Algorithm 1.

HapCompass-Tumor

input : Sequence reads, variant calls, and number of distinct haplotypes k

output: k haplotypes

GC ← spanning tree cycle basis

CC ← set of conflicting simple cycles with respect to GC

for cC ∈ CC do

Remove edge with smallest likelihood in cC

Reconstruct GC

end

Compute Gg

CN ← set of non-conflicting simple cycles with respect to GC

for cN ∈ CN do

Compute Gh with respect to cN

Compute matchings at each level of Gh

Compute disjoint paths of Gh

Increase the weight of each edge e between levels shared by Gh and Gg in Gg

proportional to the number of disjoint paths using edge e and the likelihood of

each edge (Equations 1 and 2)

end

Compute a maximum weight spanning tree of the adjacencies in Gg

Output the haplotype assembly computed from the spanning tree of Gg