Figure 2.

A new divide-and-conquer (DAC) strategy outperforms RP in hollow viscera tumors, detecting better intratumor heterogeneity (ITH) and equally well budding. (A) For low values of budding density, DAC detects more ITH than RP (red vs blue lines). DAC and RP performed equally well in detecting budding (magenta vs red lines). (B) Similar to panel (A), but intermediate values of budding density. (C) Similar to panel (A), but high values of budding density. (A–C) Percentage of either ITH or budding (B) detection (mean ± SE) as a function of the percentage of ITH density defining for each tumor. SE was calculated across N different repetitions of the same strategy and across M different tumors. Simulations parameters: L = 30 (side of 3 × L rectangle), H (number of sites with ITH) varying from 1 to 80 (or equivalently the ITH density ρ varying from approximately 0 to 89%), B (number of sites with budding) varying from 1 to 30 (or equivalently the budding density σ varying from approximately 0 to 100%), N = 50 (repetitions number for the two RP and DAC strategies), and M = 15 (number of simulated tumors). For the two strategies, RP and DAC, the total number of blocks for each repetition was equal to Q, which as explained in our previous approach, was modeling the laboratory costs. Hereon, we chose Q = 9 for both DAC and RP. For DAC, the Q sites consisted in three different parallel tissue stripes chosen at random (i.e., occupying i = 1, i = 2, and i = 3 sites and three random j’s, from j = 1, …, L). For RP, we first chose a site J randomly between 2 and L − 1, and after the sites (i,J), (i,J + 1), and (i,J − 1) with i = 1, 2, and 3.