#MultiGeMS 1.0
##Overview
Multi-sample Genotype Model Selection (MultiGeMS) is a multiple sample single nucleotide variant (SNV) caller that works with alignment files of high-throughput sequencing (HTS) data. MultiGeMS calls SNVs based on a statistical model selection procedure and accounts for enzymatic substitution sequencing errors.
##Compile
MultiGeMS can be compiled using a GCC compiler with C++11 support, by running:
$ make
##Input
MultiGeMS accepts a text file, listing on seperate lines, paths of SAMtools pileup format files. To convert a SAM/BAM alignment file into the pileup format, users can use the SAMtools mpileup procedure with option -s.
##Filter
Alignment file reads with undesirable characteristics can be filtered before running MultiGeMS. For an explanation why filtering may be desirable and for a brief tutorial on how to filter SAM/BAM alignment files using SAMtools view, please see the PDF document entitled "Filtering Alignment Files" available at https://github.com/cui-lab/multigems.
##Usage
multigems -i pileuplist.txt -o multigems.out [OPTIONS]
-b INT minimum base-calling quality score considered, default is 17
-m INT minimum mapping quality score considered, default is 20
-s FLOAT maximum likelihood computing steps float value, smaller is slower yet more precise, default is 0.01
-e FLOAT EM algorithm convergence threshold, smaller is slower yet more precise, default is 0.001
-M INT maximum number of bases to be considered from each sample of each site, 0 indicates unbounded, default is 255
-f FLOAT lFDR SNV threshold, between 0 and 1, default is 0.1
-C INT number of sites to be analyzed per analysis cycle, smaller is slower in general and uses less RAM, default is 200
-t INT number of threads, can be used in conjunction with -C option for multi-thread efficiency, default is 1
-n FLOAT site non-reference allele proportion filter (sites where all samples have a non-reference proportion less than this filter are not analyzed), smaller is slower and more susceptible to false positive SNV calls, higher is faster and more susceptible to false negative SNV calls, between 0 and 1, default is 0.2
-l FLOAT sample deletion placeholder proportion filter (samples with pileup deletion placeholder proportions greater than filter are not analyzed), between 0 and 1, default is 0.05
##Output
The MultiGeMS output is similar to that of the Variant Call Format (VCF) file format. Meta-information lines are provided at the beginning of each output. Only sites less than the user-selected lFDR SNV threshold are output and are considered SNV calls. An lFDR output of "-1" indicates that the site was analyzed by MultiGeMS, but not enough information was retained post-filtering to estimate the lFDR value for the site.
##Contact
Xinping Cui
https://sites.google.com/a/bioinformatics.ucr.edu/xinping-cui/