Atom selection language¶
AtomGroups can be created by selecting atoms using the MDAnalysis atom selection language:
In : import MDAnalysis as mda In : from MDAnalysis.tests.datafiles import PSF, DCD In : u = mda.Universe(PSF, DCD) In : ala = u.select_atoms('resname ALA') In : ala Out: <AtomGroup with 190 atoms>
select_atoms() method of a
AtomGroup or a
Universe returns an
AtomGroup. These two methods have different behaviour: while
Universe.select_atoms operates on all the atoms in the universe,
AtomGroup.select_atoms only operates on the atoms within the original AtomGroup. A single selection phrase always returns an
AtomGroup with atoms sorted according to their
index in the topology. This is to ensure that there are not any duplicates,
which can happen with complicated selections. When order matters, you can pass in multiple phrases.
This page documents selection keywords and their arguments.
select_atoms() also accepts keywords that modify the behaviour of the selection string and the resulting
AtomGroup (documented further down this page). For example, you can:
Pass in named AtomGroups as arguments:
In : sph_6 = u.select_atoms("sphzone 6 protein") In : u.select_atoms("around 3 group sph_6", sph_6=sph_6) Out: <AtomGroup with 81 atoms>
Turn off periodic boundary conditions for geometric keywords with
In : u.select_atoms("around 6 protein", periodic=False) Out: <AtomGroup with 0 atoms>
Create dynamic UpdatingAtomGroups with
In : u.select_atoms("prop x < 5 and prop y < 5 and prop z < 5", updating=True) Out: <AtomGroup with 917 atoms, with selection 'prop x < 5 and prop y < 5 and prop z < 5' on the entire Universe.>
It is possible to export selections for external software packages with the help of Selection exporters.
The following describes all selection keywords currently understood by the selection parser. The following applies to all selections:
Keywords are case sensitive.
Atoms are automatically sequentially ordered in a resulting selection (see notes below on Ordered selections for how to circumvent this if necessary).
Selections are parsed left to right and parentheses can be used for grouping. For example:
In : u.select_atoms("segid DMPC and not (name H* or type OW)") Out: <AtomGroup with 0 atoms>
Currently, wildcards are implemented as a form of pattern matching: Using the
*character in a string such as
GL*selects all strings that start with “GL” such as “GLU”, “GLY”, “GLX29”, “GLN”. Only terminal wildcards (i.e. matching the last part of a name) are currently supported.
Up until version 1.0.0, MDAnalysis will ignore everything after the
u.select_atoms("resname *E") will not select atoms whose residue name ends in E, but instead select every atom.
Selects atoms that belong to a hard-coded set of standard protein residue names.
Selects the backbone atoms of a hard-coded set of protein residues. These atoms have the names: CA, C, O, N.
Selects atoms that belong to a hard-coded set of standard nucleic residue names.
Selects the backbone atoms of a hard-coded set of nucleic residues. These atoms have the names: P, O5’, C5’, C3’, O3’
Selects the atoms in nucleobases.
Selects the atoms in nucleic sugars. These have the names: C1’, C2’, C3’, C4’, O2’, O4’, O3’
- segid seg-name
select by segid (as given in the topology), e.g.
- resid residue-number-range
residcan take a single residue number or a range of numbers, followed by insertion codes. A range consists of two selections separated by a colon (inclusive) such as
resid 1A:1C. This selects all residues with
icode in ('A', 'B', 'C'). A residue number (“resid”) and icode is taken directly from the topology. Unlike
residis sensitive to insertion codes.
- resnum residue-number-range
resnumcan take a single residue number or a range of numbers. A range consists of two numbers separated by a colon (inclusive) such as
resnum 1:5. A residue number (“resnum”) is taken directly from the topology. Unlike
resnumis insensitive to insertion codes.
- resname residue-name
select by residue name, e.g.
- name atom-name
select by atom name (as given in the topology). Often, this is force field dependent. Example:
name CA(for C-alpha atoms) or
name OW(for SPC water oxygen)
- type atom-type
select by atom type; this is either a string or a number and depends on the force field; it is read from the topology file (e.g. the CHARMM PSF file contains numeric atom types). This uses the
- atom seg-name residue-number atom-name
a selector for a single atom consisting of segid resid atomname, e.g.
DMPC 1 C2selects the C2 carbon of the first residue of the DMPC segment
- altloc alternative-location
a selection for atoms where alternative locations are available, which is often the case with high-resolution crystal structures e.g.
resid 4 and resname ALA and altloc Bselects only the atoms of ALA-4 that have an altloc B record.
- moltype molecule-type
select by the
moltypetopology attribute, e.g.
moltype Protein_A. At the moment, only the TPR format defines the
all atoms not in the selection, e.g.
not proteinselects all atoms that aren’t part of a protein
the intersection of two selections, i.e. the boolean and. e.g.
protein and not resname ALAselects all atoms that belong to a protein but are not in an alanine residue
the union of two selections, i.e. the boolean or. e.g.
protein and not (resname ALA or resname LYS)selects all atoms that belong to a protein, but are not in a lysine or alanine residue
The geometric keywords below all implement periodic boundary conditions by default when valid cell dimensions are accessible from the Universe. This can be turned off by passing in the keyword
In : u.select_atoms("around 6 protein", periodic=False) Out: <AtomGroup with 0 atoms>
- around distance selection
selects all atoms a certain cutoff away from another selection, e.g.
around 3.5 proteinselects all atoms not belonging to protein that are within 3.5 Angstroms from the protein
- sphzone externalRadius selection
selects all atoms within a spherical zone centered in the center of geometry (COG) of a given selection, e.g.
sphzone 6.0 ( protein and ( resid 130 or resid 80 ) )selects the center of geometry of protein, resid 130, resid 80 and creates a sphere of radius 6.0 around the COG.
- sphlayer innerRadius externalRadius selection
selects all atoms within a spherical layer centered in the center of geometry (COG) of a given selection, e.g.,
sphlayer 2.4 6.0 ( protein and ( resid 130 or resid 80 ) )selects the center of geometry of protein, resid 130, resid 80 and creates a spherical layer of inner radius 2.4 and external radius 6.0 around the COG.
- cyzone externalRadius zMax zMin selection
selects all atoms within a cylindric zone centered in the center of geometry (COG) of a given selection, e.g.
cyzone 15 4 -8 protein and resid 42selects the center of geometry of protein and resid 42, and creates a cylinder of external radius 15 centered on the COG. In z, the cylinder extends from 4 above the COG to 8 below. Positive values for zMin, or negative ones for zMax, are allowed.
- cylayer innerRadius externalRadius zMax zMin selection
selects all atoms within a cylindric layer centered in the center of geometry (COG) of a given selection, e.g.
cylayer 5 10 10 -8 proteinselects the center of geometry of protein, and creates a cylindrical layer of inner radius 5, external radius 10 centered on the COG. In z, the cylinder extends from 10 above the COG to 8 below. Positive values for zMin, or negative ones for zMax, are allowed.
- point x y z distance
selects all atoms within a cutoff of a point in space, make sure coordinate is separated by spaces, e.g.
point 5.0 5.0 5.0 3.5selects all atoms within 3.5 Angstroms of the coordinate (5.0, 5.0, 5.0)
- prop [abs] property operator value
selects atoms based on position, using property x, y, or z coordinate. Supports the abs keyword (for absolute value) and the following operators: <, >, <=, >=, ==, !=. For example,
prop z >= 5.0selects all atoms with z coordinate greater than 5.0;
prop abs z <= 5.0selects all atoms within -5.0 <= z <= 5.0.
Similarity and connectivity¶
- same subkeyword as selection
selects all atoms that have the same subkeyword value as any atom in selection. Allowed subkeyword values are the atom properties:
name, type, resname, resid, resnum, segid, mass, charge, radius, bfactor, the groups an atom belong to:
residue, segment, fragment, and the atom coordinates
x, y, z. (Note that
bfactorcurrently only works for MMTF formats.) e.g.
same charge as proteinselects all atoms that have the same charge as any atom in protein.
- byres selection
selects all atoms that are in the same segment and residue as selection, e.g. specify the subselection after the byres keyword.
byresis a shortcut to
same residue as
- bonded selection
selects all atoms that are bonded to selection e.g.:
name H and bonded name Nselects only hydrogens bonded to nitrogens
- index index-range
selects all atoms within a range of (0-based) inclusive indices, e.g.
index 0selects the first atom in the universe;
index 5:10selects the 6th through 11th atoms, inclusive. This uses the
- bynum number-range
selects all atoms within a range of (1-based) inclusive indices, e.g.
bynum 1selects the first atom in the universe;
bynum 5:10selects 5th through 10th atoms, inclusive.
These are not the same as the 1-indexed
bynumsimply adds 1 to the 0-indexed
Preexisting selections and modifiers¶
- group group-name
selects the atoms in the
AtomGrouppassed to the function as an argument named group-name. Only the atoms common to group-name and the instance
select_atoms()was called from will be considered, unless
groupis preceded by the
globalkeyword. group-name will be included in the parsing just by comparison of atom indices. This means that it is up to the user to make sure the group-name group was defined in an appropriate
- global selection
by default, when issuing
AtomGroup, selections and subselections are returned intersected with the atoms of that instance. Prefixing a selection term with
globalcauses its selection to be returned in its entirety. As an example, the
globalkeyword allows for
lipids.select_atoms("around 10 global protein")— where
lipidsis a group that does not contain any proteins. Were
globalabsent, the result would be an empty selection since the
proteinsubselection would itself be empty. When calling
select_atoms() returns an
AtomGroup, in which the list of atoms is
constant across trajectory frame changes. If
select_atoms() is invoked with named
updating set to
UpdatingAtomGroup instance will be returned
# A dynamic selection of corner atoms: In : ag_updating = u.select_atoms("prop x < 5 and prop y < 5 and prop z < 5", updating=True) In : ag_updating Out: <AtomGroup with 917 atoms, with selection 'prop x < 5 and prop y < 5 and prop z < 5' on the entire Universe.>
It behaves just like an
object, with the difference that the selection expressions are re-evaluated
every time the trajectory frame changes (this happens lazily, only when the
UpdatingAtomGroup object is accessed so that
there is no redundant updating going on):
In : u.trajectory.next() Out: < Timestep 1 with unit cell dimensions [ 0. 0. 0. 90. 90. 90.] > In : ag_updating Out: <AtomGroup with 923 atoms, with selection 'prop x < 5 and prop y < 5 and prop z < 5' on the entire Universe.>
group selection keyword for
Preexisting selections and modifiers, one can
make updating selections depend on
AtomGroup, or even other
Likewise, making an updating selection from an already updating group will
cause later updates to also reflect the updating of the base group:
In : chained_ag_updating = ag_updating.select_atoms("resid 1:1000", updating=True) In : chained_ag_updating Out: <AtomGroup with 923 atoms, with selection 'resid 1:1000' on another AtomGroup.> In : u.trajectory.next() Out: < Timestep 2 with unit cell dimensions [ 0. 0. 0. 90. 90. 90.] > In : chained_ag_updating Out: <AtomGroup with 921 atoms, with selection 'resid 1:1000' on another AtomGroup.>
In : static_ag = ag_updating.select_atoms("resid 1:1000") In : static_ag Out: <AtomGroup with 921 atoms> In : u.trajectory.next() Out: < Timestep 3 with unit cell dimensions [ 0. 0. 0. 90. 90. 90.] > In : static_ag Out: <AtomGroup with 921 atoms>
select_atoms() sorts the atoms
AtomGroup by atom index before
returning them (this is to eliminate possible duplicates in the
selection). If the ordering of atoms is crucial (for instance when
describing angles or dihedrals) or if duplicate atoms are required
then one has to concatenate multiple AtomGroups, which does not sort
The most straightforward way to concatenate two AtomGroups is by using the
In : ordered = u.select_atoms("resid 3 and name CA") + u.select_atoms("resid 2 and name CA") In : list(ordered) Out: [<Atom 46: CA of type 22 of resname ILE, resid 3 and segid 4AKE>, <Atom 22: CA of type 22 of resname ARG, resid 2 and segid 4AKE>]
A shortcut is to provide two or more selections to
select_atoms(), which then
does the concatenation automatically:
In : list(u.select_atoms("resid 3 and name CA", "resid 2 and name CA")) Out: [<Atom 46: CA of type 22 of resname ILE, resid 3 and segid 4AKE>, <Atom 22: CA of type 22 of resname ARG, resid 2 and segid 4AKE>]
Just for comparison to show that a single selection string does not work as one might expect:
In : list(u.select_atoms("(resid 3 or resid 2) and name CA")) Out: [<Atom 22: CA of type 22 of resname ARG, resid 2 and segid 4AKE>, <Atom 46: CA of type 22 of resname ILE, resid 3 and segid 4AKE>]