notes.txt

debruijn graph is MultiDigraph (can have multiple overlapping edges), where each edge is a kmer representing an overlap between two kmers, and the 
nodes are k-1mers. 


"even" edges (same out and in-degree) allow one to visit a node without getting stuck. 
Eulerian walk visis each edge once 

given a string, find all kmers.
for each kmer, create two nodes (k-1)mers, and edge (the kmer). 
Each node is unique, but a given pair may have multiple edges (signifying repeated kmers)

Eulerian walk may be non-unique.  (because of repeats), making it impossible to accurately reconstruct the original string. (order the kmers)
gaps or uneven coverage (more of one kmer than another) can cause problems. (missing nodes, unbalanced nodes)
so, De Bruijn assemblies give up on unsolvable repeats, and yield fragmented assembiles.

instead of treating reads as kmers, kmers are built independently from each read. 

kmers are very often repeated -- number of kmers is stored as weight of edge. 
constructing De Bruijn graph is O(N); walk is O(N)


Overlap Layout Consensus assembly

BWA is like bowtie (uses Burrows-Wheeler transform), but allows indels