twas: An EM model fit jointly to transcriptome and trait data
In dkulp2/TWAS: Transcriptome-Wide Association Study Analysis

Description Usage Arguments Value See Also Examples

twas returns a model fit from two data sets from different subjects: one set of transcriptome data and the other of trait data. twas iteratively fits the trait data to the subjects in the transcriptome data and vice versa, until convergence.

twas(transcriptome, traits, tx.genotypes, trait.genotypes, gene, trait.names,
  markers, LD.reduction = FALSE, tx.id.col = 1, trait.id.col = 1, tx.max,
  rounds = 1000, epsilon = 1e-06, tiny.weights = 1e-16,
  show.progress = TRUE)

`transcriptome`	a data frame of expression levels for one or more genes. Each row is a subject. There is one ID column and the remaining columns are genes. Expression values are non-negative integers.
`traits`	a data frame of one or more traits. Each row is a subject. There is one ID column and the remaining columns are traits. Trait values are numeric.
`tx.genotypes`	a data frame of genotypes for the transcriptome subjects. Each row is a subject. There is one ID column with the same name as used in `transcriptome` and the remaining columns are markers. Genotype values are non-negative integers. Subjects with any NA genotypes are ignored.
`trait.genotypes`	a data frame of genotypes for the trait subjects. Each row is a subject. There is one ID column with the same name as used in `traits` and the remaining columns are markers. Genotype values are non-negative integers. Subjects with any NA genotypes are ignored.
`gene`	a character identifying the column in `transcriptome` of the gene of interest. Default is the first column in `transcriptome` excluding the subject ID.
`trait.names`	a character vector of one or more column names in `traits` to be used in analysis. Defaults to all columns in `traits` except the subject ID.
`markers`	a character vector of the names of the markers to use for analysis. Defaults to the shared marker names in `tx.genotypes` and `trait.genotypes`.
`LD.reduction`	a logical scalar or a function to perform LD reduction. If true, then the SNPRelate package will be used on the genotypes of the two data sets for the specified `markers` to generate a subset of `markers`, which will be used in the analysis. If a function is supplied, then it must accept a matrix of columns of genotypes and return a vector of column names of the SNPs to keep. Default is false.
`tx.id.col`	the index or name of the column containing the subject ID in the `transcriptome` and `tx_genotype` data frames. Defaults to the first column in `transcriptome`.
`trait.id.col`	the index or name of the column containing the subject ID in the `traits` and `trait_genotype` data frames. Defaults to the first column in `traits`.
`tx.max`	an integer representing the maximum possible transcript count value to model. The default is the maximum observed value in `transcriptome`. Large values increase the modeling time and some observed counts are believed to be outliers. If not set and the observed count is greater than 1000, then a warning is issues.
`rounds`	an integer for the maximum number of EM steps. Default is 1000.
`epsilon`	a numeric for the EM stopping condition. If the absolute difference is less than `epsilon` then modeling halts. Default is 1e-6.
`tiny.weights`	a numeric. Any weights that are less than or equal to `tiny.weights` are excluded from modeling. Default is 1e-16.
`show.progress`	a logical indicating whether to display progress during EM iterations. If the `progress` package is installed the it will be used to display a progress bar. Otherwise, text will be displayed. Default is true.

A list object containing the fit expression levels and trait values for all subjects in a single data frame (data), each model parameter (gamma, beta, eta), and the variance and p-value of the test statistic for each parameter.

twas.tx, twas.traits and twas.gene.markers for examples of the necessary data formats.

twas(twas.tx$transcriptome, twas.traits$traits,
     twas.tx$genotypes, twas.traits$genotypes,
     gene="LRRC16A", trait.names="logIL6",
     markers=twas.gene.markers$LRRC16A)