Biology Reference
In-Depth Information
exist as a selective process associated with a specific molecular signature.
Study of the functional relevance of this molecular signature will shed
light on the molecular diagnosis and therapy of human breast cancer.
Another study published by Di Fiore's group illustrated a beautiful
experiment to show that survival of stage 1 lung adenocarcinomas could
be predicted based on the expression of 10 genes discovered through ana-
lyzing the clinical DNA microarray data (Bianchi
et al
., 2007). The beauty
of this study is that the original clues for identifying candidate genes came
from their analyses of two pre-existing sets of clinical microarray data
from two independent places (Michigan and Harvard, USA). The original
datasets from Michigan included 86 adenocarcinomas of human lung can-
cer; those from Harvard included 84 cases. By analyzing these microarray
data, a 80-gene model was created and tested on an independent cohort of
lung cancer patients using RT-PCR. As a result, a 10-gene predictive
model exhibited a prognostic accuracy of approximately 75% in stage 1
lung adenocarcinoma when tested on two additional independent cohorts.
Potentially, this type of approach enables the discovery of similar predic-
tive molecular signatures for other types of cancer.
There are many more examples showing the application of DNA
microarray technology in identifying molecular targets and establishing
diagnostic molecular signatures for human cancers. The overall weakness
of these microarray studies is the lack of ability to make decisions based
solely on the DNA microarray data.
2.2. Blood Cancers
An excellent example for using DNA microarray technology to study
human blood cancers is the molecular classification of human acute
myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL)
(Golub
et al
., 1999). In this study, bone marrow mononuclear cells from
11 AML and 27 ALL patients diagnosed pathologically were used as an
RNA source for DNA microarray analysis, from which 50-gene predictors
that distinguish AML from ALL were derived. These 50-gene predictors
were tested and validated on 38 new samples from AML or ALL patients.
As a result, 36 of the 38 predictions agreed with the patients' clinical diag-
nosis (the remaining two were uncertain). This high prediction rate (95%)
Search WWH ::
Custom Search