Defindit Docs and Howto Home

This page last modified: Feb 28 2007
title:OMSSA Hints and Howto
description:How to choose and determine options for OMSSA and omssacl

Table of contents
How to convert average mass to OMSSA mods
Choosing values for "te", "to", "tez", "mm", etc.
Other factors that may effect mass tolerance
Fixed and variable modifications


Choosing the best values for OMSSA options can be a challenge. With
the help of several people from various institutions, I've put
together some notes. Many, many thanks to all who contributed.

How to convert average mass to OMSSA mods

The problem: your MS/MS lab gives you a data set and says "We did this
run with a static modification on Cys of 57.0215." This translates to
an average mass of Cysteine of 57.0215. You can enter this mass value
into a couple of web sites and determine the likely mod is
Carbamidomethyl C. This you can look up in the OMSSA mods list.

average mass >= 57.0215

(Registration not necessary. Click "login as guest".)

It provides all average masses, and some monoisotopic masses when you
click on the highlighted modifications.

Choosing values for "te", "to", "tez", "mm", etc.

Many of the parameters are chosen based on the instrument and data
type, although there don't seem to be hard and fast rules.  For
instance, a no-enzyme search which contains many potential
modifications and sites requires a larger "mm" setting than a tryptic
search with a limited number of possible mods.  

The mass tolerances change based on the type of instrument.  Some labs
have spent a lot of time coming up with sets of parameters, and have
multiple sets depending upon data type.

A precursor mass tolerance "te" of 0.75 (this is +/-) is very high (a
wide range) when searching high resolution ms data. I've talked to
labs that normally stay between the values of 0.01 and 0.06 for this
parameter.  On the other hand, a value of 0.75 is very low when are
searching low resolution data.  For low resolution ms data, a typical
value is in the range of 1.0 to 1.5.  Another recommendation is to set
the "tez" parameter to 0 (zero). This disables the charge-dependent
changes to the precursor mass tolerance value.  I found (by trial and
error) that a product mass tolerance "to" of 0.15 worked for a set of
spectra (where "worked" means 0.15 yielded more peptides than other
values for this setting). However, I'm told that this product mass
tolerance could be too narrow. A typical value is somewhere in the
range 0.30 to 0.50. I'm not clear why setting the value of "to" to a larger
number (seemingly more relaxed) gives fewer peptides.

An experienced person has told me that a bigger precursor mass
tolerance "te" might give more significant hits since more potential
matches will be available. The concern would be that some of these
hits were false positives. Oddly, my personal experience is that
values outside some optimal range yield fewer hits.

Other factors that may effect mass tolerance

A friend tells me that -te and -to are very much dependent on which
mass spectrometer you use.  I don't mean to cast aspersions on the
MS/MS scientists, but other important factors in mass tolerance relate
to the MS/MS calibration:

1) How well the MS/MS was calibrated

2) How long ago the MS/MS was calibrated

3) How much the temperature in the room has changed since the MS/ME was calibrated

4) Other unspecified factors

This is one area of possible improvement for the current generation of
search tools. These tools could sample the spectra, and from that
sample determine settings, especially mass accuracy. This would be a
good idea because the mass accuracy values can be highly variable. In
the near term, I may implement a crude work around by implementing an
automated software pipeline that runs a search (such as OMSSA) many
times with a range of settings. The intent is to find the "best"
settings. Here are some rough suggestions for well-calibrated

Ion trap -to 3.0 -te 0.4

Linear ion trap -to 1.5 -te 1.0

Q-TOF -to 0.2 -te 0.2

Fixed and variable modifications

Fixed modifications are better than variable modifications when
possible because it limits the range of possible matches to be
searched. In other words, the search will be faster and there may be
fewer incorrect matches (bad hits, false positives). The speed
difference can be significant. The computational overhead is less of
an issue for many of my searches because I have the luxury of a
cluster with 30 dual-cpu nodes and I have a high-throughput pipeline
that will automataically split, run in parallel, and join results
(this pipeline is open source, by the way). I have not done a direct
comparison with the same modification fixed versus variable to confirm
that search results are the same.