Figure 1
Schematic of the ISOpure algorithm. ISOpure is a two-step algorithm for computational purification. (A) As input, it takes a set of tumor profiles t1, t2,..., t N and a set of normal profiles b1, b2,..., b R . It then rounds the component values of each t n to compute profiles x n . ISOpure uses the set of normal profiles (green triangle) to estimate the total possible variation in expression due to normal tissue contamination in each tumor sample. (B) In Step 1, ISOpure estimates a shared, representative cancer profile m, and the proportion, α n , of each tumor's mRNA contributed by the cancer cells. (C) In Step 2, ISOpure uses m as the mean of a common Dirichlet prior for each individual cancer profile c n learned for each input tumor sample t n , and also estimates h n , the profile of the contaminating normal tissue in sample n.