XB-IMG-148943
Xenbase Image ID: 148943
|
Fig S2 โ Statistics of Absolute Normalization, related to Fig 1, Supplemental
Experimental Procedures
(A) Absolute normalization of RNA standards. Left: relative normalization in TPM on linear
and log scale for the three datasets. The trend shared by all spikes is clearly visible and
differs between PolyA+ and rdRNA samples. Right: absolute normalization in transcripts per
embryo on log and linear scale. Trend in spikes is removed, the sample noise of the spikes
is retained and does not contaminate the absolute normalization of native transcripts.
(B) Absolute normalization factors. Circles mark per sample factors, lines mark Gaussian
Process smoothed factors. Left: Estimates of ๐ฝ๐ (Supplemental Experimental Procedures),
for three datasets. Center-left: per dataset correction factors eโ๐ฝ๐, note discrepancy in
magnitude but not in trend between datasets. Center-right: correction for PolyA+ bias, and
averaging Clutch A and Clutch B spike levels results in excellent agreement between the
correction factors. Right: Smoothed correction factors calculated with non-stationary
Gaussian Process (Supplemental Experimental Procedures).
(C) Absolute normalization consistency and accuracy. Scatter plots of ERCC standards
spiked vs calculated for Clutch A PolyA+, Clutch A rdRNA, Clutch B PolyA+ and Clutch
A/Clutch B PolyA+ combined. Points are averaged over all samples in time to remove spikein
sample noise. ArrayControl spikes are given with fold change errors for Clutch B PolyA+,
fold changes range from 1.11-1.25. ArrayControl spikes EC02 and EC12 have are spiked in
with the same copy number and are marked with the same vertical line.Combined Clutch
A/Clutch B gives 95% confidence interval for a linear model with Gaussan noise, with slope
unity and with standard deviation ๐๐ = 0.25. Note that residuals do not depend on expression
level.
(D) Absolute normalization uncertainty model. Left: model as described in C (far-right), with
25%, 50%, 75%, 95% confidence intervals for 106 transcripts. Right: Propagation of
uncertainty model to absolute normalization of eef1a1o Clutch A polyA+, again 25%, 50%
75% and 95% of confidence intervals for the true number of eef1a1o transcripts/embryo
marked.
(E) RNA-seq detection limits โ Transcript abundance required to produce a single read with
time, detection limits increase as RNA content in the embryo increases. Top row: Detection
limits in kb, points give the per sample detection limits, lines and shaded area give Gaussian
Process median and 95% confidence interval. Bottom: Detection limit in transcripts/embryo,
calculated by averaging kb Gaussian Process distribution over all transcript lengths.
(F) The depletion of non-ribosomal RNAs in rdRNA sequencing. Comparisons between
Clutch A PolyA+ and rdRNA sequencing (top). Examples of extreme depletion: mixer and
foxh1.2. Image published in: Owens ND et al. (2016) Copyright ยฉ 2016. Image reproduced with permission of the publisher and the copyright holder. This is an Open Access article distributed under the terms of the Creative Commons Attribution License. Larger Image Printer Friendly View |