generateTS: Generate community time series

In hallucigenia-sparsa/seqtime: Time Series Analysis of Sequencing Data

Generate community time series

Description

Wrapper function for time series generation. The result folder has a specific structure that is created if it does not yet exist. It has two sub-folders, one named timeseries and the other settings. In both sub-folders, there are folders having as name the experiment identifier followed by timeseries or settings. The interaction matrix, immigration and extinction rates, carrying capacities and other randomly generated input are also stored in settings. The idea is that the time series can be re-created from the information stored in the settings folder (though not exactly, since the SOI, Ricker and neutral models have random components). Furthermore, a plot of the time series and (if required by the model) the interaction matrix is stored in settings, for control. Properties of the interaction matrix are also stored there.

Usage

generateTS(N = 100, I = 1500, tend = 100, initAbundMode = 5,
  c = 0.05, clique.size = 5, sigma = 0.05, theta = 0.002,
  Atype = "klemm", Atweak = "tweak", d = -1, PEP = 30, m = 0.02,
  deathrate = 10, algorithm = "dm", interval = 1,
  output.folder = "", output.expId = "", input.folder = "",
  input.expId = "", read.A = FALSE, read.y = FALSE, read.K = FALSE,
  read.extinct = FALSE, read.ts = FALSE, maxIter = 10)

Arguments

N	species number (for hubbell, refers to the species number in the metacommunity)
I	individual number (only for SOI and hubbell)
tend	the number of time points to generate
initAbundMode	the mode in which initial abundances are generated (described in function generateAbundances)
c	connectivity of interaction matrix (only for SOI, ricker and gLV)
clique.size	parameter for interaction matrix generation (only for SOI, ricker and glv and Atype mode klemm)
sigma	noise term (only for ricker)
theta	Dirichlet-Multinomial overdispersion parameter (only for dm)
Atype	method to generate the interaction matrix (described in function generateA)
Atweak	tweaking method for instable interaction matrix, either tweak (convert positive into negative arcs until the matrix is stable or one third of the positive arcs have been converted) or schur (apply Schur decomposition, does not guarantee a stable matrix)
d	diagonal value of the interaction matrix
PEP	positive edge percentage of interaction matrix (only for SOI, ricker and gLV)
m	immigration rate (only for hubbell)
deathrate	number of individuals that are replaced at each time step (only for hubbell)
algorithm	how time series should be generated (can be hubbell, soi, ricker, glv, dm, davida or davidb)
interval	sampling frequency (if 1, each sample is taken, if 2, every second sample is taken etc.)
output.folder	(optional) path to result folder, if specified, input parameter and time series are exported to the output folder
output.expId	(optional) the identifier of the time series generation experiment
input.folder	(optional) the folder from which previously generated parameters can be read
input.expId	(optional) the identifier of a previous experiment whose results will be read
read.A	read the interaction matrix from the given previous experiment
read.y	read the initial abundances/SOI immigration rates from the given previous experiment (for hubbell, initial abundances refer to the metacommunity)
read.K	read the carrying capacities/growth rates from the given previous experiment
read.extinct	read the SOI extinction rates from the given previous experiment
read.ts	read the previously generated time series from the given previous experiment
maxIter	maximal iteration number for finding a stable interaction matrix

Details

On the random generation of parameter values: Extinction rates are sampled from the uniform distribution with values between 0 and 1. Carrying capacities/growth rates are sampled from the uniform distribution with values between 0 and 0.5. On the implementation of the neutral model: simHubbell is used to generate neutral model time series. The 1000 first time points of the simulation are always skipped. The species number of the local community is set to 1/10th of the species number in the metacommunity. The local community is initialized with even abundances. On the interaction matrix: When the positive edge percentage is modified, symmetry is not enforced. The maximal absolute interaction strength in the interaction matrix is set to 1. Interaction matrix stability is tested with method "ricker", using an explosion bound of 10^8. For interaction stabilizing method tweak, conversion of positive into negative arcs is stopped when more than a third of the positive arcs were converted already. On the stool data: The stool data are processed as follows: for stool data set B, the last time point is omitted. The data are rarefied to 10000 reads per sample, omitting samples with insufficient read numbers. Then, using metadata on sampling days, values for missing days are interpolated with method stineman. The top-abundant 100 taxa (according to their sum across samples) are retained and small negative values introduced by the interpolation are set to zero. Other settings: For initial abundance mode 6, k is set to 0.5. The simulation step size is always 1 (larger intervals are introduced after simulation). Dirichet-Multinomial data are generated with a count number of 1000 per sample.

Value

list containing the generated interaction matrix, initial abundances (they double as immigration probabilities for soi and metapopulation proportions for hubbell), carrying capacities (they double as growth rates for glv), extinction probabilities, time series and settings

hallucigenia-sparsa/seqtime documentation built on Jan. 9, 2023, 11:53 p.m.