...
Furthermore, a fold change plot is given (Figure 3), comparing the estimated and expected fold change in Mix A vs. Mix B, for the four groups of transcripts with the known fold change values. The fold change values were calculated based on log2 transformed RPKM values for each transcript (control).
Numbered figure captions |
---|
SubtitleText | Comparison of expected number of molecules of ERCC Mix A and the estimated number of transcript molecules (log2 of RPKM values), obtained by Partek’s modification of expectation-maximization algorithm. Each dot is an ERCC control. R2 = 0.97, regression y = 0.98 * x - 74.41 |
---|
AnchorName | comp-expected |
---|
|
Image Modified |
Numbered figure captions |
---|
SubtitleText | Comparison of expected number of molecules of ERCC Mix B and the estimated number of transcript molecules (log2 of RPKM values), obtained by Partek’s modification of expectation-maximization algorithm. Each dot is an ERCC control. R2 = 0.97, regression y = 0.98 * x - 74.06 |
---|
AnchorName | comp-expected2 |
---|
|
Image Modified |
Numbered figure captions |
---|
SubtitleText | Comparison of expected fold change (log2) vs. observed fold change (log2) in ERCC Mix A vs. Mix B, for the four groups of transcripts with the known fold change values. Each dot is an ERCC control. R2 = 0.87, regression y = 0.96 * x + 0.09 |
---|
AnchorName | comp-expected3 |
---|
|
Image Modified |
To test for possible biases in abundance estimates, we combined the approaches of Li et al. (9) and Jiang et al. (12). That amounted to regressing the estimated abundance not only on the expected abundance, but also on the transcript length, GC content, and the expected fold change, subjecting all the variables to log2 transformation.
...