Important update (January 20th, 2011): the data below have been corrected for the BCR batch which is not necessarily the processing batch. The dataset needs to be reanalyzed.
Batch vs clinical traits
Number of batches is 12. Correlation between batch and center:
table(two,batchID)
batchID
two 0689 0848 0979 1096 1198 1440 1551 1633 1818 1871 1947 2043
18 0 0 12 2 0 2 0 2 0 0 0 0
21 13 0 0 0 0 0 0 4 0 0 0 0
22 9 0 0 0 6 3 0 12 2 0 2 0
33 0 0 0 0 4 5 0 0 1 0 1 0
34 0 6 0 0 0 2 0 1 7 1 1 0
37 0 0 2 6 1 0 0 1 0 0 0 0
39 0 0 0 0 0 12 0 0 4 0 0 0
43 0 3 2 0 0 0 2 1 4 0 1 0
46 0 0 5 0 0 0 0 0 2 0 0 0
51 0 0 0 3 0 0 0 0 0 0 0 1
56 1 0 0 0 0 0 0 2 2 0 0 6
60 0 20 0 0 1 0 0 0 1 2 0 2
63 0 0 0 0 0 2 0 0 1 0 4 0
66 0 18 15 0 6 0 0 0 0 0 0 0
70 0 0 0 0 0 0 0 0 2 0 0 0
77 0 0 0 0 0 0 0 0 0 0 4 10
79 0 0 0 0 0 0 0 0 0 0 1 0
85 0 0 0 0 0 0 0 0 3 0 1 0
90 0 0 0 0 0 0 0 0 0 0 1 0
92 0 0 0 0 0 0 0 0 0 0 0 2
94 0 0 0 0 0 0 0 0 0 0 1 0
96 0 0 0 0 0 0 0 0 0 0 0 2
98 0 0 0 0 0 0 0 0 0 0 0 1
Significant batch/clinical traits correlations (complete list can be found here):
Unknown macro: {csv} LUSC,DataType,NumberOfNAs,Test,Pvalue
tumor_stage,factor,27,Pearson's Chi-squared test,8.78E-14
year_of_initial_pathologic_diagnosis,integer,23,Kruskal-Wallis rank sum test,7.95E-12
days_to_form_completion,integer,30,Kruskal-Wallis rank sum test,1.48E-09
primary_tumor_pathologic_spread,factor,23,Pearson's Chi-squared test,1.96E-09
distant_metastasis_pathologic_spread,factor,29,Pearson's Chi-squared test,3.77E-05
days_to_last_followup,integer,42,Kruskal-Wallis rank sum test,7.68E-05
vital_status,factor,23,Pearson's Chi-squared test,2.37E-03
year_of_tobacco_smoking_onset,integer,116,Kruskal-Wallis rank sum test,3.12E-03
year_of_tobacco_smoking_cessation,integer,88,Kruskal-Wallis rank sum test,5.84E-03
days_to_last_known_alive,integer,75,Kruskal-Wallis rank sum test,7.37E-03
residual_tumor,factor,46,Pearson's Chi-squared test,2.00E-02
lymphnode_pathologic_spread,factor,23,Pearson's Chi-squared test,5.48E-02
age_at_initial_pathologic_diagnosis,integer,30,Kruskal-Wallis rank sum test,9.24E-02
days_to_birth,integer,30,Kruskal-Wallis rank sum test,9.73E-02
Batch vs survival
Again, for this type of cancer clinical traits file contains days to last known alive but it has more NAs than days to the last follow up so I will use the latter for construction of the survival object.
Call:
coxph(formula = survivalObject ~ batchVector)
n= 223, number of events= 92
coef exp(coef) se(coef) z Pr(>|z|)
batchVector0848 -0.14217 0.86748 0.37975 -0.374 0.70813
batchVector0979 -0.23685 0.78911 0.42661 -0.555 0.57877
batchVector1096 1.66699 5.29619 0.60925 2.736 0.00622 **
batchVector1198 -0.13837 0.87077 0.42245 -0.328 0.74325
batchVector1440 -0.25689 0.77345 0.38754 -0.663 0.50741
batchVector1633 0.27021 1.31025 0.37760 0.716 0.47423
batchVector1818 -0.21253 0.80853 0.46395 -0.458 0.64688
batchVector1947 -0.05172 0.94959 0.48598 -0.106 0.91524
batchVector2043 0.06161 1.06355 1.03684 0.059 0.95261
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
exp(coef) exp(-coef) lower .95 upper .95
batchVector0848 0.8675 1.1528 0.4121 1.826
batchVector0979 0.7891 1.2672 0.3420 1.821
batchVector1096 5.2962 0.1888 1.6046 17.481
batchVector1198 0.8708 1.1484 0.3805 1.993
batchVector1440 0.7735 1.2929 0.3619 1.653
batchVector1633 1.3102 0.7632 0.6251 2.746
batchVector1818 0.8085 1.2368 0.3257 2.007
batchVector1947 0.9496 1.0531 0.3663 2.462
batchVector2043 1.0636 0.9402 0.1394 8.116
Rsquare= 0.041 (max possible= 0.973 )
Likelihood ratio test= 9.36 on 9 df, p=0.4044
Wald test = 12.53 on 9 df, p=0.1849
Score (logrank) test = 15.53 on 9 df, p=0.07747
On overall, correlation of batch with survival is not significant. There is one batch (1096) that seems to be somewhat more involved and it has only 11 patients. When I removed all patients from that batch no other batches showed completely insignificant correlation with survival.
DNA methylation
27k (normal Level 1 format: methylated intensities are in the first column, unmethylated intensities are in the forth column), convert to M value, didn't slit into the red and green. 134 patients, matched to the technical clinical information 133 patients. Work with them. SVD:
After fixing the inconsistencies in the concentration column (I had 0.13 ug/uL and .13 ug/uL and like for all other concentrations), converting concentration and plate_column to factors here is the summary of the technical variables:
> summary(tech)
batchID amount concentration plate_column plate_row
0689:22 13.3 uL:97 0.13 ug/uL:11 1:37 A :19
0848:46 26.7 uL:36 0.14 ug/uL:35 2:35 B :19
0979:36 0.15 ug/uL:59 3:26 C :18
1096:11 0.16 ug/uL:23 4:11 D :16
1198:18 0.17 ug/uL: 5 5:16 G :16
6: 8 E :15
(Other):30
shortDay
10-3-2010 :46
14-7-2010 :11
18-11-2009:22
30-8-2010 :18
5-5-2010 :36
Correlation of the first 6 PCs with the tech variables:
> x
batchID amount concentration plate_column plate_row shortDay
V1 1.219915e-14 1.512302e-15 0.22695068 0.004764525 0.5279767 1.219915e-14
V2 1.906561e-01 4.091184e-01 0.96296059 0.292069252 0.1656324 1.906561e-01
V3 7.464626e-02 6.984062e-02 0.22476713 0.467718779 0.5705236 7.464626e-02
V4 2.670836e-01 7.040891e-01 0.78009461 0.638345228 0.2242532 2.670836e-01
V5 5.721394e-01 7.689742e-01 0.42172107 0.348612502 0.6017037 5.721394e-01
V6 2.122370e-13 9.075951e-02 0.07250977 0.015916359 0.7132870 2.122370e-13
Start by removing the batch:
This looks strange. I never looked at the data distribution after normalization. Is it ok?
Correlation with the technical variables:
> x
batchID amount concentration plate_column plate_row shortDay
V1 0.9882080 0.9838389 0.9665557 0.45086339 0.08627734 0.9882080
V2 0.9975700 0.8632969 0.7730067 0.60414536 0.57298681 0.9975700
V3 0.9998216 0.9394499 0.3675916 0.02054405 0.88934477 0.9998216
V4 0.9994017 0.9313953 0.8917591 0.77577312 0.15250595 0.9994017
V5 0.9993497 0.9959595 0.2796044 0.45763689 0.62038204 0.9993497
V6 0.9766655 0.6485570 0.8141023 0.62257929 0.25956284 0.9766655
Removing batch fixed correlations with other technical variables. Consider data to be normalized?
ExpressionSet object is available.
