What is the source of the information served on the PACdb database?

Information on the relationship between SNPs and expression transcript levels that is served on PACdb comes from a series of publications describing studies characterizing eQTLs in cell lines from HapMap CEU and YRI samples for which transcript levels were assayed using the Affymetrix Human Exon 1.0 ST Array. The relationships between genetic variants and pharmacologic phenotypes come from genome-wide association studies (GWAS) on a broad array of anti-cancer drugs. Information on physical or functional annotation served on PACdb comes from public resources, including the HapMap (release 23a), NCBI (dbSNP 129), or is information created by us using data downloaded from these public resources. All papers relevant for data served on PACdb are available here.

How and by whom is PACdb going to be maintained?

PACdb will be maintained and kept up to date by the Pharmacogenetics of Anticancer Agents Research Group based at the University of Chicago, which is funded by the NIHGMS U0161393.

What are your immediate plans for PACdb?

In order to encourage the development of bioinformatics applications that utilize or integrate the data served on PACdb, we are developing an application programming interface (API), written in Simple Object Access Protocol (SOAP). SOAP enables the exchange of structured information, written in the ubiquitous Extensible Markup Language (XML), with other databases or other applications. The architecture allows application developers to utilize real-time data from PACdb -- drug phenotype data and gene expression data.

What drug data are stored in PACdb?

Currently, PACdb serves data on the following drugs: carboplatin, cisplatin, etoposide, daunorubicin, and cytarabine (Ara-C). PACdb will be updated regularly to include other drugs.

What p-values are being returned in PACdb?

All p-values in the results are uncorrected.

How does PACdb determine the relationship between genotype and cytotoxicity?

Genome-wide association studies were performed for each of the drugs. Quantitative Trait Disequilibrium Test (QTDT) was performed separately on each quantitative phenotype in each population to detect genotype-phenotype associations. Multiple comparisons were adjusted for by Q-VALUE. The Genotype-Cytotoxicity Query tool in PACdb allows the user to specify a p-value or q-value threshold and retrieves a list of SNPs that show associations with drug-induced cytotoxicity for a particular drug in a particular population.

What association metric is being used to describe the relationship between gene expression and cytotoxicity?

The Expression - Phenotype Query Tool evaluates whether gene expression level significantly predicts cytotoxicity measures (e.g., IC50). Since the HapMap samples include trios, we used a general linear model, in which the covariance structure within a trio was modeled using a Toeplitz covariance structure with two diagonal bands to allow for familial dependencies in the data. In this scheme, each trio is a unit; mother and father IC50 are treated as independent while the offspring IC50 is allowed to co-vary with both father and mother.

The model was programmed using the PROC MIXED procedure in SAS/STAT. PROC MIXED constructs a Type III L matrix (see SAS User Guide), which is used to compute the F-statistic:

We define the following:

NDF = row rank of L

DDF = the denominator degrees of freedom calculated using DDFM=BETWITHIN.

A p-value is calculated for the F-test from the area that exceeds this statistic from an F distribution using the NDF and DDF degrees of freedom. The F statistic gives the magnitude of the test for association between the predictor (log2-transformed gene expression) and the response (e.g., log2-transformed IC50). Small p-values (e.g., 0.05) indicate significant association.

What output format does PACdb support?

PACdb allows downloads of query results in HTML or excel format.

How do I find out more about PACdb?

See online tutorial. Or contact us.