Sample A: Cover Page of Thesis, Project, or Dissertation Proposal

for the 24 samples in the Lu, et al. raw data files [33] (filtered as described above), and the probe
and ProbeSet presence were then predicted.
Results
The results reported here are divided into three sections: (1) providing evidence for the BaFl
cleansing process, including details on the number of probes removed for each probe sequence
filter, evidence of sample batch processing variability, and exemplar ProbeSets affected by such
confounding factors. (2) the stages and effects of the BaFL probe cleansing pipeline, including
graphic representation of each filtering step and the final profile consistency across the two
datasets. (3) the research enrichment the BaFL pipeline facilitates, including elucidating potential
transcript regions of interest and a priori predictions of independent datasets. For the established
algorithms we have used the validated samples from the datasets as input, and validated
ProbeSets from the author’s lists to compare their classification performance to each other and to
the results of our method. The classification performance of the author’s original gene lists is
used to see whether the sample cleansing protocol alone has a significant effect. We show that
while previous efforts have identified informative genes, the methods are over-tuned to technical
properties (lab specific) or non-primary factor biological properties (such as SNPs), rather than
the desired biological response to the principle factor (here, disease state specific).
41