Supplementary Figures 6 - 11 from Single-Molecule Genomic Data Delineate Patient-Specific Tumor Profiles and Cancer Stem Cell Organization

PDF file - 63K, Figure S6. Inferred parameters G and a for ZNF454 and SLC5A7. For the transit amplifying stages parameter G and the apoptosis rate a, we report agreement of multiple molecular clocks. Figure S7. Heterogeneity induced by different gland splitting algorithms. In our model we implemented the gland fission mechanism as the equal division of the cell population into the two daughter cells. Although this is the most commonly advocated method of gland splitting, an alternative scenario is to consider a single founding cell for the new gland. We predict that this would substantially increase the level of heterogeneity in the glands, due to the increased number of cell divisions required by the founding CSC to grow the new gland. We do not consider this alternative process in our analysis. Figure S8. Prior distributions used in SCAI. We used non-informative priors for our analysis. However, some parameter values are less likely to yield a tumor. For example, a high apoptotic rate may completely extinguish the CSC population when this is small enough. For this reason, the symmetric division rate and apoptosis rate are not taken to be uniform. The tumor age is a secondary parameter that results from the combination of all the other parameters. Figure S9. Validation of the SCAI framework with an 8 CpG-long synthetic dataset. Using synthetic dataset with a molecular clock made by 8 CpGs produces very similar results to a synthetic dataset with 16 CpGs-long clocks. Figure S10. Patient-specific inference results for a single copy of IRX2. Inference results considering a haploid version of the IRX2 locus in each cell (e.g. due to copy number loss), are very similar to the results obtained by considering 2 copies of IRX2 per cell. This confirms the robustness of our analysis to copy number alterations. Moreover copy number changes of the IRX2 locus in CRC are reported in an extremely limited number of cases (~0.3%). Figure S11. Expected variation in summary statistics for different sequencing depths. The simulation of different sampling depths from the same gland reveals a considerable difference in summary statistics, particularly for the number of patterns, singletons, Kolmogorov distance and Shannon index.