Crux Frequently Asked Questions
- What is the difference between Tide and Comet?
The differences are as follows:
Tide converts your FASTA file to an index (either as part
tide-searchor by using
tide-index), whereas Comet does searching directly from the FASTA file. The indexing step allows Tide to pre-compute many parts of the search procedure, thereby making the search faster.
- For each match, Comet reports the total number of candidate peptides, whereas Tide reports the number of distinct candidate peptides.
Comet provides an option to report an E-value in addition to the raw
XCorr score, whereas Tide provides options to report a variety of
different types of scores. The methods for computing Tide's
various scores are described in the documentation
- Some options are available only in Comet (various options related to different types of theoretical fragment ions, max_fragment_charge, nucleotide_reading_frame, etc.) whereas others are only available in Tide (e.g., keep-terminal-aminos).
Overall, for a given set of spectra, the XCorr scores computed by Comet and by Tide should be quite similar to one another, assuming that the various search parameters are set similarly between the two algorithms. Below is a table summarizing the correspondence between the two sets of parameters. The specified values will yield nearly identical XCorr scores from the two search engines.
|Tide parameter||Tide value||Comet parameter||Comet value|
|top-match||5||num_results, num_output_lines||6, 5|
|min-mass, max-mass||200, 7200||digest_mass_range||200 7200|
Here is a scatter plot of XCorr scores from a search run with the parameters listed above:
Crux is written in C++, so in principle it should work on virtually any modern operating system. We provide pre-compiled binaries for Linux, MacOS and Windows.
/tmp/crux: /usr/lib64/libstdc++.so.6: version `CXXABI_1.3.8' not found (required by /tmp/crux) /tmp/crux: /usr/lib64/libstdc++.so.6: version `GLIBCXX_3.4.20' not found (required by /tmp/crux) /tmp/crux: /usr/lib64/libstdc++.so.6: version `GLIBCXX_3.4.21' not found (required by /tmp/crux)p
Crux depends on the C and C++ runtime libraries; however, newer versions of the runtime libraries are not backwards compatible with older versions. We build the Linux binary distributions for Crux using GCC 4.9. If your Linux system dates from before GCC 4.9 was available, then it may have versions of the runtime library that are incompatible with the Crux binary distribution. In this case, you might consider updating your operating system. Alternatively, you can build Crux from the source code distribution. To build Crux from source you'll ned to install the C/C++ development tools, and CMake. This can conveniently be done using
sudo su - yum group install "Development Tools" yum install cmake
Then follow the instructions in the in the installation tutorial.
Each amino acid can be characterized either by its monoisotopic mass,
which is the mass of the most abundant isotopic form of that amino acid,
or the average mass, which is a weighted average of the masses of all
the isotopic forms. By default, Crux uses average mass to calculate
peptide masses, though this behavior can be controlled by the
comet. For fragments,
tide-searchalways uses the monoisotopic mass, whereas
cometallows selection via
- The neutral mass of a peptide is not the sum of the masses of its amino acids. The N-terminus and C-terminus of the peptide together contribute an additional water molecule, whose mass is either 18.010564684 Da (monoisotopic) or 18.0153 Da (average).
- A charged peptide has an additional approximately 1 Da mass for each charge, corresponding to the mass of a hydrogen atom. The exact mass to be added depends on whether we are using the monoisotopic mass (1.007825035) or average mass (1.00794).
- When a peptide fragments, each b-ion will have a mass equal to the sum of its amino acids plus one hydrogen for each charge on the fragment, while each y-ion will have a mass equal to the sum of its amino acids plus water plus one hydrogen for each charge on the fragment.
By default, both Tide and Comet add a static modification of
57.021464 Da to all cysteines. This is because in most protein
preparation protocols the peptides are alkylated with iodoacetamide,
resulting in carbamidomethylation of cysteine. The alkylation step
prevents the reformation of disulfide bonds. Other alykylation
reagents may be used, in which case the appropriate mass shift can be
specified with the
The XCorr score is essentially a dot product between a preprocessed
form of the observed spectrum and the theoretical spectrum derived
from the candidate peptide. In order to compute this dot product, the
masses of the fragments in both spectra are assigned to discrete mass
bins. This can be viewed as a form of rounding, but with more control
over the discrete masses. Two parameters,
mz-bin-offset, control the size and location of the bins,
and are used to convert fragment masses according to this formula:
binned mass = floor( ( original mass /
bin-width ) + 1.0 -
The default values
bin-offset=0.40 are suitable for most high-resolution
Uri Keich, Kaipo Tamura, and William Stafford Noble. "Averaging strategy to reduce variability in target-decoy estimates of false discovery rate." Journal of Proteome Research, 18(2):585-593, 2019.
When you build your peptide index using tide-index, you have to specify the number of decoy databases to include and you have to set allow-dups to be T. It's also important that you leave the decoy-format option at its default value of shuffle, since if you reverse the peptides then all of the databases would be identical. For example, to build an index with five decoys per target from a fasta file called
human.fasta, you could do this:
crux tide-index --num-decoys-per-target 5 --allow-dups T human.fasta human
Also, when you search a file of spectra (
HepG2_f4.mzML) using the
humanindex, you must specify concat=F:
crux tide-search --use-tailor-calibration T --concat F HepG2_f4.mzML human
This command stores the search results in the default output directory (
crux-output) in files called
tide-search.decoy.txt. Notably, the decoy file will contain five times as many PSMs as the target file, with a column called "decoy index" to indicate which database the decoy comes from. Finally, you can run aTDC by calling
assign-confidence. This command will use the "decoy index" column to do the necessary averaging over the decoy databases:
crux assign-confidence --score "tailor score" crux-output/tide-search.target.txt
The final results are stored in a file called
- When the combined p-value score function is used (
- When the XCorr score function is used (
--exact-p-valuecan only be set to true when
- When the residue-evidence score function is used (
--exact-p-valuecan only be set to true when
--mz-bin-width=1.0005079. Note that high-resolution information in the residue-evidence score function is captured by
Crux calculates the ratio of two values: (1) the sum of the intensities from the peaks in the fragmentation spectrum whose m/z is greater than the precursor m/z versus (2) the sum of the peaks whose m/z is smaller than the precursor. If this ratio is greater than or equal to a calculated threshold based upon the location of the precursor m/z and the max m/z in the spectrum, Crux then assigns both +2 and +3 as possible charge states to the spectrum. Otherwise, Crux assigns +1 as the charge state. The algorithm is based upon the observation that fragmentation spectra collected from +1 peptides should have no peaks above the precursor m/z. In contrast, a peptide of charge state greater than +1 can generate fragment ions of lower charge whose m/z is greater than the precursor m/z, thus indicating a multiply charged precursor.
Note that, in addition,
precursor_charge parameter. If the first number in
this parameter is set to 0, i.e.,
precursor_charge = 0 0,
then the charge state rule above is applied. However, if a user
specifies a precursor charge range, i.e.,
precursor_charge = 1
5, then Comet will search every spectrum through this range of
assumed charge states for every spectrum whose precursor charge
the enzymatic cleavage rules. When
search_enzyme_number=0, then any subsequence of a
protein may be considered as a candidate peptide. For other values of
these parameters, the residues at the termini of the protein
subsequence must follow specific rules. For example, trypsin requires
that the preceeding residue must be an R or K and the following
residue may not be a P.
these rules must be true at both ends of the peptide. When it
num_enzyme_termini=1, then the rules must be true at
at least one of the ends. The
allowed_missed_cleavages parameters control the
maximum number of cleavage sites that may lie within the peptide
Note that if
allowed_missed_cleaveages parameters are not used.
The Crux search tools (
select candidate peptides for each spectrum based on its precursor
singly charged mass (m+h) or the mass-to-charge (m/z) and an assumed
charge state (specified in the input file). If the m+h and charge is
provided in the input file (e.g., from the Z lines of an MS2 file),
then the precursor mass is calculated from the m+h minus the mass of a
proton. Otherwise, the precursor mass is calculated from the precursor
m/z and an assumed charge. A mass window is defined in one of three
ways based on the
precursor-window-typeis set to
mz, then the window is calculated as spectrum precursor m/z ±
precursor-windowand the resulting range is converted to mass using the charge state with the formula:
Mass=m/z * charge - MASS_PROTON * charge, where MASS_PROTON=1.00727646677
precursor-window-typeis set to
mass, then the window is defined as the precursor mass ±
precursor-window-typeis set to
ppm(parts per million), then the window spans from the precursor mass / (1.0 + window * 1e-6) to precursor mass / (1.0 - window * 1e-6).
Candidate peptides are those whose mass falls within the defined window. The peptide mass is computed as the sum of the average amino acid masses plus 18 Da for the terminal OH group. Candidate peptides can also be constrained by minimum and maximum allowed length.
Although the syntax for specifying amino acid modifications differs between Comet and Tide, the logic of how those modifications are applied is similar.
First, it's important to understand that the search tools support several different types of modifications. A static modification is one that applies to every occurrence of an amino acid, whereas a variable modification means that the amino acid can occur in either form (modified or unmodified). Furthermore, unlike standard modifications that occur on amino acid side-chains, a terminal modification is applied to the molecular group at the N- or C-terminus of a peptide or protein. Typically, protein terminal modifications correspond to biological events, whereas peptide terminal modifications (which by definition must be applied after enzymatic digestion) are produced by the experimental workflow.
Conceptually, all static modifications (standard ones and terminal ones) are applied first. Thereafter, variable modifications are applied on top of static modifications. Tide applies at most one static modification and one variable modification at any location, terminal or non-terminal. Comet is similar, except that modifications can also separately be applied to terminal locations. For example, the peptide DEAGFK can have a variable modification on the side chain of the lysine residue as well as a variable modification on the peptide C-terminus.
In practice, it is an error to (try to) specify two different static modifications for the same type of residue, but perfectly legal to specify multiple variable modifications for the same residue.
In the search engine output, you will see the unmodified peptide (e.g., K[-16.9981]VLGHIR in Comet's "plain_peptide" column) and the peptide with variable modifications indicated in brackets (e.g., K[-16.9981]VLGHIR in Comet's "peptide" column and Tide's "sequence" column). The full set of modifications (including static modifications) is specified as a comma-separated list in the "modifications" column, like this: 1_S_45.0294,1_S_45.0239_n,1_V_-16.9981. This specification indicates that the first amino acid has a standard static modification of 45.0294 as well as a static n-terminal modification of 45.0239. In addition, the same residue has a variable modification of -16.9981.
Note that Comet allows slightly more flexibility than Tide in specifying different types of modifications. Terminal modifications in Comet can be specified to occur within a range of amino acids near the terminus. Also, via the "binary set" flag, a modification can be required to appear on all or none of the residues in a given peptide.
Peptide and protein terminal modifications are similar in their logic.
(c/n)term-peptide-mods-spec option applies modifications to the terminals
of all peptides, whereas
(c/n)term-protein-mods-spec applies modifications only
to peptide terminals that are also protein terminals. If both peptide and protein
mod specs are provided, then both will be applied. It is possible to simultaneously
specify peptide and protein terminal mods for the same type of residue. If both
are static then protein terminals will have both of them (sum of the mod masses).
If both are variable then they will be applied one at a time.
Note that the Tide index does not store information about the origin of variable mods. As a result, the Tide search output, unlike Comet, does not discriminate between terminal and non-terminal mods. Thus, Tides modified sequence column always has format K[-16.9981]VLGHIR, whereas in Comet the same peptide would be specified as n[-16.9981]VLGHIR for a terminal mod and K[-16.9981]VLGHIR for a non-terminal mod. Similarly, the modifications column in Tide search results always has the format [position]_V_[mass]. Static modifications do not have this issue: they will have a _c or _n suffix for peptide terminal mods and _C or _N for protein terminal mods.
Reducing the precursor mass tolerance has two main benefits. One is
reduced search time, and the other is improved statistical power to
detect matches. With a smaller window, fewer candidates are tested
against a spectrum. As a result, the statistical confidence measure
calculated after multiple testing correction will be more significant.
Of course, the flipside is that if you make the precursor window too
small, then you may end up throwing out correct identifications.
Control over the size of the window is provided by the using
Isotopically labeled data is accommodated by specifying various static or
variable modifications. The specific types of modifications differ according
to the labeling scheme, and the syntax for specifying the modifications
differs between Tide (which uses options like
Comet (which uses options like
add_Nterm_peptide). In each of
the following entries, we provide first the Tide options, followed by the
--mods-spec K+144.10253 --nterm-peptide-mods-spec X+144.10253
add_Nterm_peptide = 144.10253
add_K_lysine = 144.10253
clear_mz_range = 113.5 117.5
The 4-plex reagent has different monoisotopic mass values for 114 (144.105918), 115 (144.09599), and 116/117 (144.102063). The mass value used above is derived from averaging the three monoisotopic masses.
--mods-spec K+304.2022 --nterm-peptide-mods-spec X+304.2022
add_Nterm_peptide = 304.2022
add_K_lysine = 304.2022
clear_mz_range = 112.5 121.5
The mass modification above is the average of the two different set of masses for 115/118/119/121 (304.199040) and 113/114/116/117 (304.205360).
--mods-spec K+225.155833 --nterm-peptide-mods-spec X+225.155833
add_Nterm_peptide = 225.155833
add_K_lysine = 225.155833
clear_mz_range = 125.5 127.5
--mods-spec K+229.162932 --nterm-peptide-mods-spec X+229.162932
add_Nterm_peptide = 229.162932
add_K_lysine = 229.162932
clear_mz_range = 127.5 131.5
There are a number of different SILAC reagents with a ~4 Da modification (based on combinations of C13 and N15), each with different sites of specificity. The example below is for the 15N(4) reagent applied to R residues. You should adjust the modification mass and residue(s) applied to as necessary.
To perform a mixed light/heavy search using a variable modification search:
variable_mod01 = 3.988140 R 1 3 -1 0 0
Note that, in the example above, Comet operates in “binary” mode, which means that it only considers peptides in which all or none of the lysine residues are light/heavy. Tide does not support binary mode searching, so it will consider all peptides with up to 3 (in the above specification) heavy lysines.
To search just the heavy labeled sample, you can apply a static modification:
add_K_lysine = 3.988140
This example uses the 13C(6) SILAC mass, assuming it is applied to both K and R. You should adjust as necessary. There is at least one more SILAC reagent with ~6 Da modification mass and different residue specificity: 13C(5) 15N(1)
To perform a mixed light/heavy search using a variable modification:
variable_mod01 = 6.020129 KR 1 3 -1 0 0
To search just the heavy labeled sample, you can apply a static modification:
add_K_lysine = 6.020129
add_R_arginine = 6.020129
The example below is for 13C(6) 15N(2) on K residues. Variable modification search:
variable_mod01 = 8.014199 K 1 3 -1 0 0
Static modification for just the heavy labeled search:
add_K_lysine = 8.014199
100% 15N label:
The N15 labeling adds 0.997035 mass difference for every nitrogen in each amino acid. So for a 100% N15 analysis, specify a static modification (between 1 and 4) to every amino acid residue, like this:
add_G_glycine = 0.997035 # added to G - avg. 57.0513, mono. 57.02146 add_A_alanine = 0.997035 # added to A - avg. 71.0779, mono. 71.03711 add_S_serine = 0.997035 # added to S - avg. 87.0773, mono. 87.03203 add_P_proline = 0.997035 # added to P - avg. 97.1152, mono. 97.05276 add_V_valine = 0.997035 # added to V - avg. 99.1311, mono. 99.06841 add_T_threonine = 0.997035 # added to T - avg. 101.1038, mono. 101.04768 add_C_cysteine = 0.997035 # added to C - avg. 103.1429, mono. 103.00918 add_L_leucine = 0.997035 # added to L - avg. 113.1576, mono. 113.08406 add_I_isoleucine = 0.997035 # added to I - avg. 113.1576, mono. 113.08406 add_N_asparagine = 1.99407 # added to N - avg. 114.1026, mono. 114.04293 add_D_aspartic_acid = 0.997035 # added to D - avg. 115.0874, mono. 115.02694 add_Q_glutamine = 1.99407 # added to Q - avg. 128.1292, mono. 128.05858 add_K_lysine = 1.99407 # added to K - avg. 128.1723, mono. 128.09496 add_E_glutamic_acid = 0.997035 # added to E - avg. 129.1140, mono. 129.04259 add_M_methionine = 0.997035 # added to M - avg. 131.1961, mono. 131.04048 add_O_ornithine = 1.99407 # added to O - avg. 132.1610, mono 132.08988 add_H_histidine = 2.99111 # added to H - avg. 137.1393, mono. 137.05891 add_F_phenylalanine = 0.997035 # added to F - avg. 147.1739, mono. 147.06841 add_U_selenocysteine = 0.997035 # added to U - avg. 150.3079, mono. 150.95363 add_R_arginine = 3.98814 # added to R - avg. 156.1857, mono. 156.10111 add_Y_tyrosine = 0.997035 # added to Y - avg. 163.0633, mono. 163.06333 add_W_tryptophan = 1.99407 # added to W - avg. 186.0793, mono. 186.07931 add_B_user_amino_acid = 0.0000 # added to B - avg. 0.0000, mono. 0.00000 add_J_user_amino_acid = 0.0000 # added to J - avg. 0.0000, mono. 0.00000 add_X_user_amino_acid = 0.0000 # added to X - avg. 0.0000, mono. 0.00000 add_Z_user_amino_acid = 0.0000 # added to Z - avg. 0.0000, mono. 0.00000
If you need to adjust the cysteine mass to account for carboxyamidomethylation, then use 58.018499 instead of 0.997035 for C.
add_C_cysteine = 0.997035
Patches implementing new features can submitted directly as pull requests on github, or can be be emailed to the development team at firstname.lastname@example.org for review and inclusion in subsequent releases of Crux.
Thin air. The name is not an acronym or a reference to anything in particular.