BOL

Fermi is a de novo assembler with a particular focus on assembling Illumina short sequence reads from a mammal-sized genome. In addition to the role of a typical assembler, fermi also aims to preserve heterozygotes which are often collapsed by other assemblers. Its ultimate goal is to find a minimal set of
unitigs to represent all the information in raw reads.

Fermi follows the overlap-layout-consensus paradigm and uses the FM-DNA-index (FMD-index) as the key data structure. It is inspired by the string graph assembler (Simpson and Durbin, 2010 and 2012) and has a similar workflow.

As a typical de novo assembler, fermi tends to produce contigs with slightly longer N50. However, the major weakness of fermi is the high misassembly rate. Although fermi provides a tool to fix misassemblies by using paired-end reads to achieve an accuracy comparable to other assemblers, this is not a favorable solution.

Fermi is designed to be used on a multi-core Linux machine with large shared memory. The easiest way to run fermi is to use the run-fermi.pl script. It generates a Makefile. The actual assembly is done by invoking make. Premature assembly processes can be resumed. Here is an example:

run-fermi.pl -dAPe ./fermi -p NA12878 -t16 -f18 reads*.fq.gz > NA12878.mak
make -f NA12878.mak -j16