Automated resonance assignment with FLYA (EMBO 2013)

From CYANA Wiki
Jump to navigation Jump to search

In this tutorial we want to determine the resonance assignments and the structure of a protein using the program CYANA.

CYANA setup for the EMBO Practical Course NMR in Basel (20–27 July 2013)

Please follow the following steps carefully (exact Linux commands are given below; you may copy them to a terminal):

  1. From your home directory, change to the directory 'guentert'.
  2. Execute the setup script.
  3. Download the input data for the practical.
  4. Unpack the input data for the practical.
  5. Move it to different directory, 'flyaembo1' (such that you later unpack the original data again to run another calculation)
  6. Change into the subdirectory 'flyaembo1' where the input data for the practical are.
  7. Test whether CYANA can be started by typing its name, 'cyana'.
  8. Exit from CYANA by typing 'q' or 'quit'.
cd ~/guentert
source setup-guentert
wget 'http://www.cyana.org/wiki/images/3/31/Flyaembo.tgz'
tar zxf Flyaembo.tgz
mv flyaembo flyaembo1
cd flyaembo1
cyana
___________________________________________________________________

CYANA 3.96 (linux-intel)
 
Copyright (c) 2002-12 Peter Guentert. All rights reserved.
___________________________________________________________________

    Time-limited license valid until 2013-12-31.
    Library file "/home/guentert/src/cyana-3.96/lib/cyana.lib" read, 38 residue types.
    Sequence file "demo.seq" read, 114 residues.
cyana> q

If all worked, you are ready to go!

If you want to return to your practical later, using your own Linux or Mac OS X computer, you can download the demo version of CYANA from here. This is not necessary on the workshop computers where the software is already installed.

Experimental input data

The protein sequence is stored in three-letter code in the file 'demo.seq'.

The following spectra have been measured:

  • [1H,13C]-HSQC (called 'C13H1' in FLYA)
  • [1H,15N]-HSQC (called 'N15H1' in FLYA)
  • HNCA
  • HN(CO)CA (called 'HN_CO_CA' in FLYA)
  • HNCO
  • HN(CA)CO (called 'HN_CA_CO' in FLYA)
  • CBCANH
  • CBCACONH
  • HBHACONH
  • HCCH-TOCSY (called 'HCCH24' in FLYA)
  • HCCH-COSY (called 'HCCH7' in FLYA)
  • C(CO)NH (called 'C_CO_NH' in FLYA)
  • HC(CO)NH (called 'HC_CO_NH' in FLYA)

Peak lists in XEASY format that have been prepared by automatic peak picking with the program NMRView are stored in files XXX.peaks, where XXX denotes the FLYA spectrum type.

Each peak list starts with a header that defines the experiment type and the order of dimensions. For instance, for HNCA.peaks:

# Number of dimensions 3
#FORMAT xeasy3D
#INAME 1 HN
#INAME 2 C
#INAME 3 N
#SPECTRUM HNCA  HN C N
      5   6.475  58.033  98.548 1 U   2.769E+02  0.000E+00 e 0     0     0     0
      6   6.476  62.123  98.126 1 U   2.571E+01  0.000E+00 e 0     0     0     0
      7   6.475  54.017  98.159 1 U   2.547E+01  0.000E+00 e 0     0     0     0

The first line specifies the number of dimensions (3 in this case). The '#SPECTRUM' lines gives the experiment type (HNCA, which refers to the corresponding experiment definition in the CYANA library), followed by an identifier for each dimension of the peak list (HN C N) that specifies which chemical shift is stored in the corresponding dimension of the peak list. These labels must match those in the corresponding experiment definition in the general CYANA library (see below). After the '#SPECTRUM' line follows one line for every peak. For example, the first peak in the 'HNCA.peaks' list has

  • Peak number 5
  • HN chemical shift 6.475 ppm
  • C (CA) chemical shift 58.033 ppm
  • N chemical shift 98.548 ppm

The other data are irrelevant for automated chemical shift assignment with FLYA. In particular, the peak volume or intensity (2.769E+02) is not used by the algorithm.

Hint: The formats of other CYANA files are described in the CYANA 3.0 Reference Manual.

FLYA initialization script

The CYANA commands to run the automated assignment calculation are stored in two CYANA scripts or "macros".

One has the fixed name 'init.cya' and is executed automatically each time CYANA is started. It can also be called any one wants to reinitialize the program. It contains normally at least two commands that read the CYANA library and the protein sequence:

cyanalib
read demo.seq

The command 'cyanalib' reads the standard CYANA library. The second command reads the protein sequence.

Experiment definitions in the CYANA library

When you start CYANA, the program reads the library and displays the full path name of the library file. You can open the standard library file to inspect, for example, the NMR experiment definitions that define which expected peaks are generated by FLYA. For instance, the definition for the HNCA spectrum (search for 'HNCA' is

SPECTRUM HNCA  HN N C
 0.980  HN:H_AMI  N:N_AM*  C_A*  C_BYL  C:C_ALI
 0.800  HN:H_AMI  N:N_AMI  C_BYL  C_ALI  N_AMI  C:C_ALI

The first line corresponds to the '#SPECTRUM' line in the peak list. It specifies the experiment name and a label for the atoms that are detected in each dimension of the spectrum. The number of labels defines the dimensionality of the experiment (3 in case of HNCA).

Each line below defines a (formal) magnetization transfer pathway that gives rise to an expected peak. in the case of HNCA there are two lines, corresponding to the intraresidual and sequential peak. For instance, the definition for the intraresidual peak starts with the probability to observe the peak (0.980), followed by a series of atom types, e.g. H_AMI for amide proton etc. An expected peak is generated for each molecular fragment in which these atom types occur connected by single covalent bonds. The atoms whose chemical shifts appear in the spectrum are identified by their labels followed by ':', e.g. for HNCA 'HN:', 'N:', and 'C:'.

FLYA execution script

The CYANA script 'RUNFLYA.cya' contains the commands to perform the automated resonance assignment. It starts with the specification of the names of the input peak lists:

noesy:=N15NOESY,C13NOESY
scalar:=C13H1,N15H1,HCCH24,HCCH7,HNCA,HN_CA_CO,HNCO,HN_CO_CA,CBCANH,CBCACONH,HBHACONH,C_CO_NH,HC_CO_NH
#scalar:=C13H1,N15H1,CBCACONH,C_CO_NH,HC_CO_NH

"Through-space" experiments are given in the 'noesy:=...' statement; "through-bond" ones in the 'scalar:=...' statement. The peak list names are separated by commas (without blanks!).

The commands above will use all available peak lists. You can choose any subset of them by modifying the 'noesy:=...' or 'scalar:=...' statements.

Try it! For instance, to make only backbone assignment using the triple resonance backbone assignment spectra, you can set

noesy:=
scalar:=HNCA,HN_CA_CO,HNCO,HN_CO_CA,CBCANH,CBCACONH,HBHACONH

These are followed by tolerances for chemical shift matching:

tolerance:=0.03,0.03,0.4
assigncs_accH:=tolerance(1)
assigncs_accC:=tolerance(3)
assigncs_accN:=tolerance(3)

In this case a tolerance of 0.03 ppm will be used for protons, and 0.4 ppm will be used for carbon and nitrogen.

Parameters for the FLYA algorithm come next:

run_assign_population:=20
run_assign_iterations:=15000
run_assign_reference:=ref.prot
analyze_assign_group:=CONBB: N H CA C CB / CONSOLIDATED, CONALL: CONSOLIDATED, BB: N H CA C CB, ALL:*
randomseed  := 3771

These define

  • The population size for the genetic algorithm, i.e. how many assignments form one generation (20; chosen smaller than in normal production runs in order to speed up the calculation)
  • The maximal number of iterations during local optimization (15000).
  • A file with reference chemical shifts (ref.prot). These will not be used by the algorithm but only for a subsequent analysis of its results.
  • Groups of atoms for which assignment statistics will be calculated and reported in the 'flya.txt' output file.
  • The seed value for the random number generator. Any positive integer number can be used.

There might be a command to restrict the generation expected peaks to a subset of atoms:

command select_atoms
  atoms select "* - CZ ?H* @ARG - ?Z @LYS"
end

Here, the zeta atoms of Arg and Lys are excluded, i.e. no expected peaks will be generated for these atoms (because they are only rarely observed in the spectra).

Specific labeling can be handled in the same way, and peak list-specific atom selections can be defined (not in this practical):

command XXX_select
  atoms select "..."
end

Finally, there is the command to start the FLYA algorithm:

flya refprot=ref.prot noesy=$noesy scalar=$scalar runs=5 #stage=1

Here, the given parameters of the 'flya' command specify that

  • The results will be compared with the reference chemical shifts in the file 'ref.prot' (which have been determined independently by conventional methods).
  • The input peak lists that will be used (as defined above).
  • The number of independent runs of the algorithm, from which the consolidated shift will be calculated (chosen smaller than in normal production runs in order to speed up the calculation).

Run the FLYA calculation

To run the FLYA calculation, you start CYANA and execute the 'RUNFLYA.cya' script:

cyana "nproc=5; RUNFLYA"

By specifying the 'nproc=5' command, the 5 independent runs of the algorithm will be performed in parallel. On a computer with multiple processors this will speed up the calculation, which is expected to take a few minutes.

FLYA output files

The FLYA algorithm will produce the following output files:

  • flya.prot: Consensus assigned chemical shifts. This file contains a chemical shift for every atom that has been assigned to least one peak.
  • flya.tab: Table with details about the chemical shift assignment of each atom (comparison with reference shifts). In this file you can see for each atom whether the assignment is "strong" (self-consistent) or "weak" (only tentative).
  • flya.txt: Assignment statistics
  • flya.pdf: Graphical representation of the assignment results
  • XXX_exp.peaks: List of expected peaks, corresponding to input peak list XXX.peaks
  • XXX_asn.peaks: Assigned peak list, corresponding to input peak list XXX.peaks

The flya.txt file

This output file starts with overall assignment statistics for each group of atoms as defined by 'analyze_assign_group:=...' in RUNFLYA.cya':

   ____________________________________________________________

   CHEMICAL SHIFT ASSIGNMENT
   ____________________________________________________________

   SEED: 1
   assigned from frequencies

   CONBB: REF 493 [FREQ 493 [ASN 493 T 489(99.2%) F 4 [INNERRES 0]]] ADDFREQ 11 [ASN 11]

   CONALL: REF 1141 [FREQ 1141 [ASN 1139 T 1073(94.0%) F 68 [INNERRES 45]]] ADDFREQ 30 [ASN 29]

   BB: REF 512 [FREQ 512 [ASN 512 T 502(98.0%) F 10 [INNERRES 4]]] ADDFREQ 30 [ASN 28]

   ALL: REF 1264 [FREQ 1261 [ASN 1259 T 1151(91.1%) F 110 [INNERRES 63]]] ADDFREQ 89 [ASN 83]
  • REF is the number of reference assignments (in the selected group)
  • FREQ is the number of atoms (with reference assignments) that could in principle be assigned by FLYA, i.e. for which at least one expected peak was generated.
  • ASN is the number of atoms assigned by FLYA
  • T is the number of correct FLYA assignments that agree (within the tolerance) with the reference. The percentage is relative to ASN.
  • F is the number of erroneous FLYA assignments that do not agree (within the tolerance) with the reference.
  • INNERRES are erroneous assignment for which the correct assignment is in the same residue.
  • ADDFREQ are additional atoms (without reference assignments) that could in principle be assigned by FLYA.
  • ASN are additional atoms (without reference assignments) that were assigned by FLYA.

Further below comes a table with information about each peak list:

   Lists      #Expected  noRef   noPeak   Assigned        Match    #Measured Assigned  exp/meas Assigned
   N15NOESY      1497     174     502   932( 62.26%)   789( 52.71%)   3008    768( 25.53%)   1.2
   C13NOESY      4864     299    2091  2811( 57.79%)  2220( 45.64%)  10886   2258( 20.74%)   1.2
   C13H1          556      27      85   464( 83.45%)   406( 73.02%)    407    350( 86.00%)   1.3
   N15H1          135      16       4   127( 94.07%)   113( 83.70%)    131    112( 85.50%)   1.1
   HCCH24        2734      79    1403  1295( 47.37%)  1039( 38.00%)   2363   1015( 42.95%)   1.3
   HCCH7         1926      69    1102   852( 44.24%)   686( 35.62%)   2005    675( 33.67%)   1.3
   HNCA           219      15      19   195( 89.04%)   184( 84.02%)    329    177( 53.80%)   1.1
   HN_CA_CO       211      15      11   194( 91.94%)   180( 85.31%)    246    175( 71.14%)   1.1
   HNCO           105       7       1   100( 95.24%)    96( 91.43%)    158     97( 61.39%)   1.0
   HN_CO_CA       105       7       0   103( 98.10%)    96( 91.43%)    158     98( 62.03%)   1.1
   CBCANH         415      26      41   363( 87.47%)   344( 82.89%)    623    335( 53.77%)   1.1
   CBCACONH       216      13      18   195( 90.28%)   182( 84.26%)    324    185( 57.10%)   1.1
   HBHACONH       288      20      82   205( 71.18%)   188( 65.28%)    364    183( 50.27%)   1.1
   C_CO_NH        370      16      53   309( 83.51%)   277( 74.86%)    365    284( 77.81%)   1.1
   HC_CO_NH       540      22     225   319( 59.07%)   282( 52.22%)    442    259( 58.60%)   1.2
   ALL          14181     805    5637  8464( 59.69%)  7082( 49.94%)  21809   6971( 31.96%)   1.2

It contains the following data:

  • #Expected: Total number of expected peaks
  • noRef: Number of expected peaks with missing reference shifts
  • noPeak: Number of expected peaks for wich no peak can be measured
  • Assigned: Number of expected peaks that could be assigned based on the reference chemical shift assignments. The theoretical maximum of 100% corresponds to the situation that the spectra “explain” all expected peaks. Each expected peak can be mapped to at most one measured peak. Remaining expected peaks correspond to missing peaks in the measured peak list.
  • Match: Number of assigned peaks that fit (within tolerance) reference shifts. The theoretical maximum of 100% corresponds to having all measured peaks assigned. Note that several expected peaks can be mapped to the same measured peak, i.e. the assignments of measured peaks can be unambiguous or ambiguous. Remaining unassigned measured peaks are likely to be artifacts.
  • #Measured: Total number of peaks in peak list
  • Assigned: Number of measured peaks that could be assigned to expected peaks
  • exp/meas: Ratio of assigned expected and measured peaks

There is more information on the results of the assignment calculation in the 'flya.txt' file (not described here).

The flya.tab file

This file provides information about the chemical shift assignment of each individual atom:

   Atom  Residue      Ref   Shift     Dev  Extent  inside   inref
   N     LYS   11 120.779 120.805  -0.026     5.0   100.0   100.0  consolidated=
   H     LYS   11   7.516   7.514   0.002     5.0   100.0   100.0  consolidated=
   CA    LYS   11  55.129  54.976   0.153     5.0   100.0   100.0  consolidated=
   HA    LYS   11   4.244   4.241   0.003     5.0    99.9   100.0  consolidated=
   CB    LYS   11  31.380  31.357   0.023     5.0   100.0   100.0  consolidated=
   HB2   LYS   11   1.490   1.486   0.004     5.0   100.0   100.0  consolidated=
   HB3   LYS   11   1.490   1.487   0.003     5.0   100.0   100.0  consolidated=
   CG    LYS   11  25.362  25.427  -0.065     5.0   100.0   100.0  consolidated=
   HG2   LYS   11   1.270   1.263   0.007     5.0   100.0   100.0  consolidated=
   HG3   LYS   11   1.446   1.456  -0.010     5.0    99.8   100.0  consolidated=
   CD    LYS   11  29.374  29.445  -0.071     5.0    65.6    60.0  =
   HD2   LYS   11   1.495   1.487   0.008     5.0    59.9    60.0  =
   HD3   LYS   11   1.495   2.045  -0.550     5.0    91.4     0.0  consolidated! (HB3 10)
   CE    LYS   11  42.023  42.013   0.010     5.0   100.0   100.0  consolidated=
   HE2   LYS   11   2.744   2.749  -0.005     5.0    99.9   100.0  consolidated=
   HE3   LYS   11   2.824   2.758   0.066     5.0    74.2     0.0  ! (HE2)
   C     LYS   11 174.616 176.282  -1.666     5.0    80.0    20.0  consolidated! (C 10)
  • Ref: Chemical shift value in the reference chemical shift list (ref.prot). It was not used in the calculation.
  • Shift: Consensus chemical shift value from FLYA
  • Dev = Ref - Shift
  • Extent: Number of runs in which the atom was assigned by FLYA.
  • Inside: Percentage of chemical shift values from the (5) independent runs of FLYA that agree (within the tolerance) with the consensus value.
  • inref: Percentage of chemical shift values from the (5) independent runs of FLYA that agree (within the tolerance) with the reference value.
  • Outcome of the assignment:
    • consolidated: "strong" assignment, i.e. Inside > 80%.
    • =: Assignment that agrees with reference, i.e. Dev < tolerance.
    • !: Assignment that does not agree with the reference, i.e. Dev > tolerance.
    • (atom name): Correct assignment, if within the same residue (no residue number given), or the neighboring residues.

The flya.pdf file

This PDF file provides a graphical representation of the 'flya.tab' file. Each assignment for an atom is represented by a colored rectangle.

flya.pdf
  • Green: Assignment by FLYA agrees with the manually determined reference assignment (within tolerance)
  • Red: Assignment by FLYA does not agree with the manually determined reference assignment
  • Blue: Assigned by FLYA but no reference available
  • Black: With reference assignment but not assigned by FLYA.

Respective light colors indicate assignments not classified as strong by the chemical shift consolidation. The row labeled HN/Hα shows for each residue HN on the left and Hα in the center. The N/Cα/C’ row shows for each residue the N, Cα, and C’ assignments from left to right. The rows β-η show the side-chain assignments for the heavy atoms in the center and hydrogen atoms to the left and right. In the case of branched side-chains, the corresponding row is split into an upper part for one branch and a lower part for the other branch.

Using input chemical shifts: shift predictions or partial assignments (optional)

Input chemical shift can be used in three ways.

These shifts will only be used for comparison (e.g. in flya.tab, flya.txt, flya.pdf):

run_assign_reference:=ref.prot

Shifts and standard deviations in the file 'predicted.prot' (not provided in this practical) will replace the general statistics from cyana.lib (CSTABLE):

run_assign_statistics:=predicted.prot

Shifts in the file 'fix.prot' will be fixed to the input values

run_assign_fix:=fix.prot

The latter approach can for instance be used to perform sidechain assignment when the backbone assignment is already known.

If you want to do this, you can first make a list of only the reference backbone chemical shifts. Start CYANA. In CYANA, enter the commands

read ref.prot
atom set "* - H N CA CB C" shift=none
write fix.prot
q

The file 'fix.prot' will contain the reference chemical shifts only for the backbone (and CB) atoms H, N, CA, CB, C'. Now you can repeat the assignment calculation by inserting the 'run_assign_fix:=fix.prot' statement and choosing only the input peak lists that are relevant for sidechain assignment:

run_assign_fix:=fix.prot
noesy:=N15NOESY,C13NOESY
scalar:=C13H1,N15H1,HCCH24,HCCH7,HBHACONH,C_CO_NH,HC_CO_NH

Fully automated structure calculation

Automated resonance assignment, automated NOE restraint assignment, and the structure calculation by torsion angle dynamics can be combined by running the 'flya' command in 'RUNFLYA.cya' with the additional parameter 'stage=1' (which was commented out so far):

flya refprot=ref.prot noesy=$noesy scalar=$scalar runs=5 stage=1

To speed up the calculation, you can set in 'RUNFLYA.cya' (above the 'flya' command):

structures:=25,5
steps=4000

These tell the program to calculate, in each cycle, 25 conformers, and to analyze the best 5 of them. 4000 torsion angle dynamics steps will be applied per conformer.

7 cycle of automated NOE assignment and structure calculation will be performed. The final structure will be 'final.pdb'. To get an overview of the calculation, type (at the Linux prompt) the command

cyanatable -lp