Notes: 4 spectra of same patient in K2099 and K2199 were collected on 4 different dates.
Status:
old normalized data set
graph : 100%
QC : 100%
new normalized data set
graph : 0%
QC : 0%
Leave One Out : waiting clinical info of JFCR
Output
old normalization
intensity graph: .\old normalization\graph
QC: .\old normalization\QC\result
QC summary: .\old normalization\QC
new normalization
Data Sets Background
data sets with old normalization method
data come from two hospitals: Kanazawa and JFCR.
data from Kanazawa have laser intensity 2099 and 2199, JFCR have laser intensity 2399.
"baseline" data are samples before taking Iressa, "day14" data are samples after taking Iressa 14 days.
all data sets have done smoothing, baseline correction, and calibration by investigator, then binned and normalized by Huiming. Data haven't taken log transformation.
each patient have 3 samples except patient31 in JFCR_BaseLine_2399.
all samples are collected on same day, but spectra are collected on different day. Kanazawa2199 is collected a few months before Kanazawa 2099.
Kanazawa2199 includes both pre and post samples.
data set and patient info
data set 1 = Kanazawa2099 pre: 26 (3 spectra per patient, all 3 spectra are collected in 3 different days)
data set 2 = Kanazawa2099 post: 17 (3 spectra per patient,all 3 spectra are collected in 3 different days)
data set 3 = JFCR2399 pre : 43 (3 spectra per patient except patient 31 only have 2 spectra)
data set 4 = JFCR2399 post: 35 (3 spectra per patient)
data set 5 = Kanazawa2199 : pre - 26 (1 spectrum per patient collected different day with K2099), post - 17 (1 spectrum per patient,collected different day with K2099)
data sets with new normalization method
same background as old data sets except using different way to normaliz
Analysis Requests
old normalization
graph of expression(sort by peaks and spectra)
each data set alone
1 vs. 2
1 vs. 3
1 vs. 5 pre part
2 vs. 4
2 vs. 5 post part
3 vs. 4
1 and 5 pre part (include cluster to show day to day variance)
2 and 5 post part (include cluster to show day to day variance)
1 (cluster to show day to day variance)
2 (cluster to show day to day variance)
K2099 and K2199 pre: day to day (sort by date)
k2099 and k2199 post: day to day (sort by date)
K2099 and K2199 pre: by laser power
k2099 and k2199 post: by laser power
K2099(avg) and K2199 pre: by laser power
k2099(avg) and k2199 post: by laser power
k2099(avg) and JFCR(avg) pre: by hospital
k2099(avg) and JFCR(avg) post: by hospital
clustering
d1: K2099 pre and K2199 pre (total 26 patients)
d2: K2099 post and K2199 post (total 17 patients)
d3: JFCR pre (total 43 patients)
d4: JFCR post (total 35 patients)
d5: K2099 pre
d6: K2099 post
pre : average of K2099 vs. K2199 (Kappa agreement)
post : average of K2099 vs. K2199 (Kappa agreement)
QC analysis on all 5 data sets - method: use pre vs. pre, post vs. post
new normalization
intensity expression
each data set alone
pre & post: K2099 vs. K2199 sort by date
pre & post: K2099 vs. K2199 sort by laser power
pre & post: K2099(avg) vs. K2199 sort by date
pre & post: K2099(avg) vs. K2199 sort by laser power
pre & post: K2099(avg) vs. JFCR(avg) sort by hospital
clustering
pre & post: K2099
pre & post: K2099 vs. K2199
pre & post: K2099(avg) vs. K2199
pre & post: JFCR
QC analysis on all data set plus pre/post K2099 vs. K2199
Leave One Out
generate the 3rd data set form data set 1&2 (as testing set), data set 3&5(as training set) : value = (post - pre)/pre
run paired TTest if both pre and post have intensity (proc means)
Change Log:
Contact Info:
Primary Study Contact Person: Fumiko Taguchi (6-3819)