Gene expression anti-profiles as a basis for accurate universal cancer signatures

摘要: Early screening for cancer is arguably one of the greatest public health advances over last fifty years. However, many tests are invasive (digital rectal exams), expensive (mammograms, imaging) or both (colonoscopies). This has spurred growing interest in developing genomic signatures that can be used diagnosis and prognosis. progress been slowed by heterogeneity profiles lack effective computational prediction tools this type data. We developed anti-profiles as a first step towards translating experimental findings suggesting stochastic across-sample hyper-variability expression specific genes stable general property into predictive diagnostic signatures. Using single-chip microarray normalization quality assessment methods, we an anti-profile colon tissue biopsy samples. To demonstrate translational potential our findings, applied signature samples, without any further retraining normalization, to screen patients based on measurements from peripheral blood independent study (AUC 0.89). method achieved higher accuracy than underlying commercially available 0.81). also confirmed existence hyper-variable across range types found significant proportion tissue-specific cancer. Based these observations, universal accurately distinguishes normal regardless (ten-fold cross-validation AUC > 0.92). have introduced new approach specifically takes advantage gene heterogeneity. demonstrated successfully develop peripheral-blood diagnostics highly accurate signature. By using no would required before their application clinical settings. Our results suggest may inexpensive non-invasive tests.