return to case studies

Gene Expression for DNA Microarrays

Client

The client is a developer of a revolutionary, high-throughput DNA-chip platform technology used by the pharmaceutical and biotechnology industries as a multi-gene screening tool for disease diagnosis and drug discovery.

Challenge

To mine and explore an extensive set of microarray expression data extracted from heterogeneous tissue samples in order to identify genes of potential pharmacological or clinical significance. The identified genes would then be used in the construction of DNA microarrays for disease diagnostics and targeted drug development.

Objectives

AnVil's goal was to provide the client with a set of genes that could be used to differentiate normal tissue from two different types of cancer with a high degree of accuracy. The strategy was for AnVil to use its proprietary data mining tools and processes along with its domain and data mining expertise to generate results that would then be compared with a classifier derived from genes in the literature that were associated with the specific cancer of interest.

Solutions

AnVil first undertook an analysis of the data quality, discovering that three patient samples were statistical outliers and thus suspect. Upon further investigation, AnVil learned that one of these patients was undergoing drug therapy at the time the sample was collected. Applying advanced data mining and visualization techniques to expression data derived from cancerous tissue specimens, AnVil identified a set of genes that could accurately differentiate between normal tissue and two cancer subtypes.

Results

Click to Enlarge

AnVil prepared a 30-gene complex classifier from three independent gene sets that was designed to maximize separation of class members. The classifier prepared by AnVil was 99 percent accurate in predicting sample class as judged by 11 different analytical classifiers. (Accuracy was 100 percent after the patient treated with chemotherapy was excluded from the data.) In addition, a list of genes present in the original gene set, and implicated in both the cancer of interest and carcinogenesis, was assembled from public databases, the biomedical literature, and primary publications on cancer microarray studies using the tissue of interest to produce a 20-gene classifier of 90 percent accuracy. Surprisingly, the newly identified 30-gene classifier, sharing only one gene in common with the literature derived set, performed significantly better at distinguishing among the different tissue pathologies than the 20-gene classifier derived from the literature.

To find out more about how AnVil can help you forge the optimal path to commercial discovery, send e-mail to info@anvilinfo.com.

back to top