Background:Some oncogenes such as ERBB2 and EGFR are over-expressed in only a subset of patients. Cancer outlier profile analysis is one of computational approaches to identify outliers in gene expression data. A database with a large sample size would be a great advantage when searching for genes over-expressed in only a subset of patients.Description:GENT (Gene Expression database of Normal and Tumor tissues) is a web-accessible database that provides gene expression patterns across diverse human cancer and normal tissues. More than 40000 samples, profiled by Affymetrix U133A or U133plus2 platforms in many different laboratories across the world, were collected from public resources and combined into two large data sets, helping the identification of cancer outliers that are over-expressed in only a subset of patients. Gene expression patterns in nearly 1000 human cancer cell lines are also provided. In each tissue, users can retrieve gene expression patterns classified by more detailed clinical information.Conclusions:The large samples size (>24300 for U133plus2 and >16400 for U133A) of GENT provides an advantage in identifying cancer outliers. A cancer cell line gene expression database is useful for target validation by in vitro experiment. We hope GENT will be a useful resource for cancer researchers in many stages from target discovery to target validation. GENT is available at http://medicalgenome.kribb.re.kr/GENT/ or http://genome.kobic.re.kr/GENT/.

The EST division of GenBank, dbEST, is widely used in many applications such as gene discovery and verification of exon–intron structure. However, the use of EST sequences in the dbEST libraries is often hampered by inconsistent terminology used to describe the library sources and by the presence of contaminated sequences. Here, we describe CleanEST, a novel database server that classified dbEST libraries and removes contaminants. We classified all dbEST libraries according to species and sequencing center. In addition, we further classified human EST libraries by anatomical and pathological systems according to eVOC ontologies. For each dbEST library, we provide two different cleansed sequences: ‘pre-cleansed’ and ‘user-cleansed’. To generate pre-cleansed sequences, we cleansed sequences in dbEST by alignment of EST sequences against well-known contamination sources: UniVec, Escherichia coli, mitochondria and chloroplast (for plant). To provide user-cleansed sequences, we built an automatic user-cleansing pipeline, in which sequences of a user-selected library are cleansed on-the-fly according to user-selected options. The server is available at http://cleanest.kobic.re.kr/ and the database is updated monthly.

No abstract

No abstract