Radiogenomic Analysis of F-18-Fluorodeoxyglucose Positron Emission Tomography and Gene Expression Data Elucidates the...

Transcriptomic profiling has enabled the neater genomic characterization of several cancers, among them colorectal cancer (CRC), through the derivation of genes with enhanced causal role and informative gene sets. However, the identification of small-sized gene signatures, which can serve as potential biomarkers in CRC, remains challenging, mainly due to the great genetic heterogeneity of the disease.

We developed and exploited an analytical framework for the integrative analysis of CRC datasets, encompassing transcriptomic data and positron emission tomography (PET) measurements. Profiling data comprised two microarray datasets, pertaining biopsy specimen from 30 untreated patients with primary CRC, coupled by their F-18-Fluorodeoxyglucose (FDG) PET values, using tracer kinetic analysis measurements. The computational framework incorporates algorithms for semantic processing, multivariate analysis, data mining and dimensionality reduction.

Transcriptomic and PET data feature sets, were evaluated for their discrimination performance between primary colorectal adenocarcinomas and adjacent normal mucosa. A composite signature was derived, pertaining 12 features: 7 genes and 5 PET variables. This compact signature manifests superior performance in classification accuracy, through the integration of gene expression and PET data.

This work represents an effort for the integrative, multilayered, signature-oriented analysis of CRC, in the context of radio-genomics, inferring a composite signature with promising results for patient stratification.