder. Next, we simulate this founder's life events using the sim_life function, and censor any events that occur after the study stop_year. Possible life events include: reproduction, disease onset, and death. We continue simulating life events for any offspring, censoring events which occur after the study stop year, until the simulation process terminates. We do not simulate life events for marry-ins, i.e. individuals who mate with either the starting founder or offspring of the starting founder.

We do not model disease remission. Rather, we impose the restriction that individuals may only experience disease onset once, and remain affected from that point on. If disease onset occurs then we apply the hazard rate for death in the affected population.

sim_RVped will only return ascertained pedigrees with at least num_affected affected individuals. That is, if a simulated pedigree does not contain at least num_affected affected individuals sim_RVped will discard the pedigree and simulate another until the condition is met. We note that even for num_affected = 2, sim_RVped can be computationally expensive. To simulate a pedigree with no proband, and without a minimum number of affected members use sim_ped instead of sim_RVped.

When simulating diseases with multiple subtypes, users may wish to apply additional ascertainment criteria using the sub_criteria argument. When supplied, this argument allows users to impose numeric subtype-specific ascertainmet criteria. For example, if and sub_criteria = list("HL", 1) then at least 1 of the num_affected disease-affected relatives must be affected by subtype "HL" for the pedigree to be asceratained. We note that the first entry of sub_criteria, i.e. the subtype label, must match the one of subtype labels in the hazards object supplied to hazard_rates. See examples.

Upon simulating a pedigree with num_affected individuals, sim_RVped chooses a proband from the set of available candidates. Candidates for proband selection must have the following qualities:

1. experienced disease onset between the years specified by ascertain_span,

2. if less than num_affected - 1 individuals experienced disease onset prior to the lower bound of ascertain_span, a proband is chosen from the affected individuals, such that there were at least num_affected affected individuals when the pedigree was ascertained through the proband.

We allow users to specify the first year that reliable diagnoses can be made using the argument first_diagnosis. All subjects who experience disease onset prior to this year are not considered when ascertaining the pedigree for a specific number of disease-affected relatives. By default, first_diagnosis = NULL so that all affected relatives, recalled by the proband, are considered when ascertaining the pedigree.

After the proband is selected, the pedigree is trimmed based on the proband's recall probability of his or her relatives. This option is included to model the possibility that a proband either cannot provide a complete family history or that they explicitly request that certain family members not be contacted. If recall_probs is missing, the default values of four times the kinship coefficient, as defined by Thompson, between the proband and his or her relatives are assumed. This has the effect of retaining all first degree relatives with probability 1, retaining all second degree relatives with probability 0.5, retaining all third degree relatives with probability 0.25, etc. Alternatively, the user may specify a list of length \(l\), such that the first \(l-1\) items represent the respective recall probabilities for relatives of degree \(1, 2, ... , l-1\) and the \(l^{th}\) item represents the recall probability of a relative of degree \(l\) or greater. For example, if recall_probs = c(1, 0.75, 0.5), then all first degree relatives (i.e. parents, siblings, and offspring) are retained with probability 1, all second degree relatives (i.e. grandparents, grandchildren, aunts, uncles, nieces and nephews) are retained with probability 0.75, and all other relatives are retained with probability 0.5. To simulate fully ascertained pedigrees, simply specify recall_probs = c(1).

In the event that a trimmed pedigree fails the num_affected condition, sim_RVped will discard that pedigree and simulate another until the condition is met. For this reason, the values specified for recall_probs affect computation time.

References

Nieuwoudt, Christina and Jones, Samantha J and Brooks-Wilson, Angela and Graham, Jinko (2018). Simulating Pedigrees Ascertained for Multiple Disease-Affected Relatives. Source Code for Biology and Medicine, 13:2.

Ken-Ichi Kojima, Therese M. Kelleher. (1962), Survival of Mutant Genes. The American Naturalist 96, 329-346.

Thompson, E. (2000). Statistical Inference from Genetic Data on Pedigrees. NSF-CBMS Regional Conference Series in Probability and Statistics, 6, I-169.

Examples

# NOT RUN {
#Read in age-specific hazards
data(AgeSpecific_Hazards)

#Simulate pedigree ascertained for multiple affected individuals
set.seed(2)
ex_RVped <- sim_RVped(hazard_rates = hazard(hazardDF = AgeSpecific_Hazards),
                      GRR = 20,
                      RVfounder = TRUE,
                      FamID = 1,
                      founder_byears = c(1900, 1905),
                      ascertain_span = c(1995, 2015),
                      num_affected = 2,
                      stop_year = 2017,
                      recall_probs = c(1, 1, 0))

# Observe: ex_RVped is a list containing two ped objects
summary(ex_RVped)

# The first is the original pedigree prior
# to proband selection and trimming
plot(ex_RVped[[1]])

# The second is the ascertained pedigree which
# has been trimmed based on proband recall
plot(ex_RVped[[2]])
summary(ex_RVped[[2]])


# NOTE: by default, RVfounder = FALSE.
# Under this setting pedigrees segregate a causal
# variant with probability equal to carrier_prob.



#---------------------------------------------------#
# Simulate Pedigrees with Multiple Disease Subtypes #
#---------------------------------------------------#
# Simulating pedigrees with multiple subtypes
# Import subtype-specific hazards rates for
# Hodgkin's lymphoma and non-Hodgkin's lymphoma
data(SubtypeHazards)
head(SubtypeHazards)

my_hazards <- hazard(SubtypeHazards,
                     subtype_ID = c("HL", "NHL"))


# Simulate pedigree ascertained for at least two individuals
# affected by either Hodgkin's lymphoma or non-Hodgkin's lymphoma.
# Set GRR = c(20, 1) so that individuals who carry a causal variant
# are 20 times more likely than non-carriers to develop "HL" but have
# same risk as non-carriers to develop "NHL".
set.seed(45)
ex_RVped <- sim_RVped(hazard_rates = my_hazards,
                      GRR = c(20, 1),
                      RVfounder = TRUE,
                      FamID = 1,
                      founder_byears = c(1900, 1905),
                      ascertain_span = c(1995, 2015),
                      num_affected = 2,
                      stop_year = 2017,
                      recall_probs = c(1, 1, 0))

plot(ex_RVped[[2]], cex = 0.6)

# Note that we can modify the ascertainment criteria so that
# at least 1 of the two disease-affected relatives are affected by
# the "HL" subtype by supplying c("HL", 1) to the sub_criteria
# argument.
set.seed(69)
ex_RVped <- sim_RVped(hazard_rates = my_hazards,
                      GRR = c(20, 1),
                      RVfounder = TRUE,
                      FamID = 1,
                      founder_byears = c(1900, 1905),
                      ascertain_span = c(1995, 2015),
                      num_affected = 2,
                      stop_year = 2017,
                      recall_probs = c(1, 1, 0),
                      sub_criteria = list("HL", 1))

plot(ex_RVped[[2]], cex = 0.6)

# }