Identifying differentially expressed genes based on their expressions in leukemia

In 1999, Golub invented a new reasonable and more efficient approach to recognize acute Leukemia cases based on gene expression data. Studying on 6817 genes derived from 38 patients, they finally introduce a complex inventory statistical method to diagnose Leukemia patients in earlier stage. Their result was absolutely outstanding, without any mistake in patient diagnosis; the result perfectly (100%) assigned tumors to known classes correctly. Here, in this study, we want to introduce a simpler method to distinguish Differentially Expressed (DE) genes and use Golub approach as a test case to evaluate our results.

We used the same data as Golub study that were produced by oligonucleotide microarry. Several statistical methods such as t-test, Wilcoxon test and permutation test were applied to identify DEs.

The Westfall and Yang based on t-statistic at significant level 0.015 (α= 0.015) recognized 48 genes significant (or DE). Of those 28 genes (58%) were exactly similar to Golub study and the rest, 20 genes (42%), were different. However, with these 48 genes, the result of assigning tumors to known classes is completely as identical as that of of Golub study.

The result of assigning tumors to known classes with these 48 genes is completely as identical as that of Golub study.