Help

[1] Overview
[2] Browse
[3] Search
[4] High-throughput
[5] Drug Discovery
[6] Download
[7] Contact

[1] Overview
Protein-coding RNAs refer to RNAs that encode proteins. This kind of RNAs control various biological functions to maintain morphology and functions of tissues, whose dysfunction often influences cancer development. Accumulating researches have proven that cancers related drugs can target protein-coding RNAs. In addition, protein-coding RNAs may play critical roles in drug resistance. These researches produced substantial data such as literatures and high-throughput microarray which were hard to be screened out by researchers. To collect and annotate these data, we developed DREAM database (Figure 1), a comprehensive manually curated database including protein-coding RNAs and drug targets or drug sensitivity associations. Current version contains 1601 scientific literatures and 195 high-throughput microarray data. Each entry in DREAM database contains detailed information on RNAs, drug, cancer, and other information.
The overview of the database is as followings:


Figure 1. The overview of the DREAM database.

[2] Browse
To browse protein-coding RNAs and cancer drugs association data in the database, please click the menu "Browse". All the data were divided into two parts called ‘Drug intervention’ and ‘Drug resistance’. Users can browse all entries in three ways: by compound name, by gene name, or by disease name. In ‘Drug intervention’ part, take a browse "glioma"-related protein-coding RNAs and drugs as an example. To browse the entries for glioma, please click "Disease name" and select the "glioma" option. The browse result will be displayed in the right panel as presented in Figure 2A. In ‘Drug resistance’ part, take a browse "cisplatin"-related protein-coding RNAs and cancers as an example. To browse the entries for cisplatin, please click "Compound name" and select the "cisplatin" option. The browse result will be displayed in the right panel as presented in Figure 2B.


Figure 2. The browse interface of the DREAM database.

[4] High-throughput
The current version of DREAM database contains 195 high-throughput microarray data across 36 cancer subtypes, including 94 protein-coding RNAs and drug intervention associations microarray data and 101 protein-coding RNAs and drug resistance associations microarray data. To browse and search these data in our database, please click the menu "High-throughput". We provide three search methods: search by compound, search by feature, or search by disease. User can set the statistical significance (p-values) and biological significance (fold changes) for different high-throughput microarray data in order to quickly identify the most interesting gene candidates associated with drug targets or drug resistance.The p-values and fold changes are calculated on the basis of case groups relative to control groups. In “drug intervention” module, the case groups are disease cells with drug intervention. The control groups are disease cells with placebo intervention such as PBS, DMSO. In “drug sensitivity” module, the case groups are drug resistant disease cells and the control groups are normal disease cells.In addition, we also offer users an interactive visualization tool such as volcano plot and gene enrichment analysis such as GO annotations and KEGG pathways analysis. The ‘High-throughput’ page is displayed in Figure 4 taking ‘temozolomide’ as an example.
1. click to choose search by compound;
2. input user’s interested drug ‘temozolomide’ and set special cut-off value such as ‘fold change>2’ and ‘p value< 0.05’;
3. detailed information returns in search results;
4. click to perform function analysis.

Figure 4. The high-throughput interface of the DREAM database.

[5] Drug Discovery
In ‘Drug Discovery’ page, user can calculate the correlation coefficient between the drug's gene expression signature and disease's expression signature to repurpose and identify novel drug indications in different cancers. The raw R code was displayed in the below table. The “drug's gene expression signature” is extracted from drug-disease related gene expression profiles in DREAM. The “disease's expression signature” is gene sets with expressed protein-coding RNAs in disease groups relative to healthy control groups. The “disease's expression signature” demands user upload by themselves and must be composed of gene symbol and fold changes. 

## import user data
data <- read.table(<user_data>, header=T, sep = "\t", quote = "")
input.gene.list <- data[,1]
input.fd.list <- as.numeric(as.character(unlist(data[,2])))
input.dat <- data.frame(input.gene.list = input.gene.list, input.fd.list = input.fd.list)
rownames(input.dat) <- input.gene.list

## import reference data
dat <- read.table(<reference_data>, header = T, sep = "\t", quote = "")
dat$ID <- paste0(dat$Drug.id, "---", dat$Drug.name, "---", dat$Dataset.ID, "---", dat$Disease)
dat.ID.list <- unique(dat$ID)

result <- t(sapply(dat.ID.list,function(x){
  #print(x)
  drug.id <- strsplit(x,"---")[[1]][1]
  drug.name <- strsplit(x,"---")[[1]][2]
  dataset.id <- strsplit(x,"---")[[1]][3]
  disease.id <- strsplit(x,"---")[[1]][4]
  tmp.gene.list <- dat$Gene.id[dat$Dataset.ID==dataset.id&
                                 dat$Drug.id==drug.id&
                                 dat$Disease==disease.id]
  tmp.fd.list <- dat$Foldchange[dat$Dataset.ID==dataset.id&
                                  dat$Drug.id==drug.id&
                                  dat$Disease==disease.id]
  tmp.dat <- data.frame(tmp.gene.list=tmp.gene.list,
                        tmp.fd.list=as.numeric(as.character(unlist(tmp.fd.list))))
  rownames(tmp.dat) <- tmp.gene.list
  
  overlap.gene <- intersect(input.gene.list,tmp.gene.list)
  
  cortest <- cor.test(input.dat[overlap.gene,"input.fd.list"],
                      tmp.dat[overlap.gene,"tmp.fd.list"])
  c(drug.id,drug.name,dataset.id,disease.id,cortest$estimate[[1]],cortest$p.value) 
}))

result <- data.frame(result)
colnames(result) <- c("drug_id","drug.name","dataset_id","disease","cor","p")
result$adj.p <- p.adjust(as.numeric(as.character(unlist(result$p))))

result.filter <- result[result$adj.p<0.05,]
result.filter$cor <- format(as.numeric(as.character(unlist(result.filter$cor))), scientific = TRUE,digit = 3)
result.filter$p <- format(as.numeric(as.character(unlist(result.filter$p))), scientific = TRUE, digit = 3)
result.filter$adj.p <- format(result.filter$adj.p, scientific = TRUE, digit = 3)
options(digits = 3)
write.table(result.filter, <output_file>, row.names = F, sep = "\t", quote = F)

In ‘Drug Discovery’ page, firstly user need to upload disease's expression signature including gene symbols and fold change (relative to a healthy control) by themselves. Then, our database will match and calculate the correlation coefficients based on the database containing drug's gene expression signature. After a while, the database will output the results including correlation coefficient, drugs, and cancers. The ‘Drug discovery’ page is displayed in Figure 5.

Figure 5. The drug-discovery interface of the DREAM database.

[6] Download
To download data in DREAM database, please click the menu ‘Download’. The DREAM provides two formats of downloadable files in TEXT and Excel formats, respectively. The ‘Download’ page is displayed in Figure 6.

Figure 6. The download interface of the DREAM database.

[7] Contact
If you have any questions, suggestions or comments, please contact us by e-mails. In ‘Contact’ page, there are the details of two corresponding writers of our team.

Running, please be patient...