Desktop version

Home arrow Economics

  • Increase font
  • Decrease font


<<   CONTENTS   >>

The R Package IntegratedJM

Identification of Biomarkers

In this section we describe the R package IntegratedJM, which can be used in order to conduct the analysis presented in this chapter. We use the EGFR project for illustration. The expression matrix X is stored in the R object dat. As mentioned in Section 16.3, the EGFR project consists of information on 3595 genes measured for 35 compounds.

> dim(dat)

Features Samples

3595 35

The vectors responseVector and covariate are the bioactivity (Y) and the fingerprint feature (Z) variables, respectively:

> length(responseVector)

[1] 35

> length(covariate)

[1] 35

The main function in the package, fitJM(), can be used to fit the joint model specified in (16.1):

> jmRes <- fitJM(dat=dat, responseVector=responseVector,

covariate=covariate, methodMultTest = 'fdr')

Note that the argument methodMultTest=’fdr’ implies that the BH-FDR approach (Amaratunga, Cabrera, and Shkedy, 2014) is used for multiplicity adjustment. The list of the top genes is obtained using the function topkGenes. The argument Effect implies that a subset of differentially expressed genes will be selected. The ranking argument specifies how to rank the genes within the selected subset. For the EGFR project, the gene KRAS has the highest absolute value for unadjusted association (see also Table 16.4):

> topkGenes(jointModelResult=jmRes, subset_type ="Effect",ranking = "Pearson"

, k=10, sigLevel = 0.05)

Genes FP-Effect p-adj(Effect) Unadj.Asso. Adj.Asso. p-adj(Adj.Asso.)

  • 1 KRAS -0.29738661 0.0008707120 0.6228461 0.3390398 0.07048685
  • 2 MAP9 -0.13059460 0.0002435665 0.6160688 0.2942768 0.12538775
  • 3 SMG1 -0.09732380 0.0019501488 0.6157263 0.3524950 0.05841850
  • 4 PTER -0.10177997 0.0019501488 0.6123932 0.3462307 0.06378699
  • 5 ODZ3 -0.13710657 0.0055083660 0.5864808 0.3521713 0.05866836
  • 6 SCAF11 -0.16406685 0.0017160848 0.5852470 0.2951218 0.12404584
  • 7 PCYOX1 -0.22745115 0.0019501488 0.5823441 0.2950562 0.12410166
  • 8 PHACTR2 -0.12632673 0.0061587590 0.5793338 0.3488484 0.06155341
  • 9 USP3 -0.06771665 0.0070062886 0.5735814 0.3453967 0.06445048
  • 10 FBXO21 -0.12135041 0.0022172747 0.5670778 0.2748839 0.15757229

The gene FOSL1 is the top ranked gene in the subset of differentially expressed and correlated genes, when ranked based on the absolute value of the fingerprint effect on the gene expression (see Table 16.3):

> topkGenes(jointModelResult=jmRes, subset_type ="Effect and Correlation",

ranking = "CovEffect1", k=10, sigLevel = 0.05)

Genes FP-Effect p-adj(Effect) Unadj.Asso. Adj.Asso. p-adj(Adj.Asso.)

  • 1 FOSL1 1.1942768 0.005728185 -0.8396638 -0.7619553 4.590942e-06
  • 2 FGFBP1 0.7872531 0.008023226 -0.8446045 -0.7813558 2.422426e-06
  • 3 SEPP1 -0.6360825 0.008776129 0.8124630 0.7342640 7.168442e-06
  • 4 SCGB2A1 -0.6139309 0.008017638 0.8288709 0.7565218 4.633984e-06
  • 5 SH2B3 0.6106726 0.005601205 -0.7942338 -0.6880690 2.586543e-05
  • 6 SLCO4A1 0.5988076 0.009534114 -0.7903577 -0.7028792 1.633291e-05
  • 7 PHLDA1 0.5752853 0.005132267 -0.8465864 -0.7683447 3.543378e-06
  • 8 RRM2 0.5645259 0.016186359 -0.7703687 -0.6957629 2.011019e-05
  • 9 TXNIP -0.5267830 0.002845200 0.7453539 0.5827579 5.588912e-04
  • 10 CDC6 0.5210562 0.011427699 -0.8022046 -0.7280215 8.268421e-06

Figure 16.15 shows the expression levels versus the bioactivity for FGFBP1 and the corresponding plot for the residuals and was produced using the function plot1gene().

FIGURE 16.15

Expression levels versus bioactivity gene FGFBP1: raw data (upper panel) and corresponding residuals (lower panel).

>plot1gene(geneName="FGFBP1",fp=covariate,fpName = "-442307337"

,responseVector=responseVector,dat=dat,resPlot=TRUE,

colP = "blue",colA = "white")

 
<<   CONTENTS   >>

Related topics