Premium
Application of a genomics‐based diagnostic test for disease in the mammalian system
Author(s) -
O'Rourke Christine,
Gordon Gavin,
Pun Pattle
Publication year - 2006
Publication title -
the faseb journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.709
H-Index - 277
eISSN - 1530-6860
pISSN - 0892-6638
DOI - 10.1096/fasebj.20.5.a1325-c
Subject(s) - genomics , diagnostic test , disease , cancer , computational biology , test (biology) , medicine , bioinformatics , biology , oncology , gene , genetics , genome , pediatrics , paleontology
We have previously shown that simple, clinically relevant tests based on the ratios of select gene expression levels can be useful in cancer diagnosis (Cancer Res, 62:4963–7, 2002) and prognosis (J Natl Cancer Inst, 95:598–605, 2003). To determine whether this approach has broad applicability while retaining high classification accuracy, we conducted analyses using previously published microarray datasets to differentiate among human cancer types and between normal human tissues and diseased tissues, i.e. leukemia, lymphoma, sarcoma, carcinoma, and tumors and tissues of CNS, prostate, ovary, uterus, kidney, thyroid, and colon. A few nonhuman specimens were also analyzed. In each case, a subset of the samples was used to form a training set to create a ratio‐based model with the remaining samples used for independent validation (i.e. the test set). We found gene ratios that were highly accurate in the training set (100% with two exceptions) were validated with similar accuracy in the test set (74–100%). Thirteen of 15 tests had statistically significant classification accuracies in the test set (p<0.05, Fisher's exact test). Some improvements were shown in classification accuracy when compared to those found in the original papers. Since they confer several advantages over other equally accurate, but more complex bioinformatics tools, we conclude that gene ratio‐based tests prospectively validated will likely find clinical utility in distinguishing between different cancer types and normal tissues and diseased tissues. This work was partly funded by Wheaton College alumni grants for student research to C.O.