Cell - 8 September 2016

Figure S2. Noise Analysis for Batch 1, Related toFigure 2
The first two rows of panels show histograms of scaled deviations Zifor haploid and diploid lineages over all cycles in replicate 1, including the predicted
distribution from the noise model (red) and counting noise limit (black). Typical read counts are 100 at the beginning of the experiment and 30 by the end. The last
row shows the SD of log-slopedsivs. read number across all replicates. We expect the variance to scale as r
^1. Scaling estimates are shown from within replicate
(red), between replicate (blue), and counting noise (black). We find that noise is independent of read count at high frequency (green) and approaches the minimal
limit at late times.