How to dissect the tumors

1. check batch effect in SGIIA
(1) Heatmap, PCA on control genes
(2) Define the batches by global PCA
2. ComBat
3. CC_IFS on SGIIA
4. Select samples with avg_consensus_idx > 0.9
5. Use samples from step 4, and combine with SGIIB samples, repeat step 2-4
6. For new cohorts, repeat step 5.

limma to derive pairwise siganture, use NTP or SVM to predict.

How to include the maximum number of tumors

How we have 201 tumor samples which have avg_consensus_idx > 0.9 in CC_IFS of ComBat248, and the new ComBat201 has cophenetic = 1.
Thus we can lower the avg_consensus_idx cutoff to include more samples until cophenetic < 1.


24% of M patients benifit from chemo.
30% of D patients may benifit from PI3K?
46% treatment unknow?

tophat

  104   nohup tophat -r 124 -o tophat124  --num-threads=4 hg19 AGS_1_sequence.fastq  AGS_2_sequence.fastq >screen.txt &

The estimated library average fragment size is 280, the read length is 60bp, so the inner distange between paired reads (--mate-inner-dist) is 160.

124=fragment size(insert size) - 2*read length