Methods. One hundred and sixty-five tumour samples were obtained by fine needle aspiration (FNA). cDNA were hybridized on Splice ArrayTM. A nearest centroid prediction rule was developed to classify lesions as malignant or benign on a training set, and its performance evaluated on an independent validation set. A two-way ANOVA model was used to identify probesets that present a differential expression between cancer and benign lesions while adjusting for scan dates. P-values were adjusted for False Discovery Rate.
Findings. Overall 120 breast cancers and 45 benign lesions were included in the study. A 1228-probeset molecular classifier for breast cancer diagnosis was generated from the training set (n=94). This signature accurately classified all samples (100% accuracy, 95% exact CI: 96-100%). In the validation set (n=71), the molecular predictor accurately classified 68 out of 71 tumours (96%, 95% CI: 88-99%). When the 165 samples were taken into account, 37 858 exon-probesets (5.4%) and 3733 genes (20%) were found to be differentially expressed between malignant and benign conditions (adjusted p-value<0.05). Pathway analyses showed that genes involved in spliceosome assembly were significantly enriched in malignant condition (permutation p=0.002). In the same population of 165 samples, 956 exon-probesets presented both a higher intensity and higher splice index in breast cancer, although located on unchanged genes.
Interpretation. The present study provides a thorough description of differentially expressed exons between breast cancer and benign lesions, and emphasizes the contribution of spliceosome and alternative transcripts to the molecular portrait of breast malignancy. This allowed the development of a molecular classifier for breast cancer diagnosis using FNA.
ORGANISM(S): Homo sapiens
DISEASE(S): Benign Lesion,Breast Cancer
SUBMITTER: Philippe Dessen