02.08.2013 Views

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

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

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

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

implemented in <strong>or</strong>der to demonstrate that our method results in data that<br />

behaves m<strong>or</strong>e consistently upon down selection. Area under the receiver<br />

operating curve will be rep<strong>or</strong>ted f<strong>or</strong> three separate classification alg<strong>or</strong>ithms:<br />

random f<strong>or</strong>est (RF), k-nearest neighb<strong>or</strong>s (kNN), and linear discriminate<br />

analysis (LDA). F<strong>or</strong> comparison, similar datasets have been subjected to the<br />

commonly accepted RMA and dCHIP probe cleansing alg<strong>or</strong>ithms.<br />

3) The ultimate goal <strong>of</strong> any Microarray experiment is to generate a subset <strong>of</strong><br />

genes that either gives insight into a biological mechanism imp<strong>or</strong>tant to the<br />

sample state <strong>or</strong> that gives a high rate <strong>of</strong> success in predicting the state <strong>of</strong> a<br />

sample. Here we have predicted a set <strong>of</strong> genes <strong>of</strong> interest f<strong>or</strong> a two class<br />

experiment, n<strong>or</strong>mal tissues versus adenocarcinoma lung cancer tissues. This<br />

set <strong>of</strong> genes demonstrates impressive latent structure with and across datasets,<br />

as well as supervised classification perf<strong>or</strong>mance f<strong>or</strong> random f<strong>or</strong>ests, kNN and<br />

LDA. Comparisons <strong>of</strong> these genes have been made using the intensity values<br />

by the commonly used methods RMA and dCHIP, in place <strong>of</strong> ours. Finally,<br />

the relevance to the biological state <strong>of</strong> the sample <strong>of</strong> particular members <strong>of</strong><br />

the gene list was investigated by literature review<br />

4) Given the w<strong>or</strong>ld wide incidence rates <strong>of</strong> non small cell lung cancer and that<br />

incidences <strong>of</strong> NSCLC are increasing in individuals who have never smoked<br />

[90], a multiclass NSCLC dataset was constructed from the Bhattacharjee<br />

data. This dataset included the adenocarcinoma, squamous cell carcinoma<br />

23

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!