predict.baymap {BayMAP}R Documentation

Predict method for BayMAP results

Description

Predictions if T-to-C substitution positions are PAR-CLIP induced substitutions or not. Results of several PAR-CLIP experiments can be combined for a combined prediction.

Usage

## S3 method for class 'baymap'
predict(object, data, chr = "chr", pos = "pos",
            count = "count", coverage = "coverage",
            mutation = "mutation", mutation.type = "TC", 
            covariates = NULL,
            dist = c("truncated", "binomial"),
            ran = FALSE, cluster = NULL, 
            print.i = 100, thin = NULL, burn.in = 0, ...)

Arguments

object

either a baymap object or a list of baymap objects if several PAR-CLIP experiments should be analyzed jointly. If a list with baymap objects is read in, the class "baymap" should be assigned to the list prior to the analysis by class.

data

either a data frame with at least the count for mutations per genomic position, the number of reads/coverage and the mutation type (e.g., T-to-C) or a list of data frames.

chr

the name of the variable that contains the chromosome information.

pos

the name of the variable that contains the position information.

count

the name of the variable that counts the number of mutations.

coverage

the name of the variable that contains the number of reads.

mutation

the name of the variable that contains the different types of mutations.

mutation.type

the name of the mutation type that is induced by the PAR-CLIP method.

covariates

a vector containing the names for the covariates for the regression model, e.g., c("tpUTR", "cds", "fpUTR"). Intercept is automatically added as first variable.

dist

the distribution for the number of mutations. Possible entries are "truncated" (default) and "binomial.

ran

a logical value indicating if neighborhood dependencies should be included via a random effect (default) or not.

cluster

the name of the varialbe indicating to which cluster a position belongs. Only necessary if ran is set to TRUE.

print.i

a positive integer indicating if every ith iteration step should be printed.

thin

an additional thinning rate that should be applied on the baymap outcome. Must be a positive integer.

burn.in

length of burn in, i.e. number of iterations to discard at the beginning.

...

further arguments for predict.

Value

If a single PAR-CLIP data set is analyzed, the returned object is a data frame including the prior odds, the bayes factor and the posterior odds. If several PAR-CLIP data sets are analyzed, the returned object is a list combined predictions as well as separate predictions for each dara set. Posterior odds greater than one means that the probability of having a method-incuced substitution position given the data is larger than 0.5.

Note

Predictions of the separate analyzis are made for all entries of the included data set even for substitution types other than T-to-C.

Author(s)

Eva-Maria Huessler, eva-maria.huessler@uni-duesseldorf.de

References

Huessler, E., Schaefer, M. Schwender, H., Landgraf, P. (2019): BayMAP: A Bayesian hierarchical model for the analysis of PAR-CLIP data. Bioinformatics, 35(12), 1992-2000.

See Also

baymap

Examples

## Not run: 
data(data_test)
res <- baymap(data = data_test,
      inits.mu = c(0.05, 0.85, 0.2), n.iter = 4500)
data_new <- predict(res, data_test, burn.in = 3000)
  
## End(Not run)


[Package BayMAP version 2.0.1 Index]