Rapid identification of prognostic imaging biomarkers for non-small cell lung cancer by leveraging public gene expression microarray data

ABSTRACT: To rapidly identify new prognostic imaging biomarkers, we propose a bioinformatics approach that integrates gene expression and image data and leverages public gene expression data. We demonstrate our approach in non-small cell lung carcinoma patients for whom CT, PET/CT and gene expression data are available but without clinical follow-up. We extracted 180 image features and 56 high quality gene expression clusters, represented by metagenes. 115 image features were expressed in terms of metagenes, using sparse linear regression and cross-validation, with an accuracy of 65-86%. After mapping the signatures to a public gene expression dataset, 26 image features were significantly associated with recurrence-free survival and 22 with overall survival. A multivariate analysis identified multip

ORGANISM(S): Homo sapiens

SUBMITTER: Gevaert Olivier 

PROVIDER: S-ECPF-GEOD-28827 | biostudies-other |

REPOSITORIES: biostudies-other