Standard regression functions in R enabled for parallel processing over large data-frames
Kevin Blighe
2019-11-05
- 1 Introduction
- 2 Installation
- 3 Quick start
- 4 Advanced features
- 4.1 Speed up processing
- 4.1.1 ~2000 tests; blocksize, 500; cores, 2; nestedParallel, TRUE
- 4.1.2 ~2000 tests; blocksize, 500; cores, 2; nestedParallel, FALSE
- 4.1.3 ~40000 tests; blocksize, 2000; cores, 2; nestedParallel, TRUE
- 4.1.4 ~40000 tests; blocksize, 2000; cores, 2; nestedParallel, FALSE
- 4.1.5 ~40000 tests; blocksize, 5000; cores, 3; nestedParallel, TRUE
- 4.2 Modify confidence intervals
- 4.3 Remove some terms from output / include the intercept
- 4.1 Speed up processing
- 5 Acknowledgments
- 6 Session info
- 7 References
1 Introduction
In many analyses, a large amount of variables have to be tested independently against the trait/endpoint of interest, and also adjusted for covariates and confounding factors at the same time. The major bottleneck in these is the amount of time that it takes to complete these analyses.
With RegParallel, a large number of tests can be performed simultaneously. On a 12-core system, 144 variables can be tested simultaneously, with 1000s of variables processed in a matter of seconds via ‘nested’ parallel processing.
Works for logistic regression, linear regression, conditional logistic regression, Cox proportional hazards and survival models, Bayesian logistic regression, and negative binomial regression.
2 Installation
2.1 1. Download the package from Bioconductor
if (!requireNamespace('BiocManager', quietly = TRUE))
install.packages('BiocManager')
BiocManager::install('RegParallel')Note: to install development version:
2.2 2. Load the package into R session
3 Quick start
For this quick start, we will follow the tutorial (from Section 3.1) of RNA-seq workflow: gene-level exploratory analysis and differential expression. Specifically, we will load the ‘airway’ data, where different airway smooth muscle cells were treated with dexamethasone.
Normalise the raw counts in DESeq2 and produce regularised log counts:
library(DESeq2)
dds <- DESeqDataSet(airway, design = ~ dex + cell)
dds <- DESeq(dds, betaPrior = FALSE)
rlogcounts <- assay(rlog(dds, blind = FALSE))
rlogdata <- data.frame(colData(airway), t(rlogcounts))3.1 Perform the most basic logistic regression analysis
Here, we fit a binomial logistic regression model to the data via glmParallel, with dexamethasone as the dependent variable.
res1 <- RegParallel(
data = rlogdata[ ,1:3000],
formula = 'dex ~ [*]',
FUN = function(formula, data)
glm(formula = formula,
data = data,
family = binomial(link = 'logit')),
FUNtype = 'glm',
variables = colnames(rlogdata)[10:3000])
res1[order(res1$P, decreasing=FALSE),]## Variable Term Beta StandardError
## 1: ENSG00000095464 ENSG00000095464 43.27934 2.593463e+01
## 2: ENSG00000071859 ENSG00000071859 12.96251 7.890287e+00
## 3: ENSG00000069812 ENSG00000069812 -44.37139 2.704021e+01
## 4: ENSG00000072415 ENSG00000072415 -19.90841 1.227527e+01
## 5: ENSG00000073921 ENSG00000073921 14.59470 8.999831e+00
## ---
## 2817: ENSG00000068831 ENSG00000068831 110.84893 2.729072e+05
## 2818: ENSG00000069020 ENSG00000069020 -186.45744 4.603615e+05
## 2819: ENSG00000083642 ENSG00000083642 -789.55666 1.951104e+06
## 2820: ENSG00000104331 ENSG00000104331 394.14700 9.749138e+05
## 2821: ENSG00000083097 ENSG00000083097 -217.48873 5.398191e+05
## Z P OR ORlower ORupper
## 1: 1.6687854476 0.09515991 6.251402e+18 5.252646e-04 7.440065e+40
## 2: 1.6428433092 0.10041536 4.261323e+05 8.190681e-02 2.217017e+12
## 3: -1.6409412536 0.10080961 5.367228e-20 5.165170e-43 5.577191e+03
## 4: -1.6218306224 0.10483962 2.258841e-09 8.038113e-20 6.347711e+01
## 5: 1.6216635641 0.10487540 2.179701e+06 4.761313e-02 9.978541e+13
## ---
## 2817: 0.0004061781 0.99967592 1.383811e+48 0.000000e+00 NA
## 2818: -0.0004050239 0.99967684 1.053326e-81 0.000000e+00 NA
## 2819: -0.0004046717 0.99967712 0.000000e+00 0.000000e+00 NA
## 2820: 0.0004042891 0.99967742 1.499223e+171 0.000000e+00 NA
## 2821: -0.0004028919 0.99967854 3.514357e-95 0.000000e+00 NA3.2 Perform a basic linear regression
Here, we will perform the linear regression using both glmParallel and lmParallel. We will appreciate that a linear regression is the same using either function with the default settings.
Regularised log counts from our DESeq2 data will be used.
rlogdata <- rlogdata[ ,1:2000]
res2 <- RegParallel(
data = rlogdata,
formula = '[*] ~ cell',
FUN = function(formula, data)
glm(formula = formula,
data = data,
method = 'glm.fit'),
FUNtype = 'glm',
variables = colnames(rlogdata)[10:ncol(rlogdata)],
p.adjust = "none")
res3 <- RegParallel(
data = rlogdata,
formula = '[*] ~ cell',
FUN = function(formula, data)
lm(formula = formula,
data = data),
FUNtype = 'lm',
variables = colnames(rlogdata)[10:ncol(rlogdata)],
p.adjust = "none")
subset(res2, P<0.05)## Variable Term Beta StandardError t
## 1: ENSG00000001461 cellN061011 -0.46859875 0.10526111 -4.451775
## 2: ENSG00000001461 cellN080611 -0.84020922 0.10526111 -7.982143
## 3: ENSG00000001461 cellN61311 -0.87778101 0.10526111 -8.339082
## 4: ENSG00000001561 cellN080611 -1.71802758 0.13649920 -12.586357
## 5: ENSG00000001561 cellN61311 -1.05328889 0.13649920 -7.716448
## ---
## 519: ENSG00000092108 cellN061011 -0.12721659 0.01564082 -8.133625
## 520: ENSG00000092108 cellN61311 -0.12451203 0.01564082 -7.960708
## 521: ENSG00000092148 cellN080611 -0.34988071 0.10313461 -3.392467
## 522: ENSG00000092200 cellN080611 0.05906656 0.01521063 3.883241
## 523: ENSG00000092208 cellN080611 -0.28587683 0.08506716 -3.360602
## P OR ORlower ORupper
## 1: 0.0112313246 0.6258787 0.5092039 0.7692873
## 2: 0.0013351958 0.4316202 0.3511586 0.5305181
## 3: 0.0011301853 0.4157043 0.3382098 0.5109554
## 4: 0.0002293465 0.1794197 0.1373036 0.2344544
## 5: 0.0015182960 0.3487887 0.2669157 0.4557753
## ---
## 519: 0.0012429963 0.8805429 0.8539591 0.9079544
## 520: 0.0013489163 0.8829276 0.8562718 0.9104133
## 521: 0.0274674209 0.7047722 0.5757851 0.8626549
## 522: 0.0177922771 1.0608458 1.0296864 1.0929482
## 523: 0.0282890537 0.7513552 0.6359690 0.8876762## Variable Term Beta StandardError t
## 1: ENSG00000001461 cellN061011 -0.46859875 0.10526111 -4.451775
## 2: ENSG00000001461 cellN080611 -0.84020922 0.10526111 -7.982143
## 3: ENSG00000001461 cellN61311 -0.87778101 0.10526111 -8.339082
## 4: ENSG00000001561 cellN080611 -1.71802758 0.13649920 -12.586357
## 5: ENSG00000001561 cellN61311 -1.05328889 0.13649920 -7.716448
## ---
## 519: ENSG00000092108 cellN061011 -0.12721659 0.01564082 -8.133625
## 520: ENSG00000092108 cellN61311 -0.12451203 0.01564082 -7.960708
## 521: ENSG00000092148 cellN080611 -0.34988071 0.10313461 -3.392467
## 522: ENSG00000092200 cellN080611 0.05906656 0.01521063 3.883241
## 523: ENSG00000092208 cellN080611 -0.28587683 0.08506716 -3.360602
## P OR ORlower ORupper
## 1: 0.0112313246 0.6258787 0.5092039 0.7692873
## 2: 0.0013351958 0.4316202 0.3511586 0.5305181
## 3: 0.0011301853 0.4157043 0.3382098 0.5109554
## 4: 0.0002293465 0.1794197 0.1373036 0.2344544
## 5: 0.0015182960 0.3487887 0.2669157 0.4557753
## ---
## 519: 0.0012429963 0.8805429 0.8539591 0.9079544
## 520: 0.0013489163 0.8829276 0.8562718 0.9104133
## 521: 0.0274674209 0.7047722 0.5757851 0.8626549
## 522: 0.0177922771 1.0608458 1.0296864 1.0929482
## 523: 0.0282890537 0.7513552 0.6359690 0.88767623.3 Perform the most basic negative binomial logistic regression analysis
Here, we will utilise normalised, unlogged counts from DESeq2. Unlogged counts in RNA-seq naturally follow a negative binomial / Poisson-like distribution. glm.nbParallel will be used.
nbcounts <- round(counts(dds, normalized = TRUE), 0)
nbdata <- data.frame(colData(airway), t(nbcounts))
res4 <- RegParallel(
data = nbdata[ ,1:3000],
formula = '[*] ~ dex',
FUN = function(formula, data)
glm.nb(formula = formula,
data = data),
FUNtype = 'glm.nb',
variables = colnames(nbdata)[10:3000],
p.adjust = "fdr")
res4[order(res4$Theta, decreasing = TRUE),]## Variable Term Beta StandardError Z
## 1: ENSG00000102226 dextrt -0.139286172 0.01862107 -7.4800293
## 2: ENSG00000102910 dextrt -0.030278805 0.01582026 -1.9139254
## 3: ENSG00000063601 dextrt -0.002822867 0.02656550 -0.1062607
## 4: ENSG00000083642 dextrt -0.128217080 0.02412001 -5.3157976
## 5: ENSG00000023041 dextrt 0.134325359 0.02728332 4.9233512
## ---
## 2817: ENSG00000006128 dextrt 1.386294361 1.73450917 0.7992430
## 2818: ENSG00000029559 dextrt -1.386294361 1.73450917 -0.7992430
## 2819: ENSG00000101197 dextrt -1.386294361 1.73450917 -0.7992430
## 2820: ENSG00000102109 dextrt -0.318453731 1.39587712 -0.2281388
## 2821: ENSG00000069122 dextrt 0.552068582 1.61318296 0.3422232
## P Theta SEtheta 2xLogLik Dispersion
## 1: 7.430607e-14 2.020721e+08 1.524455e+10 -73.91316 1
## 2: 5.562969e-02 1.122474e+08 6.246419e+09 -78.06984 1
## 3: 9.153755e-01 6.328822e+07 4.095363e+09 -68.80847 1
## 4: 1.061911e-07 3.113562e+07 1.579693e+09 -72.68055 1
## 5: 8.507458e-07 3.000904e+07 1.590205e+09 -70.09368 1
## ---
## 2817: 4.241495e-01 2.843297e-01 3.514737e-01 -15.24150 1
## 2818: 4.241495e-01 2.843297e-01 3.514737e-01 -15.24150 1
## 2819: 4.241495e-01 2.843297e-01 3.514737e-01 -15.24150 1
## 2820: 8.195383e-01 2.664530e-01 1.500629e-01 -41.63690 1
## 2821: 7.321830e-01 1.984580e-01 1.233914e-01 -37.96063 1
## OR ORlower ORupper P.adjust
## 1: 0.8699790 0.83880015 0.9023169 2.624401e-12
## 2: 0.9701750 0.94055425 1.0007286 1.401173e-01
## 3: 0.9971811 0.94658900 1.0504772 9.941050e-01
## 4: 0.8796624 0.83904459 0.9222465 1.361660e-06
## 5: 1.1437649 1.08420938 1.2065918 8.333173e-06
## ---
## 2817: 4.0000000 0.13354976 119.8055313 6.161820e-01
## 2818: 0.2500000 0.00834686 7.4878457 6.161820e-01
## 2819: 0.2500000 0.00834686 7.4878457 6.161820e-01
## 2820: 0.7272727 0.04715465 11.2168280 9.303492e-01
## 2821: 1.7368421 0.07355573 41.0113595 8.703426e-013.4 Survival analysis via Cox Proportional Hazards regression
For this example, we will load breast cancer gene expression data with recurrence free survival (RFS) from Gene Expression Profiling in Breast Cancer: Understanding the Molecular Basis of Histologic Grade To Improve Prognosis. Specifically, we will encode each gene’s expression into Low|Mid|High based on Z-scores and compare these against RFS while adjusting for tumour grade in a Cox Proportional Hazards model.
First, let’s read in and prepare the data:
library(Biobase)
library(GEOquery)
# load series and platform data from GEO
gset <- getGEO('GSE2990', GSEMatrix =TRUE, getGPL=FALSE)
x <- exprs(gset[[1]])
# remove Affymetrix control probes
x <- x[-grep('^AFFX', rownames(x)),]
# transform the expression data to Z scores
x <- t(scale(t(x)))
# extract information of interest from the phenotype data (pdata)
idx <- which(colnames(pData(gset[[1]])) %in%
c('age:ch1', 'distant rfs:ch1', 'er:ch1',
'ggi:ch1', 'grade:ch1', 'node:ch1',
'size:ch1', 'time rfs:ch1'))
metadata <- data.frame(pData(gset[[1]])[,idx],
row.names = rownames(pData(gset[[1]])))
# remove samples from the pdata that have any NA value
discard <- apply(metadata, 1, function(x) any(is.na(x)))
metadata <- metadata[!discard,]
# filter the Z-scores expression data to match the samples in our pdata
x <- x[,which(colnames(x) %in% rownames(metadata))]
# check that sample names match exactly between pdata and Z-scores
all((colnames(x) == rownames(metadata)) == TRUE)## [1] TRUE # create a merged pdata and Z-scores object
coxdata <- data.frame(metadata, t(x))
# tidy column names
colnames(coxdata)[1:8] <- c('Age', 'Distant.RFS', 'ER',
'GGI', 'Grade', 'Node',
'Size', 'Time.RFS')
# prepare certain phenotypes
coxdata$Age <- as.numeric(gsub('^KJ', '', coxdata$Age))
coxdata$Distant.RFS <- as.numeric(coxdata$Distant.RFS)
coxdata$ER <- factor(coxdata$ER, levels = c(0, 1))
coxdata$Grade <- factor(coxdata$Grade, levels = c(1, 2, 3))
coxdata$Time.RFS <- as.numeric(gsub('^KJX|^KJ', '', coxdata$Time.RFS))With the data prepared, we can now apply a Cox Proportional Hazards model independently for each probe in the dataset against RFS.
In this we also increase the default blocksize to 2000 in order to speed up the analysis.
library(survival)
res5 <- RegParallel(
data = coxdata,
formula = 'Surv(Time.RFS, Distant.RFS) ~ [*]',
FUN = function(formula, data)
coxph(formula = formula,
data = data,
ties = 'breslow',
singular.ok = TRUE),
FUNtype = 'coxph',
variables = colnames(coxdata)[9:ncol(coxdata)],
blocksize = 2000,
p.adjust = "BH")
res5 <- res5[!is.na(res5$P),]
res5## Variable Term Beta StandardError Z
## 1: X1007_s_at X1007_s_at 0.3780639987 0.3535022 1.0694811914
## 2: X1053_at X1053_at 0.1177398813 0.2275041 0.5175285346
## 3: X117_at X117_at 0.6265036787 0.6763106 0.9263549892
## 4: X121_at X121_at -0.6138126274 0.6166626 -0.9953783151
## 5: X1255_g_at X1255_g_at -0.2043297829 0.3983930 -0.5128849375
## ---
## 22211: X91703_at X91703_at -0.4124539527 0.4883759 -0.8445419981
## 22212: X91816_f_at X91816_f_at 0.0482030943 0.3899180 0.1236236554
## 22213: X91826_at X91826_at 0.0546751431 0.3319572 0.1647053850
## 22214: X91920_at X91920_at -0.6452125945 0.8534623 -0.7559942684
## 22215: X91952_at X91952_at -0.0001396044 0.7377681 -0.0001892254
## P LRT Wald LogRank HR HRlower
## 1: 0.2848529 0.2826716 0.2848529 0.2848400 1.4594563 0.72994385
## 2: 0.6047873 0.6085603 0.6047873 0.6046839 1.1249515 0.72024775
## 3: 0.3542615 0.3652989 0.3542615 0.3541855 1.8710573 0.49706191
## 4: 0.3195523 0.3188303 0.3195523 0.3186921 0.5412832 0.16162940
## 5: 0.6080318 0.6084157 0.6080318 0.6077573 0.8151935 0.37337733
## ---
## 22211: 0.3983666 0.3949865 0.3983666 0.3981244 0.6620237 0.25419512
## 22212: 0.9016133 0.9015048 0.9016133 0.9016144 1.0493838 0.48869230
## 22213: 0.8691759 0.8691994 0.8691759 0.8691733 1.0561974 0.55103934
## 22214: 0.4496526 0.4478541 0.4496526 0.4498007 0.5245510 0.09847349
## 22215: 0.9998490 0.9998490 0.9998490 0.9998490 0.9998604 0.23547784
## HRupper P.adjust LRT.adjust Wald.adjust LogRank.adjust
## 1: 2.918050 0.9999969 0.9999969 0.9999969 0.9999969
## 2: 1.757056 0.9999969 0.9999969 0.9999969 0.9999969
## 3: 7.043097 0.9999969 0.9999969 0.9999969 0.9999969
## 4: 1.812712 0.9999969 0.9999969 0.9999969 0.9999969
## 5: 1.779809 0.9999969 0.9999969 0.9999969 0.9999969
## ---
## 22211: 1.724169 0.9999969 0.9999969 0.9999969 0.9999969
## 22212: 2.253373 0.9999969 0.9999969 0.9999969 0.9999969
## 22213: 2.024453 0.9999969 0.9999969 0.9999969 0.9999969
## 22214: 2.794191 0.9999969 0.9999969 0.9999969 0.9999969
## 22215: 4.245498 0.9999969 0.9999969 0.9999969 0.9999969We now take the top probes from the model by Log Rank p-value and use biomaRt to look up the corresponding gene symbols.
res5 <- res5[order(res5$LogRank, decreasing = FALSE),]
final <- subset(res5, LogRank < 0.01)
probes <- gsub('^X', '', final$Variable)
library(biomaRt)
mart <- useMart('ENSEMBL_MART_ENSEMBL')
mart <- useDataset("hsapiens_gene_ensembl", mart)
annotLookup <- getBM(mart = mart,
attributes = c('affy_hg_u133a',
'ensembl_gene_id',
'gene_biotype',
'external_gene_name'),
filter = 'affy_hg_u133a',
values = probes,
uniqueRows = TRUE)Two of the top hits include CXCL12 and MMP10. High expression of CXCL12 was previously associated with good progression free and overall survival in breast cancer in (doi: 10.1016/j.cca.2018.05.041.)[https://www.ncbi.nlm.nih.gov/pubmed/29800557] , whilst high expression of MMP10 was associated with poor prognosis in colon cancer in (doi: 10.1186/s12885-016-2515-7)[https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4950722/].
We can further explore the role of these genes to RFS by dividing their gene expression Z-scores into tertiles for low, mid, and high expression:
# extract RFS and probe data for downstream analysis
survplotdata <- coxdata[,c('Time.RFS', 'Distant.RFS',
'X203666_at', 'X205680_at')]
colnames(survplotdata) <- c('Time.RFS', 'Distant.RFS',
'CXCL12', 'MMP10')
# set Z-scale cut-offs for high and low expression
highExpr <- 1.0
lowExpr <- 1.0
# encode the expression for CXCL12 and MMP10 as low, mid, and high
survplotdata$CXCL12 <- ifelse(survplotdata$CXCL12 >= highExpr, 'High',
ifelse(x <= lowExpr, 'Low', 'Mid'))
survplotdata$MMP10 <- ifelse(survplotdata$MMP10 >= highExpr, 'High',
ifelse(x <= lowExpr, 'Low', 'Mid'))
# relevel the factors to have mid as the reference level
survplotdata$CXCL12 <- factor(survplotdata$CXCL12,
levels = c('Mid', 'Low', 'High'))
survplotdata$MMP10 <- factor(survplotdata$MMP10,
levels = c('Mid', 'Low', 'High'))Plot the survival curves and place Log Rank p-value in the plots:
library(survminer)
ggsurvplot(survfit(Surv(Time.RFS, Distant.RFS) ~ CXCL12,
data = survplotdata),
data = survplotdata,
risk.table = TRUE,
pval = TRUE,
break.time.by = 500,
ggtheme = theme_minimal(),
risk.table.y.text.col = TRUE,
risk.table.y.text = FALSE)
Survival analysis via Cox Proportional Hazards regression.
ggsurvplot(survfit(Surv(Time.RFS, Distant.RFS) ~ MMP10,
data = survplotdata),
data = survplotdata,
risk.table = TRUE,
pval = TRUE,
break.time.by = 500,
ggtheme = theme_minimal(),
risk.table.y.text.col = TRUE,
risk.table.y.text = FALSE)
Survival analysis via Cox Proportional Hazards regression.
3.5 Perform a conditional logistic regression
In this example, we will re-use the Cox data for the purpose of performing conditional logistic regression with tumour grade as our grouping / matching factor. For this example, we will use ER status as the dependent variable and also adjust for age.
x <- exprs(gset[[1]])
x <- x[-grep('^AFFX', rownames(x)),]
x <- scale(x)
x <- x[,which(colnames(x) %in% rownames(metadata))]
coxdata <- data.frame(metadata, t(x))
colnames(coxdata)[1:8] <- c('Age', 'Distant.RFS', 'ER',
'GGI', 'Grade', 'Node',
'Size', 'Time.RFS')
coxdata$Age <- as.numeric(gsub('^KJ', '', coxdata$Age))
coxdata$Grade <- factor(coxdata$Grade, levels = c(1, 2, 3))
coxdata$ER <- as.numeric(coxdata$ER)
coxdata <- coxdata[!is.na(coxdata$ER),]
res6 <- RegParallel(
data = coxdata,
formula = 'ER ~ [*] + Age + strata(Grade)',
FUN = function(formula, data)
clogit(formula = formula,
data = data,
method = 'breslow'),
FUNtype = 'clogit',
variables = colnames(coxdata)[9:ncol(coxdata)],
blocksize = 2000)
subset(res6, P < 0.01)## Variable Term Beta StandardError Z
## 1: X204667_at X204667_at 0.9940504 0.3628087 2.739875
## 2: X205225_at X205225_at 0.4444556 0.1633857 2.720285
## 3: X207813_s_at X207813_s_at 0.8218501 0.3050777 2.693904
## 4: X212108_at X212108_at 1.9610211 0.7607284 2.577820
## 5: X219497_s_at X219497_s_at -1.0249671 0.3541401 -2.894242
## P LRT Wald LogRank HR HRlower
## 1: 0.006146252 0.006808415 0.02212540 0.02104525 2.7021573 1.3270501
## 2: 0.006522559 0.010783544 0.01941078 0.01701248 1.5596409 1.1322713
## 3: 0.007062046 0.037459927 0.02449358 0.02424809 2.2747043 1.2509569
## 4: 0.009942574 0.033447973 0.03356050 0.03384960 7.1065797 1.6000274
## 5: 0.003800756 0.005153233 0.01387183 0.01183245 0.3588083 0.1792329
## HRupper
## 1: 5.502169
## 2: 2.148319
## 3: 4.136257
## 4: 31.564132
## 5: 0.718302 getBM(mart = mart,
attributes = c('affy_hg_u133a',
'ensembl_gene_id',
'gene_biotype',
'external_gene_name'),
filter = 'affy_hg_u133a',
values = c('204667_at',
'205225_at',
'207813_s_at',
'212108_at',
'219497_s_at'),
uniqueRows=TRUE)## affy_hg_u133a ensembl_gene_id gene_biotype external_gene_name
## 1 219497_s_at ENSG00000119866 protein_coding BCL11A
## 2 212108_at ENSG00000113194 protein_coding FAF2
## 3 207813_s_at ENSG00000161513 protein_coding FDXR
## 4 205225_at ENSG00000091831 protein_coding ESR1Oestrogen receptor (ESR1) comes out top - makes sense! Also, although 204667_at is not listed in biomaRt, it overlaps an exon of FOXA1, which also makes sense in relation to oestrogen signalling.
4 Advanced features
Advanced features include the ability to modify block size, choose different numbers of cores, enable ‘nested’ parallel processing, modify limits for confidence intervals, and exclude certain model terms from output.
4.1 Speed up processing
First create some test data for the purpose of benchmarking:
options(scipen=10)
options(digits=6)
# create a data-matrix of 20 x 60000 (rows x cols) random numbers
col <- 60000
row <- 20
mat <- matrix(
rexp(col*row, rate = .1),
ncol = col)
# add fake gene and sample names
colnames(mat) <- paste0('gene', 1:ncol(mat))
rownames(mat) <- paste0('sample', 1:nrow(mat))
# add some fake metadata
modelling <- data.frame(
cell = rep(c('B', 'T'), nrow(mat) / 2),
group = c(rep(c('treatment'), nrow(mat) / 2), rep(c('control'), nrow(mat) / 2)),
dosage = t(data.frame(matrix(rexp(row, rate = 1), ncol = row))),
mat,
row.names = rownames(mat))4.1.1 ~2000 tests; blocksize, 500; cores, 2; nestedParallel, TRUE
With 2 cores instead of the default of 3, coupled with nestedParallel being enabled, a total of 2 x 2 = 4 threads will be used.
df <- modelling[ ,1:2000]
variables <- colnames(df)[4:ncol(df)]
ptm <- proc.time()
res <- RegParallel(
data = df,
formula = 'factor(group) ~ [*] + (cell:dosage) ^ 2',
FUN = function(formula, data)
glm(formula = formula,
data = data,
family = binomial(link = 'logit'),
method = 'glm.fit'),
FUNtype = 'glm',
variables = variables,
blocksize = 500,
cores = 2,
nestedParallel = TRUE,
p.adjust = "BY")
proc.time() - ptm## user system elapsed
## 5.352 2.540 4.4884.1.2 ~2000 tests; blocksize, 500; cores, 2; nestedParallel, FALSE
df <- modelling[ ,1:2000]
variables <- colnames(df)[4:ncol(df)]
ptm <- proc.time()
res <- RegParallel(
data = df,
formula = 'factor(group) ~ [*] + (cell:dosage) ^ 2',
FUN = function(formula, data)
glm(formula = formula,
data = data,
family = binomial(link = 'logit'),
method = 'glm.fit'),
FUNtype = 'glm',
variables = variables,
blocksize = 500,
cores = 2,
nestedParallel = FALSE,
p.adjust = "BY")
proc.time() - ptm## user system elapsed
## 3.904 0.424 4.674Focusing on the elapsed time (as system time only reports time from the last core that finished), we can see that nested processing has negligible improvement or may actually be slower under certain conditions when tested over a small number of variables. This is likely due to the system being slowed by simply managing the larger number of threads. Nested processing’s benefits can only be gained when processing a large number of variables:
4.1.3 ~40000 tests; blocksize, 2000; cores, 2; nestedParallel, TRUE
df <- modelling[ ,1:40000]
variables <- colnames(df)[4:ncol(df)]
system.time(RegParallel(
data = df,
formula = 'factor(group) ~ [*] + (cell:dosage) ^ 2',
FUN = function(formula, data)
glm(formula = formula,
data = data,
family = binomial(link = 'logit'),
method = 'glm.fit'),
FUNtype = 'glm',
variables = variables,
blocksize = 2000,
cores = 2,
nestedParallel = TRUE))## user system elapsed
## 102.748 15.732 64.1534.1.4 ~40000 tests; blocksize, 2000; cores, 2; nestedParallel, FALSE
df <- modelling[,1:40000]
variables <- colnames(df)[4:ncol(df)]
system.time(RegParallel(
data = df,
formula = 'factor(group) ~ [*] + (cell:dosage) ^ 2',
FUN = function(formula, data)
glm(formula = formula,
data = data,
family = binomial(link = 'logit'),
method = 'glm.fit'),
FUNtype = 'glm',
variables = variables,
blocksize = 2000,
cores = 2,
nestedParallel = FALSE))## user system elapsed
## 92.940 0.704 96.823Performance is system-dependent and even increasing cores may not result in huge gains in time. Performance is a trade-off between cores, forked threads, blocksize, and the number of terms in each model.
4.1.5 ~40000 tests; blocksize, 5000; cores, 3; nestedParallel, TRUE
In this example, we choose a large blocksize and 3 cores. With nestedParallel enabled, this translates to 9 simultaneous threads.
df <- modelling[,1:40000]
variables <- colnames(df)[4:ncol(df)]
system.time(RegParallel(
data = df,
formula = 'factor(group) ~ [*] + (cell:dosage) ^ 2',
FUN = function(formula, data)
glm(formula = formula,
data = data,
family = binomial(link = 'logit'),
method = 'glm.fit'),
FUNtype = 'glm',
variables = variables,
blocksize = 5000,
cores = 3,
nestedParallel = TRUE))## user system elapsed
## 253.840 15.744 60.7964.2 Modify confidence intervals
df <- modelling[ ,1:500]
variables <- colnames(df)[4:ncol(df)]
# 99% confidfence intervals
RegParallel(
data = df,
formula = 'factor(group) ~ [*] + (cell:dosage) ^ 2',
FUN = function(formula, data)
glm(formula = formula,
data = data,
family = binomial(link = 'logit'),
method = 'glm.fit'),
FUNtype = 'glm',
variables = variables,
blocksize = 150,
cores = 3,
nestedParallel = TRUE,
conflevel = 99)## Variable Term Beta StandardError Z P
## 1: gene1 gene1 0.0147941 0.0298028 0.496401 0.6196114
## 2: gene1 cellB:dosage 0.3379872 0.9162545 0.368879 0.7122177
## 3: gene1 cellT:dosage -0.1262878 0.8989954 -0.140477 0.8882835
## 4: gene2 gene2 -0.1925899 0.0930030 -2.070793 0.0383782
## 5: gene2 cellB:dosage 1.0481032 1.1686462 0.896852 0.3697977
## ---
## 1487: gene496 cellB:dosage 0.1194649 0.8906723 0.134129 0.8933007
## 1488: gene496 cellT:dosage -0.5660039 0.9987356 -0.566720 0.5709041
## 1489: gene497 gene497 0.0123133 0.0593510 0.207466 0.8356459
## 1490: gene497 cellB:dosage 0.2026582 0.8720848 0.232384 0.8162401
## 1491: gene497 cellT:dosage -0.2496903 0.8719957 -0.286344 0.7746150
## OR ORlower ORupper
## 1: 1.014904 0.9399085 1.09588
## 2: 1.402123 0.1323735 14.85152
## 3: 0.881361 0.0869914 8.92959
## 4: 0.824820 0.6491121 1.04809
## 5: 2.852236 0.1405587 57.87795
## ---
## 1487: 1.126894 0.1136360 11.17506
## 1488: 0.567790 0.0433445 7.43775
## 1489: 1.012389 0.8688675 1.17962
## 1490: 1.224654 0.1295506 11.57676
## 1491: 0.779042 0.0824303 7.36266 # 95% confidfence intervals (default)
RegParallel(
data = df,
formula = 'factor(group) ~ [*] + (cell:dosage) ^ 2',
FUN = function(formula, data)
glm(formula = formula,
data = data,
family = binomial(link = 'logit'),
method = 'glm.fit'),
FUNtype = 'glm',
variables = variables,
blocksize = 150,
cores = 3,
nestedParallel = TRUE,
conflevel = 95)## Variable Term Beta StandardError Z P
## 1: gene1 gene1 0.0147941 0.0298028 0.496401 0.6196114
## 2: gene1 cellB:dosage 0.3379872 0.9162545 0.368879 0.7122177
## 3: gene1 cellT:dosage -0.1262878 0.8989954 -0.140477 0.8882835
## 4: gene2 gene2 -0.1925899 0.0930030 -2.070793 0.0383782
## 5: gene2 cellB:dosage 1.0481032 1.1686462 0.896852 0.3697977
## ---
## 1487: gene496 cellB:dosage 0.1194649 0.8906723 0.134129 0.8933007
## 1488: gene496 cellT:dosage -0.5660039 0.9987356 -0.566720 0.5709041
## 1489: gene497 gene497 0.0123133 0.0593510 0.207466 0.8356459
## 1490: gene497 cellB:dosage 0.2026582 0.8720848 0.232384 0.8162401
## 1491: gene497 cellT:dosage -0.2496903 0.8719957 -0.286344 0.7746150
## OR ORlower ORupper
## 1: 1.014904 0.9573193 1.075953
## 2: 1.402123 0.2327388 8.447014
## 3: 0.881361 0.1513309 5.133105
## 4: 0.824820 0.6873769 0.989746
## 5: 2.852236 0.2886902 28.179866
## ---
## 1487: 1.126894 0.1966714 6.456908
## 1488: 0.567790 0.0801793 4.020804
## 1489: 1.012389 0.9012141 1.137280
## 1490: 1.224654 0.2216631 6.766020
## 1491: 0.779042 0.1410317 4.3033334.3 Remove some terms from output / include the intercept
# remove terms but keep Intercept
RegParallel(
data = df,
formula = 'factor(group) ~ [*] + (cell:dosage) ^ 2',
FUN = function(formula, data)
glm(formula = formula,
data = data,
family = binomial(link = 'logit'),
method = 'glm.fit'),
FUNtype = 'glm',
variables = variables,
blocksize = 150,
cores = 3,
nestedParallel = TRUE,
conflevel = 95,
excludeTerms = c('cell', 'dosage'),
excludeIntercept = FALSE)## Variable Term Beta StandardError Z P
## 1: gene1 (Intercept) -0.25899877 0.9929603 -0.2608350 0.7942198
## 2: gene1 gene1 0.01479414 0.0298028 0.4964011 0.6196114
## 3: gene2 (Intercept) 1.01789937 1.0221176 0.9958730 0.3193118
## 4: gene2 gene2 -0.19258994 0.0930030 -2.0707927 0.0383782
## 5: gene3 (Intercept) -1.32452837 1.2527867 -1.0572657 0.2903903
## ---
## 990: gene495 gene495 0.00681142 0.0717827 0.0948895 0.9244026
## 991: gene496 (Intercept) -0.15901005 0.8510482 -0.1868402 0.8517859
## 992: gene496 gene496 0.04224890 0.0600048 0.7040918 0.4813756
## 993: gene497 (Intercept) -0.07518363 0.9680382 -0.0776660 0.9380938
## 994: gene497 gene497 0.01231331 0.0593510 0.2074660 0.8356459
## OR ORlower ORupper
## 1: 0.771824 0.1102323 5.404152
## 2: 1.014904 0.9573193 1.075953
## 3: 2.767375 0.3732845 20.516164
## 4: 0.824820 0.6873769 0.989746
## 5: 0.265928 0.0228239 3.098408
## ---
## 990: 1.006835 0.8746950 1.158937
## 991: 0.852988 0.1608901 4.522268
## 992: 1.043154 0.9274111 1.173342
## 993: 0.927573 0.1391081 6.185059
## 994: 1.012389 0.9012141 1.137280 # remove everything but the variable being tested
RegParallel(
data = df,
formula = 'factor(group) ~ [*] + (cell:dosage) ^ 2',
FUN = function(formula, data)
glm(formula = formula,
data = data,
family = binomial(link = 'logit'),
method = 'glm.fit'),
FUNtype = 'glm',
variables = variables,
blocksize = 150,
cores = 3,
nestedParallel = TRUE,
conflevel = 95,
excludeTerms = c('cell', 'dosage'),
excludeIntercept = TRUE)## Variable Term Beta StandardError Z P
## 1: gene1 gene1 0.01479414 0.0298028 0.4964011 0.6196114
## 2: gene2 gene2 -0.19258994 0.0930030 -2.0707927 0.0383782
## 3: gene3 gene3 0.07621485 0.0574384 1.3268968 0.1845428
## 4: gene4 gene4 -0.12218117 0.1062646 -1.1497826 0.2502334
## 5: gene5 gene5 0.18939677 0.1064759 1.7787764 0.0752764
## ---
## 493: gene493 gene493 0.05156487 0.0625298 0.8246448 0.4095733
## 494: gene494 gene494 0.03148510 0.0611921 0.5145291 0.6068821
## 495: gene495 gene495 0.00681142 0.0717827 0.0948895 0.9244026
## 496: gene496 gene496 0.04224890 0.0600048 0.7040918 0.4813756
## 497: gene497 gene497 0.01231331 0.0593510 0.2074660 0.8356459
## OR ORlower ORupper
## 1: 1.014904 0.957319 1.075953
## 2: 0.824820 0.687377 0.989746
## 3: 1.079194 0.964291 1.207790
## 4: 0.884988 0.718596 1.089908
## 5: 1.208520 0.980893 1.488972
## ---
## 493: 1.052917 0.931470 1.190200
## 494: 1.031986 0.915350 1.163484
## 495: 1.006835 0.874695 1.158937
## 496: 1.043154 0.927411 1.173342
## 497: 1.012389 0.901214 1.1372805 Acknowledgments
RegParallel would not exist were it not for initial contributions from:
Jessica Lasky-Su, Myles Lewis, Michael Barnes
Thanks also to Horacio Montenegro and GenoMax for testing.
6 Session info
## R version 3.6.1 (2019-07-05)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 18.04.3 LTS
##
## Matrix products: default
## BLAS: /home/biocbuild/bbs-3.10-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.10-bioc/R/lib/libRlapack.so
##
## locale:
## [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
## [3] LC_TIME=en_US.UTF-8 LC_COLLATE=C
## [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
## [7] LC_PAPER=en_US.UTF-8 LC_NAME=C
## [9] LC_ADDRESS=C LC_TELEPHONE=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## attached base packages:
## [1] stats4 parallel stats graphics grDevices utils datasets
## [8] methods base
##
## other attached packages:
## [1] survminer_0.4.6 ggpubr_0.2.3
## [3] ggplot2_3.2.1 biomaRt_2.42.0
## [5] GEOquery_2.54.0 DESeq2_1.26.0
## [7] magrittr_1.5 airway_1.6.0
## [9] SummarizedExperiment_1.16.0 DelayedArray_0.12.0
## [11] BiocParallel_1.20.0 matrixStats_0.55.0
## [13] Biobase_2.46.0 GenomicRanges_1.38.0
## [15] GenomeInfoDb_1.22.0 IRanges_2.20.0
## [17] S4Vectors_0.24.0 BiocGenerics_0.32.0
## [19] RegParallel_1.4.0 arm_1.10-1
## [21] lme4_1.1-21 Matrix_1.2-17
## [23] MASS_7.3-51.4 survival_2.44-1.1
## [25] stringr_1.4.0 data.table_1.12.6
## [27] doParallel_1.0.15 iterators_1.0.12
## [29] foreach_1.4.7 knitr_1.25
##
## loaded via a namespace (and not attached):
## [1] minqa_1.2.4 colorspace_1.4-1 ggsignif_0.6.0
## [4] ellipsis_0.3.0 htmlTable_1.13.2 XVector_0.26.0
## [7] base64enc_0.1-3 rstudioapi_0.10 bit64_0.9-7
## [10] AnnotationDbi_1.48.0 xml2_1.2.2 codetools_0.2-16
## [13] splines_3.6.1 geneplotter_1.64.0 zeallot_0.1.0
## [16] Formula_1.2-3 nloptr_1.2.1 km.ci_0.5-2
## [19] broom_0.5.2 annotate_1.64.0 cluster_2.1.0
## [22] dbplyr_1.4.2 readr_1.3.1 compiler_3.6.1
## [25] httr_1.4.1 backports_1.1.5 assertthat_0.2.1
## [28] lazyeval_0.2.2 limma_3.42.0 acepack_1.4.1
## [31] htmltools_0.4.0 prettyunits_1.0.2 tools_3.6.1
## [34] coda_0.19-3 gtable_0.3.0 glue_1.3.1
## [37] GenomeInfoDbData_1.2.2 dplyr_0.8.3 rappdirs_0.3.1
## [40] Rcpp_1.0.2 vctrs_0.2.0 nlme_3.1-141
## [43] xfun_0.10 lifecycle_0.1.0 XML_3.98-1.20
## [46] zoo_1.8-6 zlibbioc_1.32.0 scales_1.0.0
## [49] hms_0.5.2 RColorBrewer_1.1-2 yaml_2.2.0
## [52] curl_4.2 memoise_1.1.0 gridExtra_2.3
## [55] KMsurv_0.1-5 rpart_4.1-15 latticeExtra_0.6-28
## [58] stringi_1.4.3 RSQLite_2.1.2 highr_0.8
## [61] genefilter_1.68.0 checkmate_1.9.4 boot_1.3-23
## [64] rlang_0.4.1 pkgconfig_2.0.3 bitops_1.0-6
## [67] evaluate_0.14 lattice_0.20-38 purrr_0.3.3
## [70] labeling_0.3 htmlwidgets_1.5.1 bit_1.1-14
## [73] tidyselect_0.2.5 R6_2.4.0 generics_0.0.2
## [76] Hmisc_4.2-0 DBI_1.0.0 withr_2.1.2
## [79] pillar_1.4.2 foreign_0.8-72 abind_1.4-5
## [82] RCurl_1.95-4.12 nnet_7.3-12 tibble_2.1.3
## [85] crayon_1.3.4 survMisc_0.5.5 BiocFileCache_1.10.0
## [88] rmarkdown_1.16 progress_1.2.2 locfit_1.5-9.1
## [91] grid_3.6.1 blob_1.2.0 digest_0.6.22
## [94] xtable_1.8-4 tidyr_1.0.0 openssl_1.4.1
## [97] munsell_0.5.0 askpass_1.17 References
Blighe (2018)
Blighe, Kevin. 2018. “RegParallel: Standard regression functions in R enabled for parallel processing over large data-frames.” https://github.com/kevinblighe.