Extended Data Fig. 5: Mutational phasing for UV damage in mitotic sisters.

a, Mutation density for UV (upper) and ROS (lower) across all chromosomes. Heatmaps represent 10 mb sliding genomic windows with a 1 megabase step. Mutation density in windows is represented as mutations per 10 mb window. Each row is a single sister genome and rows are sorted by total UV mutation counts from highest to lowest. b, rl20 analysis (see Methods) as in Fig. 4a, but distinguishing between ROS mutations shared between mitotic sisters at time point 0 (dark blue) and ROS mutations unique to each individual sister, acquired after the first division (light blue). Red dots represent UV mutations. The y axis depicts the longest set of runs for a single mutation type, accounting for 20% of all informative mutations. The x axis is the significance in -log10(p-value) for seeing such a run length given random assignment to strands, calculated using a two-sided Wald-Wolfowitz runs test. c, Schematic depicting determination of background C > T mutation when modeling phasing for UV damage using Bernoulli trials. Upper box: Cells accumulate Non-UV C > T mutations in culture (yellow bars with red border) before UV damage. The ratio of C > T/ C > A for all 7 sister pairs is shown in the boxplot inset (Boxplot elements are as described in Fig. 3c, albeit without notches), and the average is ~ 0.3. Lower box: Total C > A mutation counts unique to each sister is multiplied by 0.3 to estimate the amount of total background C > T mutations (overlaid yellow bars). This background C > T estimate is then divided by the total C > T mutation counts to estimate the error adjustment for phased Bernoulli trials, which has a mean of 11.8% and ranges from 5% to 16% depending on UV total mutations. d, Error rate is used to adjust success or failure probability for completely phased segments. An example error rate of 12% is shown, whereby each phased segment probability is adjusted by 6%, as half of these background mutations will be randomly out of phase.