Detection and extraction of key
structural regions (patterns)
Lukáš Pravda
@WebchemTools
ncbr.muni.cz/webchemistry
2
Outline
 Variety of structural patterns
 How can we find them?
 Software tools
 PatternQuery
 Channels/tunnels/pores
 Identification of channels and their properties
 Software tools
 MOLE
 Practical session with examples
Variety of structural patterns
Lukáš Pravda
@WebchemTools
ncbr.muni.cz/webchemistry
4
Variety of structural patterns
5
Detection
 The goal is to identify and possibly extract
biological regions of ones interest within
biomolecular structure.
 Including but not limited to:
 Active/binding/interaction sites
 Sequences of amino acids or nucleic acids
 Pockets/channels or void.
 Super secondary motifs.
6
 Database wide detection enables us to carry
out experiments which not has been feasible
before.
 Output of these searches are often an input
for further analyses:
 Structural and functional assignment of newly
determined structures.
 Comparative analyses
 Design and engineering of novel functional sites
 Study of binding modes of certain atoms/residues
OK, but wait why do we need them?
7
Software tools
 A plethora of different software tools – these
are usually a single purpose:
 Detection of ligands
 Binding site identification
 Pockets/cavities
 Channels
 In house scripts and tools
 The question is, can we do any better?
8
PatternQuery
 Web-based application designed for
detection and extraction of molecular
(sub)structures - patterns of user interest.
 Uses unique python like query language to
define composition, topology and
connectivity of these patterns.
 Allows querying single structures as well as the
entire PDB or its subset based on a number of
criteria (organism of origin, resolution, date of
release, …)
9
How does it look like?
http://ncbr.muni.cz/PatternQuery
10
PatternQuery – Structure of language
 Generator queries
 Atoms(), Residues(), RegularMotifs()
 Modifier queries
 ConnectedResidues(), AmbientAtoms(), Filter()
 Combinatory queries
 Or(), Near(), Cluster()
 So far some 50 different queries, which can
be readily used!
11
PatternQuery – Thinking in queries
 Find binding pocket of all ligands in the
protein structure (distance <= 4Å)
12
PatternQuery – Thinking in queries
 Find binding pocket of all ligands in the
protein structure (distance <= 4Å)
13
Build a query I
Atoms("Ca"). AmbientResidues(4)
Atoms("Ca")
14
Build a query II
Atoms("Ca")
. AmbientResidues(4)
. Filter(lambda l: l.Count(Atoms() > 6))
15
 Post-translationaly modified aminoacids
 Het atoms not covalently bound to protein
 Residues with a sugar moiety
Biologically interesting queries I.
HetResidues()
. Filter(lambda l: l.IsNotConnectedTo(AminoAcids()))
ModifiedResidues()NotAminoAcids()
. Filter(lambda l: HetResidues() == 6))
Or(Rings(4 * ["C"] + ["O"]).ConnectedResidues(0),
Rings(4 * ["C"] + ["O"]).ConnectedResidues(0))
16
 PA Lec-B sugar binding site
Biologically interesting queries II.
Near(4, Atoms("Ca"), Atoms("Ca"))
.AmbientResidues(3)
.Filter(lambda l:
l.Count(Or(Rings(5 * ["C"] + ["O"]), Rings(4 * ["C"] + ["O"]))) > 0)
.Filter(lambda l: l.Count(Atoms("P")) == 0)
Questions?
OK, let’s move ON!
@WebchemTools
ncbr.muni.cz/webchemistry
Channels
Lukáš Pravda
@WebchemTools
ncbr.muni.cz/webchemistry
19
Protein empty voids
20
What are the channels/tunnels?
 A type of protein empty void.
 Connects active/binding site
with the bulk solvent.
 Spans through membrane
 They greatly influence
protein specificity, selectivity
and rate of chemical
processes.
 They look pretty(-ish )
21
How can we find them?
 Over the time a number of approaches has been
developed.
 Presently the most successful one relies on
Delaunay Triangulation and Dijkstra’s algorithm.
 Other approaches involves:
 Grid search
 Slice and optimization algorithms
 Sphere-filling methods
22
Software tools
 MOLE
 CAVER
 MolAxis
 ChexVis
 BetaVoid
 HOLE
 And others…
23
Use case – aquaporin 0
 Large family of proteins permitting permeation of
various molecules – mainly water.
 Channel is a tight fit for water molecules.
 How can water permeate through the channel,
while protons don’t?
 ar/R region in blue
24
Use case – bunyavirus
 Negative-strand RNA viruses are serious human
pathogens (Crimean-congo fever, Lassa fever,
influenza).
 How one can kill a virus?
 Design a channel inhibitor!
25
Physicochemical properties – channel duality
26
MOLE computation
 Input: Protein structure + set of parameters
 Output: Channel profile, properties and lining r.
27
Result analysis - properties
 Channel length vs channel radius
 Check presence of bottlenecks and local
narrowings.
 Channel flexibility
28
Result analysis - properties
 Hydropathy, polarity, mutability, formal charge
 Evaluate independent layers as well as entire
channel.
29
Where are my channels - I?
 Q: No channel has been identified  Why?
 A: Wrong set up of ProbeRadius, InteriorThreshold or
Filtering criteria.
 A: Substrate is blocking channel
 Cyclooxygenase-2 (PDB ID: 4cox) complexed with non-selective inhibitor indomethacin
30
Where are my channels - II?
 Q: No channel has been identified  Why?
 A: Active site is located on the surface on its vicinity
 Pocket-like channel found in tyrosine kinase EPh4 (PDB ID 2vwx)
31
Where are my channels - III?
 Q: No channel has been identified  Why?
 A: No channel is there whatsoever
32
Where are my channels - IV?
 Q: None of the found channels is relevant to me?
 A: Multiple reasons. Usually wrong set up of
ProbeRadius or InteriorThreshold parameters
Questions?
After a break we can continue with the hands-on
experience!
https://goo.gl/f5YrcE
@WebchemTools
ncbr.muni.cz/webchemistry