EMIRGE reconstructs full length ribosomal genes from short read sequencing data. In the process, it also provides estimates of the sequences' abundances.
EMIRGE uses a modification of the EM algorithm to iterate between estimating the expected value of the abundance of all SSU sequences present in a sample and estimating the probabilities for each read that a specific sequence generated that read. At the end of each iteration, those probabilities are used to re-calculate (correct) a consensus sequence for each reference SSU sequence, and the mapping is repeated, followed by the estimations of probabilities. The iterations should usually stop when the reference sequences no longer change from one iteration to the next. Practically, 40-80 iterations is usually sufficient for many samples. Right now EMIRGE uses Bowtie alignments internally, though in theory a different mapper could be used.
EMIRGE was designed for Illumina reads in FASTQ format, from pipeline version >= 1.3
There are two versions of EMIRGE:
- emirge.py -- this version was designed for metagenomic data
- emirge_amplicon.py -- this version was designed for PCR amplicon data, and can handle up to a few million reads where the entire sequencing allocation is devoted to a single gene. In theory it could also be used for RNASeq data where rRNA makes up a large percentage of the reads.
- Validated 4/5/2018