eMF stands for easy Model Free. Why did I have to write another model free program? It was originally written in order to implement distributed correlation time approaches such as Cole-Cole and Lorentzian spectral density functions for the model free analysis of unfolded or disordered proteins.
During my time with Drs. Peter E. Wright and Jane H. Dyson at The Scripps Research Institute in La Jolla, CA, I studied the ps-ns dynamics of wild type and mutant prion proteins. Prion protein has a very long stretch of disordered amino acid sequence at the N-terminus which significantly alters observed NMR relaxation rates of the whole protein including the C-terminal structured domain. Apparently, it should be described by an ensemble of vastly different overall structures due to the disordered N-terminal domain.
eMF was born out of an attempt to resolve this problem by applying the 'distributed' correlation time. It was written to be easy and straightforward to use and to put together recent model free protocols.
- Compatible with GSL 2.x and GSL 1.x
- Using CMake default FindGSL (requires cmake >= 3.2)
- separate config and data files
- reduced spectral mapping
- estimation and optimization of diffusion tensor (isotropic, axially symmetric, asymmetric, local, distributed)
- fit extended Lipari-Szabo models with selection options of BIC and AIC
- xmgrace plot of select model spectral density function
Please cite the following paper if you use eMF.
- Prion Proteins with Pathogenic and Protective Mutations Show Similar Structure and Dynamics. Sung-Hun Bae, Giuseppe Legname, Ana Serban, Stanley B. Prusiner, Peter E. Wright, and H. Jane Dyson Biochemistry 2009 48 (34), 8120-8128 http://pubs.acs.org/doi/abs/10.1021/bi900923b
You need the following two packages for proper installation:
- CMake
- GNU Scientific Library (GSL)
$ mkdir build
$ cd build
$ cmake ..
$ make
$ make install
By default, an executable binary file(eMF) and
a perl script(eMF-digest) will be installed in ~/bin directory.
You may change the destination directory defined the CMakeLists.txt file :
SET(CMAKE_INSTALL_PREFIX ~/bin )
Please make sure that eMF and eMF-digest
is accessible from your working directory
by adding ~/bin or your destination directory in the 'path'.
You need to prepare config and data files. These files are plain/text file and please use the following examples files as templates.
Default config filename is emf.conf but you may use
different config filename by -c [filename].
#
# eMF configuration
#
# CONSTANTS
# const [gamma_h ###] [gamma_x ###] [csa_x ###] [r_xh ###]
const gamma_h 26.7519 # gamma for H
const gamma_x -2.71 # gamma for X
const r_xh 1.02 # bond length (A)
const csa_x -172.0 # chemical shift anisotropy (ppm)
# PARAMETERS
# par [name][-lb #][-ub #][-gr #][-cv #][-lk #][-uk #][-fx #][-in #]
# name= S2f S2s te Rex tc Dr a scf Dxx Dyy Dzz phi theta psi
# -lb : lower bound
# -ub : upper bound
# -gr : grid points
# -cv : convergence limit
# -lk : lower bound harmonic chi-square penalty
# chisq += lk * (v - lb)^2
# -uk : upper bound harmonic chi-square penalty
# chisq += uk * (v - ub)^2
# -fx : fix value
# -in : initial value
# internal dynamics
par S2f -lb 0.0 -ub 1.0 -gr 20
par S2s -lb 0.0 -ub 1.0 -gr 20
par te -lb 0.0 -ub 3.0 -gr 30
par Rex -lb 0.0 -ub 10.0 -gr 10
# diffusion tensor: isotropic (-I)
# par tc -lb 7.5 -ub 9.5 -gr 20 -cv 1e-3
# diffusion tensor: axially symmetric (-A)
# There are two possible solutions: prolate and oblate.
# prolate (z-axis is the largest, Dr >1)
# oblate (z-axis is the smallest, Dr <1)
#
# Dpar = Dzz (parallel)
# Dper = Dxx = Dyy (perpendicular)
# Dr = Dpar/Dper = Dzz/Dxx or Dzz/Dyy
#
# Dper = 1.0 / (2 * tc * (2.0 + Dr ))
# At the isotropic diffusion limit (Dr=1), Dper = Dpar = 1/(6*(tc))
#
# valid range of diffusion tensor angles
# phi : 0-360 degree
# theta : 0-180 degree
par tc -lb 7.5 -ub 9.5 -gr 20 -cv 1e-3
par Dr -lb 1.0 -ub 1.3 -gr 10 -cv 0.01
par phi -lb 0 -ub 360 -gr 24 -cv 0.1 -fx -2.385
par theta -lb 0 -ub 180 -gr 12 -cv 0.1 -fx 0.000
# diffusion tensor: anisotropic (-N)
# valid range of diffusion tensor angles
# phi : 0-360 degree
# theta : 0-180 degree
# psi : 0-360 degree
#
# par Dxx -lb 0.008 -ub 0.03 -gr 20 -cv 1e-3
# par Dyy -lb 0.008 -ub 0.03 -gr 20 -cv 1e-3
# par Dzz -lb 0.008 -ub 0.03 -gr 20 -cv 1e-3
# par phi -lb 0.0 -ub 360 -gr 24 -cv 1e-3 -in 173.128
# par theta -lb 0.0 -ub 180 -gr 12 -cv 1e-3 -in 75.123
# par psi -lb 0.0 -ub 360 -gr 24 -cv 1e-3 -in 82.351
# PDB COORDINATES (NECESSARY FOR NON-ISOTROPIC MODELS)
# coor [file name] [-h atom name] [-x atom name]
coor 1rx7.rot.pdb -h H -x N
# ADJUST EXPERIMENTAL ERRORS
# error [-field #] [-min # # #] [-scale # # #]
# -field # : specifiy field or * for all data
# -min # # # : set minimum proportioanl errors for R1,R2,NOE
# -scale # # # : scale errors for R1,R2,NOE
# error -field * -min 0.05 0.05 0.05
# i.e if given errors are less than 5% of measured values,
# set errors as 5% of measured values for all fields
# ESTIMATION OF DIFFUSION PARAMETER(S) by fitting R2/R1
# estimate [-cluster # ]
estimate -cluster 1
# OPTIMIZATION OF DIFFUSION PARAMETER(S) by fitting R1,R2,NOE
# optimize [-cluster #] [-s2 #] [-maxiter #]
#
# -cluster # : use cluster #
# -s2 # : use residues with order parameter larger than s2
# -maxiter # : maximum iteration
optimize -cluster 1 -s2 0.7
# i.e. use cluster 1 AND S2 > 0.7 for optimization of diffusion tensor
# FITTING AND SELECTING MODELS
#
# fitmodel [-select AIC|BIC] [-mc #] [-mctrim #]
#
# -select AIC : use Akke information criterion
# -select BIC : use Bayesian information criterion
fitmodel -select BIC
# column 1 : residue number
# column 2 : cluster number, used for inclusion or exclusion of data
# column 3 : B0, field strength in MHz
# column 4-5 : R1 relaxation rate and error (1/s)
# column 6-7 : R2 relaxation rate and error (1/s)
# column 8-9 : {1H}-X NOE and error
# column 10 : (only for distributed) tc distribution file name
7 1 500.38 1.6265 0.0691 10.7276 0.4488 0.7760 0.0400
7 1 600.13 1.2629 0.0436 11.1164 0.7537 0.8070 0.0880
8 1 500.38 1.6935 0.0442 10.8300 0.3911 0.7730 0.0240
8 1 600.13 1.3243 0.0559 12.1164 0.3416 0.7990 0.0220
9 0 500.38 1.6324 0.0658 15.3615 0.6703 0.7930 0.0260
9 0 600.13 1.2494 0.0476 16.6599 0.5612 0.7710 0.0660
10 1 500.38 1.7236 0.0398 10.5304 0.5007 0.7900 0.0400
10 1 600.13 1.3388 0.0534 10.7676 0.4242 0.7830 0.0420
11 1 500.38 1.5871 0.0856 10.7627 0.4155 0.8040 0.0700
11 1 600.13 1.2146 0.0245 10.0442 0.2884 0.7720 0.1080
$ eMF -h
Reduced spectral mapping only needs const definitions in the config file.
$ eMF -s dhfr.dat
Output will be like:
# REDUCED SPECTRAL DENSITY
#
# B0= 500.38 MHz, 0.87H freq.= 435.33 MHz or 2.735e+09 rad/s
# B0= 500.38 MHz, N freq.= -50.69 MHz or -3.185e+08 rad/s
# B0= 600.13 MHz, 0.87H freq.= 522.11 MHz or 3.281e+09 rad/s
# B0= 600.13 MHz, N freq.= -60.79 MHz or -3.820e+08 rad/s
# J : (ns/rad)
#
# RESID J(0) dJ(0) J(N) dJ(N) J(0.87H) dJ(0.87H) B0(MHz)
# ----- -------- -------- -------- -------- -------- -------- --------
2 2.68e+00 6.42e-02 3.52e-01 8.02e-03 6.62e-03 9.53e-04 500.38
2 2.73e+00 9.90e-02 2.56e-01 1.07e-02 4.53e-03 6.82e-04 600.13
3 3.15e+00 9.46e-02 3.43e-01 9.31e-03 5.27e-03 7.13e-04 500.38
3 2.98e+00 1.38e-01 2.46e-01 1.03e-02 4.03e-03 9.79e-04 600.13
| diffusion tensor | parameter(s) | eMF command |
|---|---|---|
| isotropic | tc | I |
| axially symmetric | tc, Dr, phi, theta | A |
| anisotropic | Dxx, Dyy, Dzz, phi, theta, psi | N |
| distributed | (datafile) | D |
Diffusion tensor should be determined before fitting the models.
In general, determination of diffusion tensor is an iterative process
and takes several trials. So, you can disable the fitmodel
at the last line in the config file
until you find a suitable diffusion tensor:
#fitmodel -select BIC
Run
$ eMF -A dhfr.dat
Once the diffusion tensor is determined, you can disable estimation and optimization and enable fitmodel:
#estimate -cluster 1
#optimize -cluster 1 -s2 0.7
fitmodel -select BIC
Run
$ eMF -A dhfr.dat
For xmgrace outputs (resid_0007.xmgr, resid_0008.xmgr, ...):
$ eMF -x resid -A dhfr.dat
Output will be like:
eMF 1.0 by Sung-Hun Bae (2008)
# data = dhfr.dat
# number of lines = 10
# number of clusters = 2 (0, 1)
# number of residues = 5
# number of magnetic fields = 2 (500.38, 600.13)
# configuration = emf.conf
# gamma H = 26.7519
# gamma X = -2.71
# CSA X (ppm) = -172
# bond length (A) = 1.02
# estimation of diffusion = No
# optimization of diffusion = No
# model selection criterion = BIC
# Monte Carlo simulation = No
# error adjusted = Yes
# R1 error minimum (%) = 0, 0
# R1 error scale = 1, 1
# R2 error minimum (%) = 0, 0
# R2 error scale = 1, 1
# hnNOE error minimum = 0, 0
# hnNOE error scale = 1, 1
# PDB coordinate file = 1rx7.rot.pdb
# number of vectors = 5
# atom H = H
# atom X = N
# diffusion tensor = global axial
# spectral density function = Lipari-Szabo extended
# Param Value [ Lower ... Upper | GRD, Step ] Convergence
# S2f 0.000 [ 0.000 ... 1.000 | 20, 5.0e-02 ] 1.0e-05
# S2s 0.000 [ 0.000 ... 1.000 | 20, 5.0e-02 ] 1.0e-05
# te 0.000 [ 0.000 ... 3.000 | 30, 1.0e-01 ] 1.0e-05
# Rex 0.000 [ 0.000 ... 10.000 | 10, 1.0e+00 ] 1.0e-05
# GF tc 8.630 [ 7.500 ... 9.500 | 20, 1.0e-01 ] 1.0e-03
# GF Dr 1.195 [ 1.000 ... 1.300 | 10, 3.0e-02 ] 1.0e-02
# GF phi -2.385 [ 0.000 ... 360.000 | 24, 1.5e+01 ] 1.0e-01
# GF theta 0.000 [ 0.000 ... 180.000 | 12, 1.5e+01 ] 1.0e-01
RESIDUE 7 tc 8.630
RESIDUE 7 Dr 1.195
RESIDUE 7 phi -2.385
RESIDUE 7 theta 0.000
RESIDUE 7 alpha 34.971
>+M1 S2 0.828 0.018 S2s 0.828 0.018 S2f 1.000 0.000 te 0.000 0.000 Rex 0.00 0.00 X2 1.099 dof 5 BIC 2.890 AIC 3.099
+M2 S2 0.828 0.018 S2s 0.828 0.018 S2f 1.000 0.000 te -1.095 0.000 Rex 0.00 0.00 X2 1.099 dof 4 BIC 4.682 AIC 5.099
+M3 S2 0.821 0.022 S2s 0.821 0.022 S2f 1.000 0.000 te 0.000 0.000 Rex 0.22 0.42 X2 0.818 dof 4 BIC 4.402 AIC 4.818
+M4 S2 0.815 0.140 S2s 0.815 0.140 S2f 1.000 0.000 te 0.009 0.217 Rex 0.28 1.41 X2 0.277 dof 3 BIC 5.652 AIC 6.277
+M5 S2 0.824 3.155 S2s 0.985 2.659 S2f 0.837 2.270 te 0.131 29.834 Rex 0.00 0.00 X2 0.699 dof 3 BIC 6.074 AIC 6.699
# EXP 500.38 1.6265 0.0691 10.7276 0.4488 0.7760 0.0400
# FIT M1 500.38 1.6351 10.4611 0.8057
# EXP 600.13 1.2629 0.0436 11.1164 0.7537 0.8070 0.0880
# FIT M1 600.13 1.2784 11.1412 0.8271
Selected models are marked by > in the eMF output.
Parameters of these models can be conveniently extracted from
the eMF output file by eMF-digest:
$ eMF-digest dhfr-test.out
Output will be like:
# Resid S2.val S2.err S2s.val S2s.err S2f.val S2f.err te.val te.err Rex.val Rex.err X2 dof BIC AIC Model
7 0.828 0.018 0.828 0.018 1.000 0.000 0.000 0.000 0.000 0.000 1.099 5.000 2.890 3.099 1
8 0.836 0.110 0.836 0.110 1.000 0.000 0.011 0.216 0.660 0.970 0.552 3.000 5.927 6.552 4
9 0.816 0.022 0.816 0.022 1.000 0.000 0.000 0.000 4.270 0.410 3.180 4.000 6.763 7.180 3
10 0.842 0.014 0.842 0.014 1.000 0.000 0.000 0.000 0.000 0.000 1.616 5.000 3.407 3.616 1
11 0.781 0.011 0.781 0.011 1.000 0.000 0.000 0.000 0.000 0.000 8.499 5.000 10.291 10.499 1