Package 'scDDboost'

Title: A compositional model to assess expression changes from single-cell rna-seq data
Description: scDDboost is an R package to analyze changes in the distribution of single-cell expression data between two experimental conditions. Compared to other methods that assess differential expression, scDDboost benefits uniquely from information conveyed by the clustering of cells into cellular subtypes. Through a novel empirical Bayesian formulation it calculates gene-specific posterior probabilities that the marginal expression distribution is the same (or different) between the two conditions. The implementation in scDDboost treats gene-level expression data within each condition as a mixture of negative binomial distributions.
Authors: Xiuyu Ma [cre, aut], Michael A. Newton [ctb]
Maintainer: Xiuyu Ma <[email protected]>
License: GPL (>= 2)
Version: 1.7.0
Built: 2024-06-30 03:43:15 UTC
Source: https://github.com/bioc/scDDboost

Help Index

A compositional model to assess expression changes from single-cell rna-seq data

Description

scDDboost is an R package to analyze changes in the distribution of single-cell expression data between two experimental conditions. Compared to other methods that assess differential expression, scDDboost benefits uniquely from information conveyed by the clustering of cells into cellular subtypes. Through a novel empirical Bayesian formulation it calculates gene-specific posterior probabilities that the marginal expression distribution is the same (or different) between the two conditions. The implementation in scDDboost treats gene-level expression data within each condition as a mixture of negative binomial distributions.

Details

The DESCRIPTION file:

Package: scDDboost
Type: Package
Title: A compositional model to assess expression changes from single-cell rna-seq data
Version: 1.7.0
Date: 2018-10-31
Authors@R: c(person("Xiuyu","Ma",email="[email protected]",role = c("cre", "aut")),person(given = "Michael A.",family = "Newton", email="[email protected]",role = "ctb"))
Description: scDDboost is an R package to analyze changes in the distribution of single-cell expression data between two experimental conditions. Compared to other methods that assess differential expression, scDDboost benefits uniquely from information conveyed by the clustering of cells into cellular subtypes. Through a novel empirical Bayesian formulation it calculates gene-specific posterior probabilities that the marginal expression distribution is the same (or different) between the two conditions. The implementation in scDDboost treats gene-level expression data within each condition as a mixture of negative binomial distributions.
License: GPL (>= 2)
Imports: Rcpp (>= 0.12.11), RcppEigen (>= 0.3.2.9.0),EBSeq, BiocParallel, mclust, SingleCellExperiment, cluster, Oscope, SummarizedExperiment, stats, methods
biocViews: SingleCell, Software, Clustering, Sequencing, GeneExpression, DifferentialExpression, Bayesian
Depends: R (>= 4.2), ggplot2
LinkingTo: Rcpp, RcppEigen, BH
Suggests: knitr, rmarkdown, BiocStyle, testthat
SystemRequirements: c++11
Roxygen: list(wrap=FALSE)
RoxygenNote: 7.1.2
VignetteBuilder: knitr
BugReports: https://github.com/wiscstatman/scDDboost/issues
URL: https://github.com/wiscstatman/scDDboost
Repository: https://bioc.r-universe.dev
RemoteUrl: https://github.com/bioc/scDDboost
RemoteRef: HEAD
RemoteSha: 68a6e0dfce9f3738a2a11c7c486467acb38ecca3
Author: Xiuyu Ma [cre, aut], Michael A. Newton [ctb]
Maintainer: Xiuyu Ma <[email protected]>

Index of help topics: 

EBS                     accelerated empirical bayesian
LL                      likelihood function for hyperparameters
                        estimation
calD                    calculate distance matrix
clusHelper              function to get intra and inter distance for
                        clusters
detK                    determine the number of clusters
extractInfo             extract count matrix from SingleCellExperiment
                        object
gCl                     gene_level cluster
gRef                    generate reference matrix
genRClus                generate random clusterings
getDD                   index of DD genes under FDR control
getSizeofDD             number of DD genes under FDR control
getZ1Z2                 function to get counts of cluster sizes at two
                        conditions
isRef                   check refinement relation between two clusters
lpt1t2                  log likelihood of z1,z2 given t1,t2
lpzgt                   log likelihood of aggregated multinomial counts
                        z given aggregated proportions t
mdd                     posterior of proportion change given mixture
                        double dirichlet prior
pat                     generating partition patterns
pdd                     calculate posterior probabilities of a gene to
                        be differential distributed
pddAggregate            function to aggregate intermediate results and
                        get prob of DD
rwMle                   MLE for random weighting parameter
scDDboost-package       A compositional model to assess expression
                        changes from single-cell rna-seq data
sim_dat                 scDDboost

Package used to score evidence of differential distribution in single-cell RNA-seq data

Author(s)

Xiuyu Ma [cre, aut], Michael A. Newton [ctb]

Maintainer: Xiuyu Ma <[email protected]>

References

https://projecteuclid.org/journals/annals-of-applied-statistics/volume-15/issue-2/A-compositional-model-to-assess-expression-changes-from-single-cell/10.1214/20-AOAS1423.short

See Also

https://github.com/wiscstatman/scDDboost/blob/master/DESCRIPTION

Examples

data(sim_dat)
dat = extractInfo(sim_dat)
data_counts = dat$count_matrix
cd = dat$condition
bp <- BiocParallel::MulticoreParam(4)
D_c = calD(data_counts,bp)
pDD = pdd(data_counts,cd,bp,D_c)

calculate distance matrix

Description

calculate distance matrix

Usage

calD(data, bp)

Arguments

data

transcripts
bp

bioc parallel parameter

Value

distance matrix

Examples

data(sim_dat)
dat <- extractInfo(sim_dat)
data_counts <- dat$count_matrix
bp <- BiocParallel::MulticoreParam(4)
D_c <- calD(data_counts,bp)

function to get intra and inter distance for clusters

Description

function to get intra and inter distance for clusters

Usage

clusHelper(D, i)

Arguments

D

distance matrix
i

number of clusters

Value

vector of intra and inter distance

determine the number of clusters

Description

determine the number of clusters

Usage

detK(D, epi = 1)

Arguments

D

distance matrix
epi

threshold for cutting off

Value

number of clusters

Examples

data(sim_dat)
dat <- extractInfo(sim_dat)
data_counts <- dat$count_matrix
bp <- BiocParallel::MulticoreParam(4)
D_c <- calD(data_counts,bp)
detK(D_c)

accelerated empirical bayesian

Description

accelerated empirical bayesian

Usage

EBS(data, conditions, gclus, sf, iter = 10, hyper, PP, stp1, stp2)

Arguments

data

single cell expression matrix, row as genes column as cells
conditions

partition of cells
gclus

partition of genes
sf

size factors
iter

maximum iteration step of EM
hyper

hyper parameters for beta distributions
PP

pattern of partitions
stp1

step size of hyperparameter alpha (shared by all units) in one step EM
stp2

step size of hyperparameter beta (unit specific) in one step EM

Value

posterior probability of mean expression pattern

extract count matrix from SingleCellExperiment object

Description

extract count matrix from SingleCellExperiment object

Usage

extractInfo(data)

Arguments

data

SingleCellExperiment object

Value

list of count matrix and condition vector

Examples

data(sim_dat) 
dat <- extractInfo(sim_dat)

gene_level cluster

Description

gene_level cluster

Usage

gCl(data, bp)

Arguments

data

transcripts
bp

bioc parallel parameter

Value

return a matrix whose row represent gene specific cluster

generate random clusterings

Description

generate random clusterings

Usage

genRClus(D, a, K)

Arguments

D

distance matrix of cells
a

paramter for weights
K

number of subtypes

Value

random generated clustering of cells

index of DD genes under FDR control

Description

index of DD genes under FDR control

Usage

getDD(pDD, FDR = 0.01)

Arguments

pDD

probability of genes being DD
FDR

fdr to be controlled

Value

index of positive genes

Examples

p_dd <- c(0.01,0.99,0.7,0.5)
getDD(p_dd)

number of DD genes under FDR control

Description

number of DD genes under FDR control

Usage

getSizeofDD(pDD, FDR = 0.01)

Arguments

pDD

estimated probability of being DD
FDR

fdr to be controlled

Value

number of positive genes

Examples

p_dd <- c(0.1,0.99,1,0.05,0.05)
getSizeofDD(p_dd)

function to get counts of cluster sizes at two conditions

Description

function to get counts of cluster sizes at two conditions

Usage

getZ1Z2(ccl, cd)

Arguments

ccl

clustering label
cd

condition label

Value

return list of counts

generate reference matrix

Description

generate reference matrix

Usage

gRef(Posp)

Arguments

Posp

possible partition of data

Value

return a matrix indicate the refinement relation between different partitions.

check refinement relation between two clusters

Description

check refinement relation between two clusters

Usage

isRef(x, y)

Arguments

x

a cluster
y

a cluster

Value

whether x refines y

likelihood function for hyperparameters estimation

Description

likelihood function for hyperparameters estimation

Usage

LL(param, x, d0)

Arguments

param

parameters to be determined by MLE
x

distance matrix of cells
d0

rate parameter of prior of 1 / true distance

Value

return hyperparameteres a.

log likelihood of z1,z2 given t1,t2

Description

log likelihood of z1,z2 given t1,t2

Usage

lpt1t2(z1, z2, pp, alpha1, alpha2)

Arguments

z1

counts of each group in condition 1
z2

counts of each group in condition 2
pp

a partition
alpha1

parameter of double dirichlet prior
alpha2

parameter of double dirichlet prior

Value

log likelihood of z1,z2 given t1,t2

log likelihood of aggregated multinomial counts z given aggregated proportions t

Description

log likelihood of aggregated multinomial counts z given aggregated proportions t

Usage

lpzgt(z, pp, alpha)

Arguments

z

counts of each group in one condition
pp

a partition
alpha

parameter of double dirichlet prior

Value

log likelihood of aggregated multinomial counts z given aggregated proportions t

posterior of proportion change given mixture double dirichlet prior

Description

posterior of proportion change given mixture double dirichlet prior

Usage

mdd(z1, z2, pat, alpha1, alpha2)

Arguments

z1

counts of each group in condition 1
z2

counts of each group in condition 2
pat

partition patterns
alpha1

parameter of double dirichlet prior
alpha2

parameter of double dirichlet prior

Value

posterior of proportion change

generating partition patterns

Description

generating partition patterns

Usage

pat(K)

Arguments

K

number of elements

Value

all possible partition of K elements

Examples

pat(3)

calculate posterior probabilities of a gene to be differential distributed

Description

calculate posterior probabilities of a gene to be differential distributed

Usage

pdd(
  data,
  cd,
  bp,
  D,
  random = TRUE,
  norm = TRUE,
  epi = 1,
  Upper = 1000,
  nrandom = 50,
  iter = 20,
  reltol = 0.001,
  stp1 = 1e-06,
  stp2 = 0.01,
  K = 0
)

Arguments

data

normalized preprocessed transcripts
cd

conditions label
bp

bioc parallel parameter
D

distance matrix of cells or cluster of cells or a given clustering
random

boolean indicator of whether randomzation has been been implemented on distance matrix
norm

boolean indicator of whether the input expression data is normalized
epi

tol for change of validity score in determining number of clusters
Upper

bound for hyper parameters optimization
nrandom

number of random generated distance matrix
iter

max number of iterations for EM
reltol

relative tolerance for optim on weighting paramters
stp1

step size of hyperparameter alpha (shared by all units) in one step EM
stp2

step size of hyperparameter beta (unit specific) in one step EM
K

number of subtypes, could be user specified or determined internally(set to 0)

Value

posterior probabilities of a gene to be differential distributed

Examples

data(sim_dat)
dat <- extractInfo(sim_dat)
data_counts <- dat$count_matrix
cd <- dat$condition
bp <- BiocParallel::MulticoreParam(4)
D_c <- calD(data_counts,bp)
pDD <- pdd(data_counts,cd,bp,D_c)

function to aggregate intermediate results and get prob of DD

Description

function to aggregate intermediate results and get prob of DD

Usage

pddAggregate(z1, z2, Posp, DE, K, REF)

Arguments

z1

counts of cluster sizes in condition 1
z2

counts of cluster sizes in condition 2
Posp

partition of cells
DE

posterior probabilities of DE patterns
K

number of clusters
REF

reference matrix indicating relation of nested partitions

Value

return vector of prob of DD

MLE for random weighting parameter

Description

MLE for random weighting parameter

Usage

rwMle(D, reltol)

Arguments

D

distance matrix of cells
reltol

tolerance of convergence

Value

MLE of random weighting parameter

scDDboost

Description

simulated data for demonstration, data are mixture negative binomial distributed

Usage

data(sim_dat)

Format

An object of class "list".

Examples

data(sim_dat)