Fig. 1

a Alteration matrices illustrating differences between the combinatorial weight function W(M) introduced in Dendrix and the probabilistic score Φ(M) used in CoMEt. Both matrices contain 4 mutually exclusive alterations whose alteration frequencies are indicated inside each bar. The samples without alterations are not shown in either matrix. Since both sets are exclusive and have the same total alteration frequency, the Dendrix weight function does not distinguish between these sets. Sets like M (blue) are common in cancer genome studies which often have a small number of recurrently mutated genes and a long tail of rarely mutated genes. The score used in CoMEt conditions on the observed frequencies of each alteration, giving more significance to the set M ^′ (green). b An example of 2×2×2 contingency table X _M for the set M={m ₁,m ₂,m ₃}, illustrating how samples are cross-classified into exclusive, co-occurring, or absent for each alteration. The test statistic ϕ(M) used by CoMEt is the sum of the highlighted exclusive cells