this is for holding javascript data
Friederike Dündar details of deseq size factor
almost 9 years ago
Commit id: d93471de9a9492dc710310037b0e4c186d389371
deletions | additions
diff --git a/section_Normalization_Methods_The_number__.tex b/section_Normalization_Methods_The_number__.tex
index 167f358..e5897e1 100644
--- a/section_Normalization_Methods_The_number__.tex
+++ b/section_Normalization_Methods_The_number__.tex
...
%-----------------------
Total Count &
$\frac{gene\,read\,count}{total\,read\,number}$ &
inappropriate if some genes are
are only expressed in one condition unique to or extremely highly expressed
\\ in one condition, but not the other\\
%\tabularnewline \midrule
%-----------------------
Trimmed Mean of M-values (TMM) &
1. calculate gene-wise $log_2$ fold changes (= M-values): $M_g = \frac{log_2( \frac{obs.\,gene\,count_1}{total\,read\,number_1} )}{log_2( \frac{obs.\,gene\,count_2}{total\,read\,number_2} )} $; 2. trimming: removal of upper and lower 30\%; 3. precision weighing: the inverse of the estimated variance is used to account for lower variance of genes with larger counts
%$log_2(TMM^r_{gk}) = \frac{\displaystyle\sum_{g \in G^*} w^r_{gk}M^r_{gk}}{\displaystyle\sum_{g \in G^*} w^r_{gk}} $
&
more robust than total count normalization; details in \citet{RobinsonOshlack2010}\\
%\tabularnewline \midrule
%-----------------------
DESeq DESeq's size factor &
1. for each gene, the geometric mean of read counts across all samples is calculated; 2. every gene count is divided by the geometric mean; 3. for each sample, the median of these ratios (skipping the genes with a geometric mean of zero) is used as the sample's size factor &
\\ more robust than total count normalization; details in \citet{Anders2010}\\
%\tabularnewline \midrule
%-----------------------
Upper quartile & & \\