Virtual CGH: an integrative approach to predict genetic abnormalities from gene expression microarray data applied in...

Area of Science:

• Genomics
• Bioinformatics
• Cancer Biology

Background:

• Comparative Genomic Hybridization (CGH) detects DNA Copy Number Alterations (CNAs) crucial in tumorigenesis.
• Gene Expression Profiling (GEP) data is abundant, offering a potential resource for inferring CNAs.
• GEP-derived CNAs may better reflect functional relevance in disease pathogenesis.

Purpose of the Study:

• To develop a computational method for predicting CNAs from GEP data.
• To leverage existing GEP data for identifying recurrent CNAs in tumors.
• To enhance the detection of functionally significant genetic abnormalities in cancer.

Main Methods:

• Proposed virtual CGH (vCGH), a novel computational approach using hidden Markov models (HMMs).
• Trained vCGH on paired GEP and CGH data from tumor samples.
• Applied vCGH to predict CNAs from GEP data in new tumor samples.

Main Results:

• vCGH achieved 80% sensitivity, 90% specificity, and 90% accuracy in predicting CNAs in Diffuse Large B-Cell Lymphomas (DLBCL).
• Identified recurrent CNAs largely concordant with experimental CGH, including gains of 1q, 3q, 7, 11q, 12, 18q21 and losses of 6q, 8p, 9p, 17p.
• Discovered novel recurrent abnormalities (e.g., gains of 1p, 2q, 6q; losses of 1q, 6p, 8q) associated with clinical outcomes in DLBCL.

Conclusions:

• Developed vCGH, a computational tool to predict genome-wide genetic abnormalities from GEP data in lymphomas.
• vCGH demonstrates potential for broader application across various tumor types.
• This method significantly enhances the identification of functionally important genetic abnormalities in cancer research.