| Title: | Long Non-Coding RNA Differential Expression Analysis |
| Version: | 1.0.0 |
| Description: | We developed an approach to detect differential expression features in long non-coding RNA low counts, using generalized linear model with zero-inflated exponential quasi likelihood ratio test. Methods implemented in this package are described in Li (2019) <doi:10.1186/s12864-019-5926-4>. |
| Depends: | R (≥ 3.5.0) |
| License: | GPL-2 | GPL-3 [expanded from: GPL (≥ 2)] |
| Encoding: | UTF-8 |
| LazyData: | true |
| RoxygenNote: | 7.0.0 |
| NeedsCompilation: | no |
| Packaged: | 2020-01-10 03:36:50 UTC; liq |
| Author: | Qian Li [aut, cre] |
| Maintainer: | Qian Li <qian.li10000@gmail.com> |
| Repository: | CRAN |
| Date/Publication: | 2020-01-13 22:20:08 UTC |
Likelihood ratio test based on ZIQML.fit()
Description
ZIQML.LRT returns the likelihood ratio test statistics and p-value based on the object returned by ZIQML.fit().
Usage
LRT(ZIQML.fit, coef = NULL)
Arguments
ZIQML.fit |
Object returned by ZIQML.fit() |
coef |
An integer or vector indicating the coefficient(s) in design matrix to be tested. coef=1 is the intercept (i.e. baseline group effect), and should not be tested. |
Value
LRT.stat |
Likelihood ratio test statistics. |
LRT.pvalue |
Likelihood ratio test p-value. |
Examples
data('hnsc.edata','design')
# 'hnsc.edata' contains FPKM of 1132 lncRNA genes and 80 samples.
# 'design' is the design matrix of tissue type (tumor vs normal).
# Fit GLM by ZIQML.fit for the first 100 genes
fit.log=ZIQML.fit(edata=hnsc.edata[1:100,],design.matrix=design)
# Likelihood ratio test to compare tumor vs normal in gene expression level.
LRT.results=LRT(fit.log,coef=2)
Group and covariate effects on lncRNA counts by Generalized Linear Model
Description
ZIQML.fit estimates the group effect on gene expression using zero-inflated exponential quasi likelihood.
Usage
ZIQML.fit(edata, design.matrix, link = "log")
Arguments
edata |
Normalized counts matrix with genes in rows and samples in columns. |
design.matrix |
Design matrix for groups and covariates, generated by model.matrix(). |
link |
Link function for the generalized linear model and likelihood function,either 'log' or 'identity'. The default is 'log'. |
Value
Estimates |
Estimated group effect on gene expression by zero-inflated exponential quasi maximum likelihood (ZIQML) estimator. |
logLikelihood |
The value of zero-inflated quasi likelihood. |
edata |
lncRNA counts or expression matrix. |
design.matrix |
The design matrix of groups and covariates. |
link |
The specified link function. |
Examples
data('hnsc.edata','design')
# 'hnsc.edata' contains FPKM of 1000 lncRNA genes and 80 samples
# 'design' is the design matrix for tissue and batch.
# For the first 100 genes
# Fit GLM by ZIQML with logarithmic link function
fit.log=ZIQML.fit(edata=hnsc.edata[1:100,],design.matrix=design,link='log')
# Fit GLM by ZIQML with identity link function
fit.identity=ZIQML.fit(edata=hnsc.edata[1:100,],design.matrix=design,link='identity')
Batch information for samples in hnsc.edata.
Description
Batch information for samples in hnsc.edata.
Usage
cov
Format
A matrix of covariate(s) in columns.
Design matrix for samples in hnsc.edata.
Description
Design matrix for samples in hnsc.edata.
Usage
design
Format
A model matrix with 80 rows (i.e. samples) and 3 columns of tissue type and batch.
lncRNA Fragments Per Killobase per Million (FPKM) in a head and neck squamous cell carcinomas (hnsc) study.
Description
lncRNA Fragments Per Killobase per Million (FPKM) in a head and neck squamous cell carcinomas (hnsc) study.
Usage
hnsc.edata
Format
A data frame of lncRNA FPKM with 1000 rows (i.e. genes) and 80 columns (i.e. samples ).
lncRNA Differential Expression (DE) analysis
Description
lncDIFF returns DE analysis results based on lncRNA counts and grouping variables.
Usage
lncDIFF(
edata,
group,
covariate = NULL,
link.function = "log",
CompareGroups = NULL,
simulated.pvalue = FALSE,
permutation = 100
)
Arguments
edata |
Normalized counts matrix with genes in rows and samples in columns. |
group |
Primary factor of interest in DE analysis, e.g., treatment groups, tissue types, other phenotypes. |
covariate |
Other variables (or covariates) associated with expression level. Input must be a matrix or data frame with each column being a covariate matching to |
link.function |
Link function for the generalized linear model, either 'log' or 'identity', default as 'log'. |
CompareGroups |
Labels of treatment groups or phenotypes of interest to be compared in DE analysis. Input must be a vector of |
simulated.pvalue |
If empirical p-values are computed, simulated.pvalue=TRUE. The default is FALSE. |
permutation |
The number of permutations used in simulating pvalues. The default value is 100. |
Value
DE.results |
Likelihood ratio test results with test statistics, p-value, FDR, DE genes, groupwise mean expression, fold change (if two groups are compared). If simulated.pvalue=TRUE, test.results also includes simulated p-value and FDR. |
full.model.fit |
Generalized linear model with zero-inflated Exponential likelihood function, estimating group effect compared to a reference group. |
References
Li, Q., Yu, X., Chaudhary, R. et al.'lncDIFF: a novel quasi-likelihood method for differential expression analysis of non-coding RNA'. BMC Genomics (2019) 20: 539.
Examples
data('hnsc.edata','tissue','cov')
# DE analysis comparing two groups (normal vs tumor) for 100 genes
result=lncDIFF(edata=hnsc.edata[1:100,],group=tissue,covariate=cov)
# Recommend at least 50 permutations if simulated.pvalue=TRUE
Tissue type for samples in hnsc.edata.
Description
Tissue type for samples in hnsc.edata.
Usage
tissue
Format
A character vector of tissue type.