OMSSA was developed by Lewis Geer at NCBI/NIH. This page is my OMSSA help page resource. Given that all parameters are specified on the omssacl command line, I wanted a single page where they are all listed for easy reference. The help text for OMSSA version 2.1.9 are listed below. Actually the download unpacks a 2.1.9 directory but the version output states 2.1.8. I'll assume they forgot to update the version string in the linux binary as the other distributions also unpack to a 2.1.9 directory.
Unfortunately OMSSA is no longer being supported at NIH.
omssacl -e 16 -i 1,4 -mf 3 -mv 1 -tem 0 -tom 0 -te 1.0 -to 0.5 -tez 1 -he 100000.0 -zcc 1 -hl 5
USAGE omssacl [-h] [-help] [-xmlhelp] [-pm param] [-d blastdb] [-umm] [-f infile] [-fx xmlinfile] [-fb dtainfile] [-fp pklinfile] [-fm pklinfile] [-foms omsinfile] [-fomx omxinfile] [-fbz2 bz2infile] [-fxml omxinfile] [-o textasnoutfile] [-ob binaryasnoutfile] [-ox xmloutfile] [-obz2 bz2outfile] [-op pepxmloutfile] [-oc csvfile] [-w] [-to pretol] [-te protol] [-tom promass] [-tem premass] [-tez prozdep] [-ti isotopes] [-teppm] [-ta autotol] [-tex exact] [-i ions] [-cl cutlo] [-ch cuthi] [-ci cutinc] [-cp precursorcull] [-v cleave] [-x taxid] [-w1 window1] [-w2 window2] [-h1 hit1] [-h2 hit2] [-hl hitlist] [-hc hitcount] [-ht tophitnum] [-hm minhit] [-hs minspectra] [-he evalcut] [-mf fixedmod] [-mv variablemod] [-mnm] [-mm maxmod] [-e enzyme] [-zh maxcharge] [-zl mincharge] [-zoh maxprodcharge] [-zt chargethresh] [-z1 plusone] [-zc calcplusone] [-zcc calccharge] [-zn negions] [-pc pseudocount] [-sb1 searchb1] [-sct searchcterm] [-sp productnum] [-scorr corrscore] [-scorp corrprob] [-no minno] [-nox maxno] [-is subsetthresh] [-ir replacethresh] [-ii iterativethresh] [-p prolineruleions] [-il] [-el] [-ml] [-mx modinputfile] [-mux usermodinputfile] [-nt numthreads] [-ni] [-ns] [-os] [-nrs] [-logfile File_Name] [-conffile File_Name] [-version] [-version-full] [-dryrun] DESCRIPTION Search engine for identifying MS/MS peptide spectra OPTIONAL ARGUMENTS -h Print USAGE and DESCRIPTION; ignore other arguments -help Print USAGE, DESCRIPTION and ARGUMENTS description; ignore other arguments -xmlhelp Print USAGE, DESCRIPTION and ARGUMENTS description in XML format; ignore other arguments -pm <String> search parameter input in xml format (overrides command line) Default = `' -d <String> Blast sequence library to search. Do not include .p* filename suffixes. Default = `nr' -umm use memory mapped sequence libraries -f <String> single dta file to search Default = `' -fx <String> multiple xml-encapsulated dta files to search Default = `' -fb <String> multiple dta files separated by blank lines to search Default = `' -fp <String> pkl formatted file Default = `' -fm <String> mgf formatted file Default = `' -foms <String> omssa oms file Default = `' -fomx <String> omssa omx file Default = `' -fbz2 <String> omssa omx file compressed by bzip2 Default = `' -fxml <String> omssa xml search request file Default = `' -o <String> filename for text asn.1 formatted search results Default = `' -ob <String> filename for binary asn.1 formatted search results Default = `' -ox <String> filename for xml formatted search results Default = `' -obz2 <String> filename for bzip2 compressed xml formatted search results Default = `' -op <String> filename for pepXML formatted search results Default = `' -oc <String> filename for csv formatted search summary Default = `' -w include spectra and search params in search results -to <Real> product ion m/z tolerance in Da Default = `0.8' -te <Real> precursor ion m/z tolerance in Da (or ppm if -teppm flag set) Default = `2.0' -tom <Integer> product ion search type (0 = mono, 1 = avg, 2 = N15, 3 = exact) Default = `0' -tem <Integer> precursor ion search type (0 = mono, 1 = avg, 2 = N15, 3 = exact, 4 = multiisotope) Default = `0' -tez <Integer> charge dependency of precursor mass tolerance (0 = none, 1 = linear) Default = `0' -ti <Integer> when doing multiisotope search, number of isotopic peaks to search. 0 = monoisotopic peak only Default = `0' -teppm search precursor masses in units of ppm -ta <Real> automatic mass tolerance adjustment fraction Default = `1.0' -tex <Real> threshold in Da above which the mass of neutron should be added in exact mass search Default = `1446.94' -i <String> id numbers of ions to search (comma delimited, no spaces) Default = `1,4' -cl <Real> low intensity cutoff as a fraction of max peak Default = `0.0' -ch <Real> high intensity cutoff as a fraction of max peak Default = `0.2' -ci <Real> intensity cutoff increment as a fraction of max peak Default = `0.0005' -cp <Integer> eliminate charge reduced precursors in spectra (0=no, 1=yes) Default = `0' -v <Integer> number of missed cleavages allowed Default = `1' -x <String> comma delimited list of taxids to search (0 = all) Default = `0' -w1 <Integer> single charge window in Da Default = `27' -w2 <Integer> double charge window in Da Default = `14' -h1 <Integer> number of peaks allowed in single charge window (0 = number of ion species) Default = `2' -h2 <Integer> number of peaks allowed in double charge window (0 = number of ion species) Default = `2' -hl <Integer> maximum number of hits retained per precursor charge state per spectrum during the search Default = `30' -hc <Integer> maximum number of hits reported per spectrum (0 = all) Default = `0' -ht <Integer> number of m/z values corresponding to the most intense peaks that must include one match to the theoretical peptide Default = `6' -hm <Integer> the minimum number of m/z matches a sequence library peptide must have for the hit to the peptide to be recorded Default = `2' -hs <Integer> the minimum number of m/z values a spectrum must have to be searched Default = `4' -he <Real> the maximum evalue allowed in the hit list Default = `1' -mf <String> comma delimited (no spaces) list of id numbers for fixed modifications Default = `' -mv <String> comma delimited (no spaces) list of id numbers for variable modifications Default = `' -mnm n-term methionine should not be cleaved -mm <Integer> the maximum number of mass ladders to generate per database peptide Default = `128' -e <Integer> id number of enzyme to use Default = `0' -zh <Integer> maximum precursor charge to search when not 1+ Default = `3' -zl <Integer> minimum precursor charge to search when not 1+ Default = `1' -zoh <Integer> maximum product charge to search Default = `2' -zt <Integer> minimum precursor charge to start considering multiply charged products Default = `3' -z1 <Real> fraction of peaks below precursor used to determine if spectrum is charge 1 Default = `0.95' -zc <Integer> should charge plus one be determined algorithmically? (1=yes) Default = `1' -zcc <Integer> how should precursor charges be determined? (1=believe the input file, 2=use a range) Default = `2' -zn <Integer> search using negative or positive ions (1=positive, -1=negative) Default = `1' -pc <Integer> minimum number of precursors that match a spectrum Default = `1' -sb1 <Integer> should first forward (b1) product ions be in search (1=no) Default = `1' -sct <Integer> should c terminus ions be searched (1=no) Default = `0' -sp <Integer> max number of ions in each series being searched (0=all) Default = `100' -scorr <Integer> turn off correlation correction to score (1=off, 0=use correlation) Default = `0' -scorp <Real> probability of consecutive ion (used in correlation correction) Default = `0.5' -no <Integer> minimum size of peptides for no-enzyme and semi-tryptic searches Default = `4' -nox <Integer> maximum size of peptides for no-enzyme and semi-tryptic searches (0=none) Default = `40' -is <Real> evalue threshold to include a sequence in the iterative search, 0 = all Default = `0.0' -ir <Real> evalue threshold to replace a hit, 0 = only if better Default = `0.0' -ii <Real> evalue threshold to iteratively search a spectrum again, 0 = always Default = `0.01' -p <String> id numbers of ion series to apply no product ions at proline rule at (comma delimited, no spaces) Default = `' -il print a list of ions and their corresponding id number -el print a list of enzymes and their corresponding id number -ml print a list of modifications and their corresponding id number -mx <String> file containing modification data Default = `mods.xml' -mux <String> file containing user modification data Default = `usermods.xml' -nt <Integer> number of search threads to use, 0=autodetect Default = `0' -ni don't print informational messages -ns depreciated flag -os use omssa 1.0 scoring -nrs turn off rank score -logfile <File_Out> File to which the program log should be redirected -conffile <File_In> Program's configuration (registry) data file -version Print version number; ignore other arguments -version-full Print extended version data; ignore other arguments -dryrun Dry run the application: do nothing, only test all preconditions -i options: 0: a 1: b 2: c 3: x 4: y 5: zdot 10: adot 11: x-CO2 12: adot-CO2 -e enzyme options: 0: Trypsin 1: Arg-C 2: CNBr 3: Chymotrypsin (FYWL) 4: Formic Acid 5: Lys-C 6: Lys-C, no P rule 7: Pepsin A 8: Trypsin+CNBr 9: Trypsin+Chymotrypsin (FYWLKR) 10: Trypsin, no P rule 11: Whole protein 12: Asp-N 13: Glu-C 14: Asp-N+Glu-C 15: Top-Down 16: Semi-Tryptic 17: No Enzyme 18: Chymotrypsin, no P rule (FYWL) 19: Asp-N (DE) 20: Glu-C (DE) 21: Lys-N (K) 22: Thermolysin, no P rule 23: Semi-Chymotrypsin (FYWL) 24: Semi-Glu-C Modifications: # : Name 23: 2-amino-3-oxo-butanoic acid T 182: Asparagine HexNAc 183: Asparagine dHexHexNAc 131: CAMthiopropanoyl K 130: ICAT heavy 129: ICAT light 9: M cleavage from protein n-term 179: MMTS on C 119: N to D conversion 83: NEM C 84: NIPCAM 87: O18 on peptide n-term 139: PNGasF in O18 water 113: SeMet 184: Serine HexNAc 185: Threonine HexNAc 24: acetylation of K 10: acetylation of protein n-term 25: amidation of peptide c-term 163: arginine to ornithine 140: beta elimination of S 141: beta elimination of T 13: beta methythiolation of D 26: beta-methylthiolation of D 3: carbamidomethyl C 31: carbamylation of K 32: carbamylation of n-term peptide 29: carboxyamidomethylation of D 30: carboxyamidomethylation of E 28: carboxyamidomethylation of H 27: carboxyamidomethylation of K 165: carboxykynurenin of W 2: carboxymethyl C 33: citrullination of R 4: deamidation of N and Q 164: dehydro of S and T 88: di-O18 on peptide n-term 35: di-iodination of Y 36: di-methylation of K 37: di-methylation of R 38: di-methylation of peptide n-term 42: farnesylation of C 46: fluorophenylalanine 43: formylation of K 44: formylation of peptide n-term 82: formylation of protein n-term 47: gamma-carboxylation of D 48: gamma-carboxylation of E 49: geranyl-geranyl 50: glucuronylation of protein n-term 51: glutathione disulfide 53: guanidination of K 136: heavy arginine-13C6 137: heavy arginine-13C6-15N4 181: heavy lysine - 13C6 15N2 180: heavy lysine - 2H4 138: heavy lysine-13C6 56: homoserine 57: homoserine lactone 64: hydroxylation of Y 59: hydroxylation of D 63: hydroxylation of F 60: hydroxylation of K 61: hydroxylation of N 62: hydroxylation of P 168: iTRAQ114 on K 169: iTRAQ114 on Y 167: iTRAQ114 on nterm 171: iTRAQ115 on K 172: iTRAQ115 on Y 170: iTRAQ115 on nterm 174: iTRAQ116 on K 175: iTRAQ116 on Y 173: iTRAQ116 on nterm 177: iTRAQ117 on K 178: iTRAQ117 on Y 176: iTRAQ117 on nterm 65: iodination of Y 67: lipoyl K 73: methyl C 74: methyl H 75: methyl N 77: methyl R 69: methyl ester of D 70: methyl ester of E 71: methyl ester of S 72: methyl ester of Y 68: methyl ester of peptide c-term 16: methylation of D 17: methylation of E 0: methylation of K 14: methylation of Q 18: methylation of peptide c-term 76: methylation of peptide n-term 11: methylation of protein n-term 78: myristoleylation of G 79: myristoyl-4H of G 81: myristoylation of K 80: myristoylation of peptide n-term G 118: n-acyl diglyceride cysteine 22: n-formyl met addition 34: oxidation of C to cysteic acid 162: oxidation of C to sulfinic acid 39: oxidation of F to dihydroxyphenylalanine 89: oxidation of H 55: oxidation of H to D 54: oxidation of H to N 1: oxidation of M 111: oxidation of P to pyroglutamic acid 90: oxidation of W 45: oxidation of W to formylkynurenin 58: oxidation of W to hydroxykynurenin 66: oxidation of W to kynurenin 85: oxidation of W to nitro 86: oxidation of Y to nitro 92: palmitoylation of C 93: palmitoylation of K 94: palmitoylation of S 95: palmitoylation of T 91: phosphopantetheine S 6: phosphorylation of S 134: phosphorylation of S with ETD loss 96: phosphorylation of S with prompt loss 7: phosphorylation of T 135: phosphorylation of T with ETD loss 97: phosphorylation of T with prompt loss 8: phosphorylation of Y 99: phosphorylation with neutral loss on C 100: phosphorylation with neutral loss on D 101: phosphorylation with neutral loss on H 132: phosphorylation with neutral loss on S 133: phosphorylation with neutral loss on T 98: phosphorylation with prompt loss on Y 5: propionamide C 104: propionyl heavy K 105: propionyl heavy peptide n-term 102: propionyl light K 103: propionyl light on peptide n-term 106: pyridyl K 107: pyridyl peptide n-term 108: pyro-cmC 109: pyro-glu from n-term E 110: pyro-glu from n-term Q 112: s-pyridylethylation of C 114: sulfation of Y 115: sulphone of M 166: sumoylation of K 40: thioacylation of K 41: thioacylation of peptide n-term 19: tri-deuteromethylation of D 20: tri-deuteromethylation of E 21: tri-deuteromethylation of peptide c-term 116: tri-iodination of Y 15: tri-methylation of K 117: tri-methylation of R 12: tri-methylation of protein n-term 52: ubiquitinylation residue