Discovar denovo is by the Broad institute. It is a total pipeline, meaning that one needs to follow the sequencing as well as assembly protocols to obtain the best results. Most of the information is the in the PPT here. The notes below are supplemental.

Basic Outline:

• Builds stacks
• Filters out crap
• Builds consensus
• Trys to merge stacks
• Using pairwise windows. If it finds a match it checks adjacent windows for overlap before they link them
• Correct paired ends against each other
• Then build using HQ pairs

• Spri beads size selection.
• No amplification in adapter phase.
• Wants a single library on an illumia machine.

• Use sequential error correction increasing in difficulty to reduce computation
• Look for isolated loser calls which means no LQ scores adjacent and corrects the scores up.
• Reduce quality scores in a sliding window. Correct artifically high quality scores with adjacent low scores

• Use the good founder pairs from previous steps to “seed” building the genome.